You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
15919 lines
482 KiB
15919 lines
482 KiB
/** |
|
* JSXTransformer v0.13.2 |
|
*/ |
|
(function(f){if(typeof exports==="object"&&typeof module!=="undefined"){module.exports=f()}else if(typeof define==="function"&&define.amd){define([],f)}else{var g;if(typeof window!=="undefined"){g=window}else if(typeof global!=="undefined"){g=global}else if(typeof self!=="undefined"){g=self}else{g=this}g.JSXTransformer = f()}})(function(){var define,module,exports;return (function e(t,n,r){function s(o,u){if(!n[o]){if(!t[o]){var a=typeof require=="function"&&require;if(!u&&a)return a(o,!0);if(i)return i(o,!0);var f=new Error("Cannot find module '"+o+"'");throw f.code="MODULE_NOT_FOUND",f}var l=n[o]={exports:{}};t[o][0].call(l.exports,function(e){var n=t[o][1][e];return s(n?n:e)},l,l.exports,e,t,n,r)}return n[o].exports}var i=typeof require=="function"&&require;for(var o=0;o<r.length;o++)s(r[o]);return s})({1:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013-2015, Facebook, Inc. |
|
* All rights reserved. |
|
* |
|
* This source code is licensed under the BSD-style license found in the |
|
* LICENSE file in the root directory of this source tree. An additional grant |
|
* of patent rights can be found in the PATENTS file in the same directory. |
|
*/ |
|
/* jshint browser: true */ |
|
/* jslint evil: true */ |
|
/*eslint-disable no-eval */ |
|
/*eslint-disable block-scoped-var */ |
|
|
|
'use strict'; |
|
|
|
var ReactTools = _dereq_('../main'); |
|
var inlineSourceMap = _dereq_('./inline-source-map'); |
|
|
|
var headEl; |
|
var dummyAnchor; |
|
var inlineScriptCount = 0; |
|
|
|
// The source-map library relies on Object.defineProperty, but IE8 doesn't |
|
// support it fully even with es5-sham. Indeed, es5-sham's defineProperty |
|
// throws when Object.prototype.__defineGetter__ is missing, so we skip building |
|
// the source map in that case. |
|
var supportsAccessors = Object.prototype.hasOwnProperty('__defineGetter__'); |
|
|
|
/** |
|
* Run provided code through jstransform. |
|
* |
|
* @param {string} source Original source code |
|
* @param {object?} options Options to pass to jstransform |
|
* @return {object} object as returned from jstransform |
|
*/ |
|
function transformReact(source, options) { |
|
options = options || {}; |
|
|
|
// Force the sourcemaps option manually. We don't want to use it if it will |
|
// break (see above note about supportsAccessors). We'll only override the |
|
// value here if sourceMap was specified and is truthy. This guarantees that |
|
// we won't override any user intent (since this method is exposed publicly). |
|
if (options.sourceMap) { |
|
options.sourceMap = supportsAccessors; |
|
} |
|
|
|
// Otherwise just pass all options straight through to react-tools. |
|
return ReactTools.transformWithDetails(source, options); |
|
} |
|
|
|
/** |
|
* Eval provided source after transforming it. |
|
* |
|
* @param {string} source Original source code |
|
* @param {object?} options Options to pass to jstransform |
|
*/ |
|
function exec(source, options) { |
|
return eval(transformReact(source, options).code); |
|
} |
|
|
|
/** |
|
* This method returns a nicely formated line of code pointing to the exact |
|
* location of the error `e`. The line is limited in size so big lines of code |
|
* are also shown in a readable way. |
|
* |
|
* Example: |
|
* ... x', overflow:'scroll'}} id={} onScroll={this.scroll} class=" ... |
|
* ^ |
|
* |
|
* @param {string} code The full string of code |
|
* @param {Error} e The error being thrown |
|
* @return {string} formatted message |
|
* @internal |
|
*/ |
|
function createSourceCodeErrorMessage(code, e) { |
|
var sourceLines = code.split('\n'); |
|
// e.lineNumber is non-standard so we can't depend on its availability. If |
|
// we're in a browser where it isn't supported, don't even bother trying to |
|
// format anything. We may also hit a case where the line number is reported |
|
// incorrectly and is outside the bounds of the actual code. Handle that too. |
|
if (!e.lineNumber || e.lineNumber > sourceLines.length) { |
|
return ''; |
|
} |
|
var erroneousLine = sourceLines[e.lineNumber - 1]; |
|
|
|
// Removes any leading indenting spaces and gets the number of |
|
// chars indenting the `erroneousLine` |
|
var indentation = 0; |
|
erroneousLine = erroneousLine.replace(/^\s+/, function(leadingSpaces) { |
|
indentation = leadingSpaces.length; |
|
return ''; |
|
}); |
|
|
|
// Defines the number of characters that are going to show |
|
// before and after the erroneous code |
|
var LIMIT = 30; |
|
var errorColumn = e.column - indentation; |
|
|
|
if (errorColumn > LIMIT) { |
|
erroneousLine = '... ' + erroneousLine.slice(errorColumn - LIMIT); |
|
errorColumn = 4 + LIMIT; |
|
} |
|
if (erroneousLine.length - errorColumn > LIMIT) { |
|
erroneousLine = erroneousLine.slice(0, errorColumn + LIMIT) + ' ...'; |
|
} |
|
var message = '\n\n' + erroneousLine + '\n'; |
|
message += new Array(errorColumn - 1).join(' ') + '^'; |
|
return message; |
|
} |
|
|
|
/** |
|
* Actually transform the code. |
|
* |
|
* @param {string} code |
|
* @param {string?} url |
|
* @param {object?} options |
|
* @return {string} The transformed code. |
|
* @internal |
|
*/ |
|
function transformCode(code, url, options) { |
|
try { |
|
var transformed = transformReact(code, options); |
|
} catch(e) { |
|
e.message += '\n at '; |
|
if (url) { |
|
if ('fileName' in e) { |
|
// We set `fileName` if it's supported by this error object and |
|
// a `url` was provided. |
|
// The error will correctly point to `url` in Firefox. |
|
e.fileName = url; |
|
} |
|
e.message += url + ':' + e.lineNumber + ':' + e.columnNumber; |
|
} else { |
|
e.message += location.href; |
|
} |
|
e.message += createSourceCodeErrorMessage(code, e); |
|
throw e; |
|
} |
|
|
|
if (!transformed.sourceMap) { |
|
return transformed.code; |
|
} |
|
|
|
var source; |
|
if (url == null) { |
|
source = 'Inline JSX script'; |
|
inlineScriptCount++; |
|
if (inlineScriptCount > 1) { |
|
source += ' (' + inlineScriptCount + ')'; |
|
} |
|
} else if (dummyAnchor) { |
|
// Firefox has problems when the sourcemap source is a proper URL with a |
|
// protocol and hostname, so use the pathname. We could use just the |
|
// filename, but hopefully using the full path will prevent potential |
|
// issues where the same filename exists in multiple directories. |
|
dummyAnchor.href = url; |
|
source = dummyAnchor.pathname.substr(1); |
|
} |
|
|
|
return ( |
|
transformed.code + |
|
'\n' + |
|
inlineSourceMap(transformed.sourceMap, code, source) |
|
); |
|
} |
|
|
|
|
|
/** |
|
* Appends a script element at the end of the <head> with the content of code, |
|
* after transforming it. |
|
* |
|
* @param {string} code The original source code |
|
* @param {string?} url Where the code came from. null if inline |
|
* @param {object?} options Options to pass to jstransform |
|
* @internal |
|
*/ |
|
function run(code, url, options) { |
|
var scriptEl = document.createElement('script'); |
|
scriptEl.text = transformCode(code, url, options); |
|
headEl.appendChild(scriptEl); |
|
} |
|
|
|
/** |
|
* Load script from the provided url and pass the content to the callback. |
|
* |
|
* @param {string} url The location of the script src |
|
* @param {function} callback Function to call with the content of url |
|
* @internal |
|
*/ |
|
function load(url, successCallback, errorCallback) { |
|
var xhr; |
|
xhr = window.ActiveXObject ? new window.ActiveXObject('Microsoft.XMLHTTP') |
|
: new XMLHttpRequest(); |
|
|
|
// async, however scripts will be executed in the order they are in the |
|
// DOM to mirror normal script loading. |
|
xhr.open('GET', url, true); |
|
if ('overrideMimeType' in xhr) { |
|
xhr.overrideMimeType('text/plain'); |
|
} |
|
xhr.onreadystatechange = function() { |
|
if (xhr.readyState === 4) { |
|
if (xhr.status === 0 || xhr.status === 200) { |
|
successCallback(xhr.responseText); |
|
} else { |
|
errorCallback(); |
|
throw new Error('Could not load ' + url); |
|
} |
|
} |
|
}; |
|
return xhr.send(null); |
|
} |
|
|
|
/** |
|
* Loop over provided script tags and get the content, via innerHTML if an |
|
* inline script, or by using XHR. Transforms are applied if needed. The scripts |
|
* are executed in the order they are found on the page. |
|
* |
|
* @param {array} scripts The <script> elements to load and run. |
|
* @internal |
|
*/ |
|
function loadScripts(scripts) { |
|
var result = []; |
|
var count = scripts.length; |
|
|
|
function check() { |
|
var script, i; |
|
|
|
for (i = 0; i < count; i++) { |
|
script = result[i]; |
|
|
|
if (script.loaded && !script.executed) { |
|
script.executed = true; |
|
run(script.content, script.url, script.options); |
|
} else if (!script.loaded && !script.error && !script.async) { |
|
break; |
|
} |
|
} |
|
} |
|
|
|
scripts.forEach(function(script, i) { |
|
var options = { |
|
sourceMap: true |
|
}; |
|
if (/;harmony=true(;|$)/.test(script.type)) { |
|
options.harmony = true; |
|
} |
|
if (/;stripTypes=true(;|$)/.test(script.type)) { |
|
options.stripTypes = true; |
|
} |
|
|
|
// script.async is always true for non-javascript script tags |
|
var async = script.hasAttribute('async'); |
|
|
|
if (script.src) { |
|
result[i] = { |
|
async: async, |
|
error: false, |
|
executed: false, |
|
content: null, |
|
loaded: false, |
|
url: script.src, |
|
options: options |
|
}; |
|
|
|
load(script.src, function(content) { |
|
result[i].loaded = true; |
|
result[i].content = content; |
|
check(); |
|
}, function() { |
|
result[i].error = true; |
|
check(); |
|
}); |
|
} else { |
|
result[i] = { |
|
async: async, |
|
error: false, |
|
executed: false, |
|
content: script.innerHTML, |
|
loaded: true, |
|
url: null, |
|
options: options |
|
}; |
|
} |
|
}); |
|
|
|
check(); |
|
} |
|
|
|
/** |
|
* Find and run all script tags with type="text/jsx". |
|
* |
|
* @internal |
|
*/ |
|
function runScripts() { |
|
var scripts = document.getElementsByTagName('script'); |
|
|
|
// Array.prototype.slice cannot be used on NodeList on IE8 |
|
var jsxScripts = []; |
|
for (var i = 0; i < scripts.length; i++) { |
|
if (/^text\/jsx(;|$)/.test(scripts.item(i).type)) { |
|
jsxScripts.push(scripts.item(i)); |
|
} |
|
} |
|
|
|
if (jsxScripts.length < 1) { |
|
return; |
|
} |
|
|
|
console.warn( |
|
'You are using the in-browser JSX transformer. Be sure to precompile ' + |
|
'your JSX for production - ' + |
|
'http://facebook.github.io/react/docs/tooling-integration.html#jsx' |
|
); |
|
|
|
loadScripts(jsxScripts); |
|
} |
|
|
|
// Listen for load event if we're in a browser and then kick off finding and |
|
// running of scripts. |
|
if (typeof window !== 'undefined' && window !== null) { |
|
headEl = document.getElementsByTagName('head')[0]; |
|
dummyAnchor = document.createElement('a'); |
|
|
|
if (window.addEventListener) { |
|
window.addEventListener('DOMContentLoaded', runScripts, false); |
|
} else { |
|
window.attachEvent('onload', runScripts); |
|
} |
|
} |
|
|
|
module.exports = { |
|
transform: transformReact, |
|
exec: exec |
|
}; |
|
|
|
},{"../main":2,"./inline-source-map":41}],2:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013-2015, Facebook, Inc. |
|
* All rights reserved. |
|
* |
|
* This source code is licensed under the BSD-style license found in the |
|
* LICENSE file in the root directory of this source tree. An additional grant |
|
* of patent rights can be found in the PATENTS file in the same directory. |
|
*/ |
|
|
|
'use strict'; |
|
/*eslint-disable no-undef*/ |
|
var visitors = _dereq_('./vendor/fbtransform/visitors'); |
|
var transform = _dereq_('jstransform').transform; |
|
var typesSyntax = _dereq_('jstransform/visitors/type-syntax'); |
|
var inlineSourceMap = _dereq_('./vendor/inline-source-map'); |
|
|
|
module.exports = { |
|
transform: function(input, options) { |
|
options = processOptions(options); |
|
var output = innerTransform(input, options); |
|
var result = output.code; |
|
if (options.sourceMap) { |
|
var map = inlineSourceMap( |
|
output.sourceMap, |
|
input, |
|
options.filename |
|
); |
|
result += '\n' + map; |
|
} |
|
return result; |
|
}, |
|
transformWithDetails: function(input, options) { |
|
options = processOptions(options); |
|
var output = innerTransform(input, options); |
|
var result = {}; |
|
result.code = output.code; |
|
if (options.sourceMap) { |
|
result.sourceMap = output.sourceMap.toJSON(); |
|
} |
|
if (options.filename) { |
|
result.sourceMap.sources = [options.filename]; |
|
} |
|
return result; |
|
} |
|
}; |
|
|
|
/** |
|
* Only copy the values that we need. We'll do some preprocessing to account for |
|
* converting command line flags to options that jstransform can actually use. |
|
*/ |
|
function processOptions(opts) { |
|
opts = opts || {}; |
|
var options = {}; |
|
|
|
options.harmony = opts.harmony; |
|
options.stripTypes = opts.stripTypes; |
|
options.sourceMap = opts.sourceMap; |
|
options.filename = opts.sourceFilename; |
|
|
|
if (opts.es6module) { |
|
options.sourceType = 'module'; |
|
} |
|
if (opts.nonStrictEs6module) { |
|
options.sourceType = 'nonStrictModule'; |
|
} |
|
|
|
// Instead of doing any fancy validation, only look for 'es3'. If we have |
|
// that, then use it. Otherwise use 'es5'. |
|
options.es3 = opts.target === 'es3'; |
|
options.es5 = !options.es3; |
|
|
|
return options; |
|
} |
|
|
|
function innerTransform(input, options) { |
|
var visitorSets = ['react']; |
|
if (options.harmony) { |
|
visitorSets.push('harmony'); |
|
} |
|
|
|
if (options.es3) { |
|
visitorSets.push('es3'); |
|
} |
|
|
|
if (options.stripTypes) { |
|
// Stripping types needs to happen before the other transforms |
|
// unfortunately, due to bad interactions. For example, |
|
// es6-rest-param-visitors conflict with stripping rest param type |
|
// annotation |
|
input = transform(typesSyntax.visitorList, input, options).code; |
|
} |
|
|
|
var visitorList = visitors.getVisitorsBySet(visitorSets); |
|
return transform(visitorList, input, options); |
|
} |
|
|
|
},{"./vendor/fbtransform/visitors":40,"./vendor/inline-source-map":41,"jstransform":22,"jstransform/visitors/type-syntax":36}],3:[function(_dereq_,module,exports){ |
|
/*! |
|
* The buffer module from node.js, for the browser. |
|
* |
|
* @author Feross Aboukhadijeh <feross@feross.org> <http://feross.org> |
|
* @license MIT |
|
*/ |
|
|
|
var base64 = _dereq_('base64-js') |
|
var ieee754 = _dereq_('ieee754') |
|
var isArray = _dereq_('is-array') |
|
|
|
exports.Buffer = Buffer |
|
exports.SlowBuffer = SlowBuffer |
|
exports.INSPECT_MAX_BYTES = 50 |
|
Buffer.poolSize = 8192 // not used by this implementation |
|
|
|
var kMaxLength = 0x3fffffff |
|
var rootParent = {} |
|
|
|
/** |
|
* If `Buffer.TYPED_ARRAY_SUPPORT`: |
|
* === true Use Uint8Array implementation (fastest) |
|
* === false Use Object implementation (most compatible, even IE6) |
|
* |
|
* Browsers that support typed arrays are IE 10+, Firefox 4+, Chrome 7+, Safari 5.1+, |
|
* Opera 11.6+, iOS 4.2+. |
|
* |
|
* Note: |
|
* |
|
* - Implementation must support adding new properties to `Uint8Array` instances. |
|
* Firefox 4-29 lacked support, fixed in Firefox 30+. |
|
* See: https://bugzilla.mozilla.org/show_bug.cgi?id=695438. |
|
* |
|
* - Chrome 9-10 is missing the `TypedArray.prototype.subarray` function. |
|
* |
|
* - IE10 has a broken `TypedArray.prototype.subarray` function which returns arrays of |
|
* incorrect length in some situations. |
|
* |
|
* We detect these buggy browsers and set `Buffer.TYPED_ARRAY_SUPPORT` to `false` so they will |
|
* get the Object implementation, which is slower but will work correctly. |
|
*/ |
|
Buffer.TYPED_ARRAY_SUPPORT = (function () { |
|
try { |
|
var buf = new ArrayBuffer(0) |
|
var arr = new Uint8Array(buf) |
|
arr.foo = function () { return 42 } |
|
return arr.foo() === 42 && // typed array instances can be augmented |
|
typeof arr.subarray === 'function' && // chrome 9-10 lack `subarray` |
|
new Uint8Array(1).subarray(1, 1).byteLength === 0 // ie10 has broken `subarray` |
|
} catch (e) { |
|
return false |
|
} |
|
})() |
|
|
|
/** |
|
* Class: Buffer |
|
* ============= |
|
* |
|
* The Buffer constructor returns instances of `Uint8Array` that are augmented |
|
* with function properties for all the node `Buffer` API functions. We use |
|
* `Uint8Array` so that square bracket notation works as expected -- it returns |
|
* a single octet. |
|
* |
|
* By augmenting the instances, we can avoid modifying the `Uint8Array` |
|
* prototype. |
|
*/ |
|
function Buffer (subject, encoding) { |
|
var self = this |
|
if (!(self instanceof Buffer)) return new Buffer(subject, encoding) |
|
|
|
var type = typeof subject |
|
var length |
|
|
|
if (type === 'number') { |
|
length = +subject |
|
} else if (type === 'string') { |
|
length = Buffer.byteLength(subject, encoding) |
|
} else if (type === 'object' && subject !== null) { |
|
// assume object is array-like |
|
if (subject.type === 'Buffer' && isArray(subject.data)) subject = subject.data |
|
length = +subject.length |
|
} else { |
|
throw new TypeError('must start with number, buffer, array or string') |
|
} |
|
|
|
if (length > kMaxLength) { |
|
throw new RangeError('Attempt to allocate Buffer larger than maximum size: 0x' + |
|
kMaxLength.toString(16) + ' bytes') |
|
} |
|
|
|
if (length < 0) length = 0 |
|
else length >>>= 0 // coerce to uint32 |
|
|
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
// Preferred: Return an augmented `Uint8Array` instance for best performance |
|
self = Buffer._augment(new Uint8Array(length)) // eslint-disable-line consistent-this |
|
} else { |
|
// Fallback: Return THIS instance of Buffer (created by `new`) |
|
self.length = length |
|
self._isBuffer = true |
|
} |
|
|
|
var i |
|
if (Buffer.TYPED_ARRAY_SUPPORT && typeof subject.byteLength === 'number') { |
|
// Speed optimization -- use set if we're copying from a typed array |
|
self._set(subject) |
|
} else if (isArrayish(subject)) { |
|
// Treat array-ish objects as a byte array |
|
if (Buffer.isBuffer(subject)) { |
|
for (i = 0; i < length; i++) { |
|
self[i] = subject.readUInt8(i) |
|
} |
|
} else { |
|
for (i = 0; i < length; i++) { |
|
self[i] = ((subject[i] % 256) + 256) % 256 |
|
} |
|
} |
|
} else if (type === 'string') { |
|
self.write(subject, 0, encoding) |
|
} else if (type === 'number' && !Buffer.TYPED_ARRAY_SUPPORT) { |
|
for (i = 0; i < length; i++) { |
|
self[i] = 0 |
|
} |
|
} |
|
|
|
if (length > 0 && length <= Buffer.poolSize) self.parent = rootParent |
|
|
|
return self |
|
} |
|
|
|
function SlowBuffer (subject, encoding) { |
|
if (!(this instanceof SlowBuffer)) return new SlowBuffer(subject, encoding) |
|
|
|
var buf = new Buffer(subject, encoding) |
|
delete buf.parent |
|
return buf |
|
} |
|
|
|
Buffer.isBuffer = function isBuffer (b) { |
|
return !!(b != null && b._isBuffer) |
|
} |
|
|
|
Buffer.compare = function compare (a, b) { |
|
if (!Buffer.isBuffer(a) || !Buffer.isBuffer(b)) { |
|
throw new TypeError('Arguments must be Buffers') |
|
} |
|
|
|
if (a === b) return 0 |
|
|
|
var x = a.length |
|
var y = b.length |
|
for (var i = 0, len = Math.min(x, y); i < len && a[i] === b[i]; i++) {} |
|
if (i !== len) { |
|
x = a[i] |
|
y = b[i] |
|
} |
|
if (x < y) return -1 |
|
if (y < x) return 1 |
|
return 0 |
|
} |
|
|
|
Buffer.isEncoding = function isEncoding (encoding) { |
|
switch (String(encoding).toLowerCase()) { |
|
case 'hex': |
|
case 'utf8': |
|
case 'utf-8': |
|
case 'ascii': |
|
case 'binary': |
|
case 'base64': |
|
case 'raw': |
|
case 'ucs2': |
|
case 'ucs-2': |
|
case 'utf16le': |
|
case 'utf-16le': |
|
return true |
|
default: |
|
return false |
|
} |
|
} |
|
|
|
Buffer.concat = function concat (list, totalLength) { |
|
if (!isArray(list)) throw new TypeError('list argument must be an Array of Buffers.') |
|
|
|
if (list.length === 0) { |
|
return new Buffer(0) |
|
} else if (list.length === 1) { |
|
return list[0] |
|
} |
|
|
|
var i |
|
if (totalLength === undefined) { |
|
totalLength = 0 |
|
for (i = 0; i < list.length; i++) { |
|
totalLength += list[i].length |
|
} |
|
} |
|
|
|
var buf = new Buffer(totalLength) |
|
var pos = 0 |
|
for (i = 0; i < list.length; i++) { |
|
var item = list[i] |
|
item.copy(buf, pos) |
|
pos += item.length |
|
} |
|
return buf |
|
} |
|
|
|
Buffer.byteLength = function byteLength (str, encoding) { |
|
var ret |
|
str = str + '' |
|
switch (encoding || 'utf8') { |
|
case 'ascii': |
|
case 'binary': |
|
case 'raw': |
|
ret = str.length |
|
break |
|
case 'ucs2': |
|
case 'ucs-2': |
|
case 'utf16le': |
|
case 'utf-16le': |
|
ret = str.length * 2 |
|
break |
|
case 'hex': |
|
ret = str.length >>> 1 |
|
break |
|
case 'utf8': |
|
case 'utf-8': |
|
ret = utf8ToBytes(str).length |
|
break |
|
case 'base64': |
|
ret = base64ToBytes(str).length |
|
break |
|
default: |
|
ret = str.length |
|
} |
|
return ret |
|
} |
|
|
|
// pre-set for values that may exist in the future |
|
Buffer.prototype.length = undefined |
|
Buffer.prototype.parent = undefined |
|
|
|
// toString(encoding, start=0, end=buffer.length) |
|
Buffer.prototype.toString = function toString (encoding, start, end) { |
|
var loweredCase = false |
|
|
|
start = start >>> 0 |
|
end = end === undefined || end === Infinity ? this.length : end >>> 0 |
|
|
|
if (!encoding) encoding = 'utf8' |
|
if (start < 0) start = 0 |
|
if (end > this.length) end = this.length |
|
if (end <= start) return '' |
|
|
|
while (true) { |
|
switch (encoding) { |
|
case 'hex': |
|
return hexSlice(this, start, end) |
|
|
|
case 'utf8': |
|
case 'utf-8': |
|
return utf8Slice(this, start, end) |
|
|
|
case 'ascii': |
|
return asciiSlice(this, start, end) |
|
|
|
case 'binary': |
|
return binarySlice(this, start, end) |
|
|
|
case 'base64': |
|
return base64Slice(this, start, end) |
|
|
|
case 'ucs2': |
|
case 'ucs-2': |
|
case 'utf16le': |
|
case 'utf-16le': |
|
return utf16leSlice(this, start, end) |
|
|
|
default: |
|
if (loweredCase) throw new TypeError('Unknown encoding: ' + encoding) |
|
encoding = (encoding + '').toLowerCase() |
|
loweredCase = true |
|
} |
|
} |
|
} |
|
|
|
Buffer.prototype.equals = function equals (b) { |
|
if (!Buffer.isBuffer(b)) throw new TypeError('Argument must be a Buffer') |
|
if (this === b) return true |
|
return Buffer.compare(this, b) === 0 |
|
} |
|
|
|
Buffer.prototype.inspect = function inspect () { |
|
var str = '' |
|
var max = exports.INSPECT_MAX_BYTES |
|
if (this.length > 0) { |
|
str = this.toString('hex', 0, max).match(/.{2}/g).join(' ') |
|
if (this.length > max) str += ' ... ' |
|
} |
|
return '<Buffer ' + str + '>' |
|
} |
|
|
|
Buffer.prototype.compare = function compare (b) { |
|
if (!Buffer.isBuffer(b)) throw new TypeError('Argument must be a Buffer') |
|
if (this === b) return 0 |
|
return Buffer.compare(this, b) |
|
} |
|
|
|
Buffer.prototype.indexOf = function indexOf (val, byteOffset) { |
|
if (byteOffset > 0x7fffffff) byteOffset = 0x7fffffff |
|
else if (byteOffset < -0x80000000) byteOffset = -0x80000000 |
|
byteOffset >>= 0 |
|
|
|
if (this.length === 0) return -1 |
|
if (byteOffset >= this.length) return -1 |
|
|
|
// Negative offsets start from the end of the buffer |
|
if (byteOffset < 0) byteOffset = Math.max(this.length + byteOffset, 0) |
|
|
|
if (typeof val === 'string') { |
|
if (val.length === 0) return -1 // special case: looking for empty string always fails |
|
return String.prototype.indexOf.call(this, val, byteOffset) |
|
} |
|
if (Buffer.isBuffer(val)) { |
|
return arrayIndexOf(this, val, byteOffset) |
|
} |
|
if (typeof val === 'number') { |
|
if (Buffer.TYPED_ARRAY_SUPPORT && Uint8Array.prototype.indexOf === 'function') { |
|
return Uint8Array.prototype.indexOf.call(this, val, byteOffset) |
|
} |
|
return arrayIndexOf(this, [ val ], byteOffset) |
|
} |
|
|
|
function arrayIndexOf (arr, val, byteOffset) { |
|
var foundIndex = -1 |
|
for (var i = 0; byteOffset + i < arr.length; i++) { |
|
if (arr[byteOffset + i] === val[foundIndex === -1 ? 0 : i - foundIndex]) { |
|
if (foundIndex === -1) foundIndex = i |
|
if (i - foundIndex + 1 === val.length) return byteOffset + foundIndex |
|
} else { |
|
foundIndex = -1 |
|
} |
|
} |
|
return -1 |
|
} |
|
|
|
throw new TypeError('val must be string, number or Buffer') |
|
} |
|
|
|
// `get` will be removed in Node 0.13+ |
|
Buffer.prototype.get = function get (offset) { |
|
console.log('.get() is deprecated. Access using array indexes instead.') |
|
return this.readUInt8(offset) |
|
} |
|
|
|
// `set` will be removed in Node 0.13+ |
|
Buffer.prototype.set = function set (v, offset) { |
|
console.log('.set() is deprecated. Access using array indexes instead.') |
|
return this.writeUInt8(v, offset) |
|
} |
|
|
|
function hexWrite (buf, string, offset, length) { |
|
offset = Number(offset) || 0 |
|
var remaining = buf.length - offset |
|
if (!length) { |
|
length = remaining |
|
} else { |
|
length = Number(length) |
|
if (length > remaining) { |
|
length = remaining |
|
} |
|
} |
|
|
|
// must be an even number of digits |
|
var strLen = string.length |
|
if (strLen % 2 !== 0) throw new Error('Invalid hex string') |
|
|
|
if (length > strLen / 2) { |
|
length = strLen / 2 |
|
} |
|
for (var i = 0; i < length; i++) { |
|
var parsed = parseInt(string.substr(i * 2, 2), 16) |
|
if (isNaN(parsed)) throw new Error('Invalid hex string') |
|
buf[offset + i] = parsed |
|
} |
|
return i |
|
} |
|
|
|
function utf8Write (buf, string, offset, length) { |
|
var charsWritten = blitBuffer(utf8ToBytes(string, buf.length - offset), buf, offset, length) |
|
return charsWritten |
|
} |
|
|
|
function asciiWrite (buf, string, offset, length) { |
|
var charsWritten = blitBuffer(asciiToBytes(string), buf, offset, length) |
|
return charsWritten |
|
} |
|
|
|
function binaryWrite (buf, string, offset, length) { |
|
return asciiWrite(buf, string, offset, length) |
|
} |
|
|
|
function base64Write (buf, string, offset, length) { |
|
var charsWritten = blitBuffer(base64ToBytes(string), buf, offset, length) |
|
return charsWritten |
|
} |
|
|
|
function utf16leWrite (buf, string, offset, length) { |
|
var charsWritten = blitBuffer(utf16leToBytes(string, buf.length - offset), buf, offset, length) |
|
return charsWritten |
|
} |
|
|
|
Buffer.prototype.write = function write (string, offset, length, encoding) { |
|
// Support both (string, offset, length, encoding) |
|
// and the legacy (string, encoding, offset, length) |
|
if (isFinite(offset)) { |
|
if (!isFinite(length)) { |
|
encoding = length |
|
length = undefined |
|
} |
|
} else { // legacy |
|
var swap = encoding |
|
encoding = offset |
|
offset = length |
|
length = swap |
|
} |
|
|
|
offset = Number(offset) || 0 |
|
|
|
if (length < 0 || offset < 0 || offset > this.length) { |
|
throw new RangeError('attempt to write outside buffer bounds') |
|
} |
|
|
|
var remaining = this.length - offset |
|
if (!length) { |
|
length = remaining |
|
} else { |
|
length = Number(length) |
|
if (length > remaining) { |
|
length = remaining |
|
} |
|
} |
|
encoding = String(encoding || 'utf8').toLowerCase() |
|
|
|
var ret |
|
switch (encoding) { |
|
case 'hex': |
|
ret = hexWrite(this, string, offset, length) |
|
break |
|
case 'utf8': |
|
case 'utf-8': |
|
ret = utf8Write(this, string, offset, length) |
|
break |
|
case 'ascii': |
|
ret = asciiWrite(this, string, offset, length) |
|
break |
|
case 'binary': |
|
ret = binaryWrite(this, string, offset, length) |
|
break |
|
case 'base64': |
|
ret = base64Write(this, string, offset, length) |
|
break |
|
case 'ucs2': |
|
case 'ucs-2': |
|
case 'utf16le': |
|
case 'utf-16le': |
|
ret = utf16leWrite(this, string, offset, length) |
|
break |
|
default: |
|
throw new TypeError('Unknown encoding: ' + encoding) |
|
} |
|
return ret |
|
} |
|
|
|
Buffer.prototype.toJSON = function toJSON () { |
|
return { |
|
type: 'Buffer', |
|
data: Array.prototype.slice.call(this._arr || this, 0) |
|
} |
|
} |
|
|
|
function base64Slice (buf, start, end) { |
|
if (start === 0 && end === buf.length) { |
|
return base64.fromByteArray(buf) |
|
} else { |
|
return base64.fromByteArray(buf.slice(start, end)) |
|
} |
|
} |
|
|
|
function utf8Slice (buf, start, end) { |
|
var res = '' |
|
var tmp = '' |
|
end = Math.min(buf.length, end) |
|
|
|
for (var i = start; i < end; i++) { |
|
if (buf[i] <= 0x7F) { |
|
res += decodeUtf8Char(tmp) + String.fromCharCode(buf[i]) |
|
tmp = '' |
|
} else { |
|
tmp += '%' + buf[i].toString(16) |
|
} |
|
} |
|
|
|
return res + decodeUtf8Char(tmp) |
|
} |
|
|
|
function asciiSlice (buf, start, end) { |
|
var ret = '' |
|
end = Math.min(buf.length, end) |
|
|
|
for (var i = start; i < end; i++) { |
|
ret += String.fromCharCode(buf[i] & 0x7F) |
|
} |
|
return ret |
|
} |
|
|
|
function binarySlice (buf, start, end) { |
|
var ret = '' |
|
end = Math.min(buf.length, end) |
|
|
|
for (var i = start; i < end; i++) { |
|
ret += String.fromCharCode(buf[i]) |
|
} |
|
return ret |
|
} |
|
|
|
function hexSlice (buf, start, end) { |
|
var len = buf.length |
|
|
|
if (!start || start < 0) start = 0 |
|
if (!end || end < 0 || end > len) end = len |
|
|
|
var out = '' |
|
for (var i = start; i < end; i++) { |
|
out += toHex(buf[i]) |
|
} |
|
return out |
|
} |
|
|
|
function utf16leSlice (buf, start, end) { |
|
var bytes = buf.slice(start, end) |
|
var res = '' |
|
for (var i = 0; i < bytes.length; i += 2) { |
|
res += String.fromCharCode(bytes[i] + bytes[i + 1] * 256) |
|
} |
|
return res |
|
} |
|
|
|
Buffer.prototype.slice = function slice (start, end) { |
|
var len = this.length |
|
start = ~~start |
|
end = end === undefined ? len : ~~end |
|
|
|
if (start < 0) { |
|
start += len |
|
if (start < 0) start = 0 |
|
} else if (start > len) { |
|
start = len |
|
} |
|
|
|
if (end < 0) { |
|
end += len |
|
if (end < 0) end = 0 |
|
} else if (end > len) { |
|
end = len |
|
} |
|
|
|
if (end < start) end = start |
|
|
|
var newBuf |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
newBuf = Buffer._augment(this.subarray(start, end)) |
|
} else { |
|
var sliceLen = end - start |
|
newBuf = new Buffer(sliceLen, undefined) |
|
for (var i = 0; i < sliceLen; i++) { |
|
newBuf[i] = this[i + start] |
|
} |
|
} |
|
|
|
if (newBuf.length) newBuf.parent = this.parent || this |
|
|
|
return newBuf |
|
} |
|
|
|
/* |
|
* Need to make sure that buffer isn't trying to write out of bounds. |
|
*/ |
|
function checkOffset (offset, ext, length) { |
|
if ((offset % 1) !== 0 || offset < 0) throw new RangeError('offset is not uint') |
|
if (offset + ext > length) throw new RangeError('Trying to access beyond buffer length') |
|
} |
|
|
|
Buffer.prototype.readUIntLE = function readUIntLE (offset, byteLength, noAssert) { |
|
offset = offset >>> 0 |
|
byteLength = byteLength >>> 0 |
|
if (!noAssert) checkOffset(offset, byteLength, this.length) |
|
|
|
var val = this[offset] |
|
var mul = 1 |
|
var i = 0 |
|
while (++i < byteLength && (mul *= 0x100)) { |
|
val += this[offset + i] * mul |
|
} |
|
|
|
return val |
|
} |
|
|
|
Buffer.prototype.readUIntBE = function readUIntBE (offset, byteLength, noAssert) { |
|
offset = offset >>> 0 |
|
byteLength = byteLength >>> 0 |
|
if (!noAssert) { |
|
checkOffset(offset, byteLength, this.length) |
|
} |
|
|
|
var val = this[offset + --byteLength] |
|
var mul = 1 |
|
while (byteLength > 0 && (mul *= 0x100)) { |
|
val += this[offset + --byteLength] * mul |
|
} |
|
|
|
return val |
|
} |
|
|
|
Buffer.prototype.readUInt8 = function readUInt8 (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 1, this.length) |
|
return this[offset] |
|
} |
|
|
|
Buffer.prototype.readUInt16LE = function readUInt16LE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 2, this.length) |
|
return this[offset] | (this[offset + 1] << 8) |
|
} |
|
|
|
Buffer.prototype.readUInt16BE = function readUInt16BE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 2, this.length) |
|
return (this[offset] << 8) | this[offset + 1] |
|
} |
|
|
|
Buffer.prototype.readUInt32LE = function readUInt32LE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 4, this.length) |
|
|
|
return ((this[offset]) | |
|
(this[offset + 1] << 8) | |
|
(this[offset + 2] << 16)) + |
|
(this[offset + 3] * 0x1000000) |
|
} |
|
|
|
Buffer.prototype.readUInt32BE = function readUInt32BE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 4, this.length) |
|
|
|
return (this[offset] * 0x1000000) + |
|
((this[offset + 1] << 16) | |
|
(this[offset + 2] << 8) | |
|
this[offset + 3]) |
|
} |
|
|
|
Buffer.prototype.readIntLE = function readIntLE (offset, byteLength, noAssert) { |
|
offset = offset >>> 0 |
|
byteLength = byteLength >>> 0 |
|
if (!noAssert) checkOffset(offset, byteLength, this.length) |
|
|
|
var val = this[offset] |
|
var mul = 1 |
|
var i = 0 |
|
while (++i < byteLength && (mul *= 0x100)) { |
|
val += this[offset + i] * mul |
|
} |
|
mul *= 0x80 |
|
|
|
if (val >= mul) val -= Math.pow(2, 8 * byteLength) |
|
|
|
return val |
|
} |
|
|
|
Buffer.prototype.readIntBE = function readIntBE (offset, byteLength, noAssert) { |
|
offset = offset >>> 0 |
|
byteLength = byteLength >>> 0 |
|
if (!noAssert) checkOffset(offset, byteLength, this.length) |
|
|
|
var i = byteLength |
|
var mul = 1 |
|
var val = this[offset + --i] |
|
while (i > 0 && (mul *= 0x100)) { |
|
val += this[offset + --i] * mul |
|
} |
|
mul *= 0x80 |
|
|
|
if (val >= mul) val -= Math.pow(2, 8 * byteLength) |
|
|
|
return val |
|
} |
|
|
|
Buffer.prototype.readInt8 = function readInt8 (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 1, this.length) |
|
if (!(this[offset] & 0x80)) return (this[offset]) |
|
return ((0xff - this[offset] + 1) * -1) |
|
} |
|
|
|
Buffer.prototype.readInt16LE = function readInt16LE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 2, this.length) |
|
var val = this[offset] | (this[offset + 1] << 8) |
|
return (val & 0x8000) ? val | 0xFFFF0000 : val |
|
} |
|
|
|
Buffer.prototype.readInt16BE = function readInt16BE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 2, this.length) |
|
var val = this[offset + 1] | (this[offset] << 8) |
|
return (val & 0x8000) ? val | 0xFFFF0000 : val |
|
} |
|
|
|
Buffer.prototype.readInt32LE = function readInt32LE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 4, this.length) |
|
|
|
return (this[offset]) | |
|
(this[offset + 1] << 8) | |
|
(this[offset + 2] << 16) | |
|
(this[offset + 3] << 24) |
|
} |
|
|
|
Buffer.prototype.readInt32BE = function readInt32BE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 4, this.length) |
|
|
|
return (this[offset] << 24) | |
|
(this[offset + 1] << 16) | |
|
(this[offset + 2] << 8) | |
|
(this[offset + 3]) |
|
} |
|
|
|
Buffer.prototype.readFloatLE = function readFloatLE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 4, this.length) |
|
return ieee754.read(this, offset, true, 23, 4) |
|
} |
|
|
|
Buffer.prototype.readFloatBE = function readFloatBE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 4, this.length) |
|
return ieee754.read(this, offset, false, 23, 4) |
|
} |
|
|
|
Buffer.prototype.readDoubleLE = function readDoubleLE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 8, this.length) |
|
return ieee754.read(this, offset, true, 52, 8) |
|
} |
|
|
|
Buffer.prototype.readDoubleBE = function readDoubleBE (offset, noAssert) { |
|
if (!noAssert) checkOffset(offset, 8, this.length) |
|
return ieee754.read(this, offset, false, 52, 8) |
|
} |
|
|
|
function checkInt (buf, value, offset, ext, max, min) { |
|
if (!Buffer.isBuffer(buf)) throw new TypeError('buffer must be a Buffer instance') |
|
if (value > max || value < min) throw new RangeError('value is out of bounds') |
|
if (offset + ext > buf.length) throw new RangeError('index out of range') |
|
} |
|
|
|
Buffer.prototype.writeUIntLE = function writeUIntLE (value, offset, byteLength, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
byteLength = byteLength >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, byteLength, Math.pow(2, 8 * byteLength), 0) |
|
|
|
var mul = 1 |
|
var i = 0 |
|
this[offset] = value & 0xFF |
|
while (++i < byteLength && (mul *= 0x100)) { |
|
this[offset + i] = (value / mul) >>> 0 & 0xFF |
|
} |
|
|
|
return offset + byteLength |
|
} |
|
|
|
Buffer.prototype.writeUIntBE = function writeUIntBE (value, offset, byteLength, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
byteLength = byteLength >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, byteLength, Math.pow(2, 8 * byteLength), 0) |
|
|
|
var i = byteLength - 1 |
|
var mul = 1 |
|
this[offset + i] = value & 0xFF |
|
while (--i >= 0 && (mul *= 0x100)) { |
|
this[offset + i] = (value / mul) >>> 0 & 0xFF |
|
} |
|
|
|
return offset + byteLength |
|
} |
|
|
|
Buffer.prototype.writeUInt8 = function writeUInt8 (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 1, 0xff, 0) |
|
if (!Buffer.TYPED_ARRAY_SUPPORT) value = Math.floor(value) |
|
this[offset] = value |
|
return offset + 1 |
|
} |
|
|
|
function objectWriteUInt16 (buf, value, offset, littleEndian) { |
|
if (value < 0) value = 0xffff + value + 1 |
|
for (var i = 0, j = Math.min(buf.length - offset, 2); i < j; i++) { |
|
buf[offset + i] = (value & (0xff << (8 * (littleEndian ? i : 1 - i)))) >>> |
|
(littleEndian ? i : 1 - i) * 8 |
|
} |
|
} |
|
|
|
Buffer.prototype.writeUInt16LE = function writeUInt16LE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 2, 0xffff, 0) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = value |
|
this[offset + 1] = (value >>> 8) |
|
} else { |
|
objectWriteUInt16(this, value, offset, true) |
|
} |
|
return offset + 2 |
|
} |
|
|
|
Buffer.prototype.writeUInt16BE = function writeUInt16BE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 2, 0xffff, 0) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = (value >>> 8) |
|
this[offset + 1] = value |
|
} else { |
|
objectWriteUInt16(this, value, offset, false) |
|
} |
|
return offset + 2 |
|
} |
|
|
|
function objectWriteUInt32 (buf, value, offset, littleEndian) { |
|
if (value < 0) value = 0xffffffff + value + 1 |
|
for (var i = 0, j = Math.min(buf.length - offset, 4); i < j; i++) { |
|
buf[offset + i] = (value >>> (littleEndian ? i : 3 - i) * 8) & 0xff |
|
} |
|
} |
|
|
|
Buffer.prototype.writeUInt32LE = function writeUInt32LE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 4, 0xffffffff, 0) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset + 3] = (value >>> 24) |
|
this[offset + 2] = (value >>> 16) |
|
this[offset + 1] = (value >>> 8) |
|
this[offset] = value |
|
} else { |
|
objectWriteUInt32(this, value, offset, true) |
|
} |
|
return offset + 4 |
|
} |
|
|
|
Buffer.prototype.writeUInt32BE = function writeUInt32BE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 4, 0xffffffff, 0) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = (value >>> 24) |
|
this[offset + 1] = (value >>> 16) |
|
this[offset + 2] = (value >>> 8) |
|
this[offset + 3] = value |
|
} else { |
|
objectWriteUInt32(this, value, offset, false) |
|
} |
|
return offset + 4 |
|
} |
|
|
|
Buffer.prototype.writeIntLE = function writeIntLE (value, offset, byteLength, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) { |
|
checkInt( |
|
this, value, offset, byteLength, |
|
Math.pow(2, 8 * byteLength - 1) - 1, |
|
-Math.pow(2, 8 * byteLength - 1) |
|
) |
|
} |
|
|
|
var i = 0 |
|
var mul = 1 |
|
var sub = value < 0 ? 1 : 0 |
|
this[offset] = value & 0xFF |
|
while (++i < byteLength && (mul *= 0x100)) { |
|
this[offset + i] = ((value / mul) >> 0) - sub & 0xFF |
|
} |
|
|
|
return offset + byteLength |
|
} |
|
|
|
Buffer.prototype.writeIntBE = function writeIntBE (value, offset, byteLength, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) { |
|
checkInt( |
|
this, value, offset, byteLength, |
|
Math.pow(2, 8 * byteLength - 1) - 1, |
|
-Math.pow(2, 8 * byteLength - 1) |
|
) |
|
} |
|
|
|
var i = byteLength - 1 |
|
var mul = 1 |
|
var sub = value < 0 ? 1 : 0 |
|
this[offset + i] = value & 0xFF |
|
while (--i >= 0 && (mul *= 0x100)) { |
|
this[offset + i] = ((value / mul) >> 0) - sub & 0xFF |
|
} |
|
|
|
return offset + byteLength |
|
} |
|
|
|
Buffer.prototype.writeInt8 = function writeInt8 (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 1, 0x7f, -0x80) |
|
if (!Buffer.TYPED_ARRAY_SUPPORT) value = Math.floor(value) |
|
if (value < 0) value = 0xff + value + 1 |
|
this[offset] = value |
|
return offset + 1 |
|
} |
|
|
|
Buffer.prototype.writeInt16LE = function writeInt16LE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 2, 0x7fff, -0x8000) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = value |
|
this[offset + 1] = (value >>> 8) |
|
} else { |
|
objectWriteUInt16(this, value, offset, true) |
|
} |
|
return offset + 2 |
|
} |
|
|
|
Buffer.prototype.writeInt16BE = function writeInt16BE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 2, 0x7fff, -0x8000) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = (value >>> 8) |
|
this[offset + 1] = value |
|
} else { |
|
objectWriteUInt16(this, value, offset, false) |
|
} |
|
return offset + 2 |
|
} |
|
|
|
Buffer.prototype.writeInt32LE = function writeInt32LE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 4, 0x7fffffff, -0x80000000) |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = value |
|
this[offset + 1] = (value >>> 8) |
|
this[offset + 2] = (value >>> 16) |
|
this[offset + 3] = (value >>> 24) |
|
} else { |
|
objectWriteUInt32(this, value, offset, true) |
|
} |
|
return offset + 4 |
|
} |
|
|
|
Buffer.prototype.writeInt32BE = function writeInt32BE (value, offset, noAssert) { |
|
value = +value |
|
offset = offset >>> 0 |
|
if (!noAssert) checkInt(this, value, offset, 4, 0x7fffffff, -0x80000000) |
|
if (value < 0) value = 0xffffffff + value + 1 |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
this[offset] = (value >>> 24) |
|
this[offset + 1] = (value >>> 16) |
|
this[offset + 2] = (value >>> 8) |
|
this[offset + 3] = value |
|
} else { |
|
objectWriteUInt32(this, value, offset, false) |
|
} |
|
return offset + 4 |
|
} |
|
|
|
function checkIEEE754 (buf, value, offset, ext, max, min) { |
|
if (value > max || value < min) throw new RangeError('value is out of bounds') |
|
if (offset + ext > buf.length) throw new RangeError('index out of range') |
|
if (offset < 0) throw new RangeError('index out of range') |
|
} |
|
|
|
function writeFloat (buf, value, offset, littleEndian, noAssert) { |
|
if (!noAssert) { |
|
checkIEEE754(buf, value, offset, 4, 3.4028234663852886e+38, -3.4028234663852886e+38) |
|
} |
|
ieee754.write(buf, value, offset, littleEndian, 23, 4) |
|
return offset + 4 |
|
} |
|
|
|
Buffer.prototype.writeFloatLE = function writeFloatLE (value, offset, noAssert) { |
|
return writeFloat(this, value, offset, true, noAssert) |
|
} |
|
|
|
Buffer.prototype.writeFloatBE = function writeFloatBE (value, offset, noAssert) { |
|
return writeFloat(this, value, offset, false, noAssert) |
|
} |
|
|
|
function writeDouble (buf, value, offset, littleEndian, noAssert) { |
|
if (!noAssert) { |
|
checkIEEE754(buf, value, offset, 8, 1.7976931348623157E+308, -1.7976931348623157E+308) |
|
} |
|
ieee754.write(buf, value, offset, littleEndian, 52, 8) |
|
return offset + 8 |
|
} |
|
|
|
Buffer.prototype.writeDoubleLE = function writeDoubleLE (value, offset, noAssert) { |
|
return writeDouble(this, value, offset, true, noAssert) |
|
} |
|
|
|
Buffer.prototype.writeDoubleBE = function writeDoubleBE (value, offset, noAssert) { |
|
return writeDouble(this, value, offset, false, noAssert) |
|
} |
|
|
|
// copy(targetBuffer, targetStart=0, sourceStart=0, sourceEnd=buffer.length) |
|
Buffer.prototype.copy = function copy (target, target_start, start, end) { |
|
if (!start) start = 0 |
|
if (!end && end !== 0) end = this.length |
|
if (target_start >= target.length) target_start = target.length |
|
if (!target_start) target_start = 0 |
|
if (end > 0 && end < start) end = start |
|
|
|
// Copy 0 bytes; we're done |
|
if (end === start) return 0 |
|
if (target.length === 0 || this.length === 0) return 0 |
|
|
|
// Fatal error conditions |
|
if (target_start < 0) { |
|
throw new RangeError('targetStart out of bounds') |
|
} |
|
if (start < 0 || start >= this.length) throw new RangeError('sourceStart out of bounds') |
|
if (end < 0) throw new RangeError('sourceEnd out of bounds') |
|
|
|
// Are we oob? |
|
if (end > this.length) end = this.length |
|
if (target.length - target_start < end - start) { |
|
end = target.length - target_start + start |
|
} |
|
|
|
var len = end - start |
|
|
|
if (len < 1000 || !Buffer.TYPED_ARRAY_SUPPORT) { |
|
for (var i = 0; i < len; i++) { |
|
target[i + target_start] = this[i + start] |
|
} |
|
} else { |
|
target._set(this.subarray(start, start + len), target_start) |
|
} |
|
|
|
return len |
|
} |
|
|
|
// fill(value, start=0, end=buffer.length) |
|
Buffer.prototype.fill = function fill (value, start, end) { |
|
if (!value) value = 0 |
|
if (!start) start = 0 |
|
if (!end) end = this.length |
|
|
|
if (end < start) throw new RangeError('end < start') |
|
|
|
// Fill 0 bytes; we're done |
|
if (end === start) return |
|
if (this.length === 0) return |
|
|
|
if (start < 0 || start >= this.length) throw new RangeError('start out of bounds') |
|
if (end < 0 || end > this.length) throw new RangeError('end out of bounds') |
|
|
|
var i |
|
if (typeof value === 'number') { |
|
for (i = start; i < end; i++) { |
|
this[i] = value |
|
} |
|
} else { |
|
var bytes = utf8ToBytes(value.toString()) |
|
var len = bytes.length |
|
for (i = start; i < end; i++) { |
|
this[i] = bytes[i % len] |
|
} |
|
} |
|
|
|
return this |
|
} |
|
|
|
/** |
|
* Creates a new `ArrayBuffer` with the *copied* memory of the buffer instance. |
|
* Added in Node 0.12. Only available in browsers that support ArrayBuffer. |
|
*/ |
|
Buffer.prototype.toArrayBuffer = function toArrayBuffer () { |
|
if (typeof Uint8Array !== 'undefined') { |
|
if (Buffer.TYPED_ARRAY_SUPPORT) { |
|
return (new Buffer(this)).buffer |
|
} else { |
|
var buf = new Uint8Array(this.length) |
|
for (var i = 0, len = buf.length; i < len; i += 1) { |
|
buf[i] = this[i] |
|
} |
|
return buf.buffer |
|
} |
|
} else { |
|
throw new TypeError('Buffer.toArrayBuffer not supported in this browser') |
|
} |
|
} |
|
|
|
// HELPER FUNCTIONS |
|
// ================ |
|
|
|
var BP = Buffer.prototype |
|
|
|
/** |
|
* Augment a Uint8Array *instance* (not the Uint8Array class!) with Buffer methods |
|
*/ |
|
Buffer._augment = function _augment (arr) { |
|
arr.constructor = Buffer |
|
arr._isBuffer = true |
|
|
|
// save reference to original Uint8Array set method before overwriting |
|
arr._set = arr.set |
|
|
|
// deprecated, will be removed in node 0.13+ |
|
arr.get = BP.get |
|
arr.set = BP.set |
|
|
|
arr.write = BP.write |
|
arr.toString = BP.toString |
|
arr.toLocaleString = BP.toString |
|
arr.toJSON = BP.toJSON |
|
arr.equals = BP.equals |
|
arr.compare = BP.compare |
|
arr.indexOf = BP.indexOf |
|
arr.copy = BP.copy |
|
arr.slice = BP.slice |
|
arr.readUIntLE = BP.readUIntLE |
|
arr.readUIntBE = BP.readUIntBE |
|
arr.readUInt8 = BP.readUInt8 |
|
arr.readUInt16LE = BP.readUInt16LE |
|
arr.readUInt16BE = BP.readUInt16BE |
|
arr.readUInt32LE = BP.readUInt32LE |
|
arr.readUInt32BE = BP.readUInt32BE |
|
arr.readIntLE = BP.readIntLE |
|
arr.readIntBE = BP.readIntBE |
|
arr.readInt8 = BP.readInt8 |
|
arr.readInt16LE = BP.readInt16LE |
|
arr.readInt16BE = BP.readInt16BE |
|
arr.readInt32LE = BP.readInt32LE |
|
arr.readInt32BE = BP.readInt32BE |
|
arr.readFloatLE = BP.readFloatLE |
|
arr.readFloatBE = BP.readFloatBE |
|
arr.readDoubleLE = BP.readDoubleLE |
|
arr.readDoubleBE = BP.readDoubleBE |
|
arr.writeUInt8 = BP.writeUInt8 |
|
arr.writeUIntLE = BP.writeUIntLE |
|
arr.writeUIntBE = BP.writeUIntBE |
|
arr.writeUInt16LE = BP.writeUInt16LE |
|
arr.writeUInt16BE = BP.writeUInt16BE |
|
arr.writeUInt32LE = BP.writeUInt32LE |
|
arr.writeUInt32BE = BP.writeUInt32BE |
|
arr.writeIntLE = BP.writeIntLE |
|
arr.writeIntBE = BP.writeIntBE |
|
arr.writeInt8 = BP.writeInt8 |
|
arr.writeInt16LE = BP.writeInt16LE |
|
arr.writeInt16BE = BP.writeInt16BE |
|
arr.writeInt32LE = BP.writeInt32LE |
|
arr.writeInt32BE = BP.writeInt32BE |
|
arr.writeFloatLE = BP.writeFloatLE |
|
arr.writeFloatBE = BP.writeFloatBE |
|
arr.writeDoubleLE = BP.writeDoubleLE |
|
arr.writeDoubleBE = BP.writeDoubleBE |
|
arr.fill = BP.fill |
|
arr.inspect = BP.inspect |
|
arr.toArrayBuffer = BP.toArrayBuffer |
|
|
|
return arr |
|
} |
|
|
|
var INVALID_BASE64_RE = /[^+\/0-9A-z\-]/g |
|
|
|
function base64clean (str) { |
|
// Node strips out invalid characters like \n and \t from the string, base64-js does not |
|
str = stringtrim(str).replace(INVALID_BASE64_RE, '') |
|
// Node converts strings with length < 2 to '' |
|
if (str.length < 2) return '' |
|
// Node allows for non-padded base64 strings (missing trailing ===), base64-js does not |
|
while (str.length % 4 !== 0) { |
|
str = str + '=' |
|
} |
|
return str |
|
} |
|
|
|
function stringtrim (str) { |
|
if (str.trim) return str.trim() |
|
return str.replace(/^\s+|\s+$/g, '') |
|
} |
|
|
|
function isArrayish (subject) { |
|
return isArray(subject) || Buffer.isBuffer(subject) || |
|
subject && typeof subject === 'object' && |
|
typeof subject.length === 'number' |
|
} |
|
|
|
function toHex (n) { |
|
if (n < 16) return '0' + n.toString(16) |
|
return n.toString(16) |
|
} |
|
|
|
function utf8ToBytes (string, units) { |
|
units = units || Infinity |
|
var codePoint |
|
var length = string.length |
|
var leadSurrogate = null |
|
var bytes = [] |
|
var i = 0 |
|
|
|
for (; i < length; i++) { |
|
codePoint = string.charCodeAt(i) |
|
|
|
// is surrogate component |
|
if (codePoint > 0xD7FF && codePoint < 0xE000) { |
|
// last char was a lead |
|
if (leadSurrogate) { |
|
// 2 leads in a row |
|
if (codePoint < 0xDC00) { |
|
if ((units -= 3) > -1) bytes.push(0xEF, 0xBF, 0xBD) |
|
leadSurrogate = codePoint |
|
continue |
|
} else { |
|
// valid surrogate pair |
|
codePoint = leadSurrogate - 0xD800 << 10 | codePoint - 0xDC00 | 0x10000 |
|
leadSurrogate = null |
|
} |
|
} else { |
|
// no lead yet |
|
|
|
if (codePoint > 0xDBFF) { |
|
// unexpected trail |
|
if ((units -= 3) > -1) bytes.push(0xEF, 0xBF, 0xBD) |
|
continue |
|
} else if (i + 1 === length) { |
|
// unpaired lead |
|
if ((units -= 3) > -1) bytes.push(0xEF, 0xBF, 0xBD) |
|
continue |
|
} else { |
|
// valid lead |
|
leadSurrogate = codePoint |
|
continue |
|
} |
|
} |
|
} else if (leadSurrogate) { |
|
// valid bmp char, but last char was a lead |
|
if ((units -= 3) > -1) bytes.push(0xEF, 0xBF, 0xBD) |
|
leadSurrogate = null |
|
} |
|
|
|
// encode utf8 |
|
if (codePoint < 0x80) { |
|
if ((units -= 1) < 0) break |
|
bytes.push(codePoint) |
|
} else if (codePoint < 0x800) { |
|
if ((units -= 2) < 0) break |
|
bytes.push( |
|
codePoint >> 0x6 | 0xC0, |
|
codePoint & 0x3F | 0x80 |
|
) |
|
} else if (codePoint < 0x10000) { |
|
if ((units -= 3) < 0) break |
|
bytes.push( |
|
codePoint >> 0xC | 0xE0, |
|
codePoint >> 0x6 & 0x3F | 0x80, |
|
codePoint & 0x3F | 0x80 |
|
) |
|
} else if (codePoint < 0x200000) { |
|
if ((units -= 4) < 0) break |
|
bytes.push( |
|
codePoint >> 0x12 | 0xF0, |
|
codePoint >> 0xC & 0x3F | 0x80, |
|
codePoint >> 0x6 & 0x3F | 0x80, |
|
codePoint & 0x3F | 0x80 |
|
) |
|
} else { |
|
throw new Error('Invalid code point') |
|
} |
|
} |
|
|
|
return bytes |
|
} |
|
|
|
function asciiToBytes (str) { |
|
var byteArray = [] |
|
for (var i = 0; i < str.length; i++) { |
|
// Node's code seems to be doing this and not & 0x7F.. |
|
byteArray.push(str.charCodeAt(i) & 0xFF) |
|
} |
|
return byteArray |
|
} |
|
|
|
function utf16leToBytes (str, units) { |
|
var c, hi, lo |
|
var byteArray = [] |
|
for (var i = 0; i < str.length; i++) { |
|
if ((units -= 2) < 0) break |
|
|
|
c = str.charCodeAt(i) |
|
hi = c >> 8 |
|
lo = c % 256 |
|
byteArray.push(lo) |
|
byteArray.push(hi) |
|
} |
|
|
|
return byteArray |
|
} |
|
|
|
function base64ToBytes (str) { |
|
return base64.toByteArray(base64clean(str)) |
|
} |
|
|
|
function blitBuffer (src, dst, offset, length) { |
|
for (var i = 0; i < length; i++) { |
|
if ((i + offset >= dst.length) || (i >= src.length)) break |
|
dst[i + offset] = src[i] |
|
} |
|
return i |
|
} |
|
|
|
function decodeUtf8Char (str) { |
|
try { |
|
return decodeURIComponent(str) |
|
} catch (err) { |
|
return String.fromCharCode(0xFFFD) // UTF 8 invalid char |
|
} |
|
} |
|
|
|
},{"base64-js":4,"ieee754":5,"is-array":6}],4:[function(_dereq_,module,exports){ |
|
var lookup = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/'; |
|
|
|
;(function (exports) { |
|
'use strict'; |
|
|
|
var Arr = (typeof Uint8Array !== 'undefined') |
|
? Uint8Array |
|
: Array |
|
|
|
var PLUS = '+'.charCodeAt(0) |
|
var SLASH = '/'.charCodeAt(0) |
|
var NUMBER = '0'.charCodeAt(0) |
|
var LOWER = 'a'.charCodeAt(0) |
|
var UPPER = 'A'.charCodeAt(0) |
|
var PLUS_URL_SAFE = '-'.charCodeAt(0) |
|
var SLASH_URL_SAFE = '_'.charCodeAt(0) |
|
|
|
function decode (elt) { |
|
var code = elt.charCodeAt(0) |
|
if (code === PLUS || |
|
code === PLUS_URL_SAFE) |
|
return 62 // '+' |
|
if (code === SLASH || |
|
code === SLASH_URL_SAFE) |
|
return 63 // '/' |
|
if (code < NUMBER) |
|
return -1 //no match |
|
if (code < NUMBER + 10) |
|
return code - NUMBER + 26 + 26 |
|
if (code < UPPER + 26) |
|
return code - UPPER |
|
if (code < LOWER + 26) |
|
return code - LOWER + 26 |
|
} |
|
|
|
function b64ToByteArray (b64) { |
|
var i, j, l, tmp, placeHolders, arr |
|
|
|
if (b64.length % 4 > 0) { |
|
throw new Error('Invalid string. Length must be a multiple of 4') |
|
} |
|
|
|
// the number of equal signs (place holders) |
|
// if there are two placeholders, than the two characters before it |
|
// represent one byte |
|
// if there is only one, then the three characters before it represent 2 bytes |
|
// this is just a cheap hack to not do indexOf twice |
|
var len = b64.length |
|
placeHolders = '=' === b64.charAt(len - 2) ? 2 : '=' === b64.charAt(len - 1) ? 1 : 0 |
|
|
|
// base64 is 4/3 + up to two characters of the original data |
|
arr = new Arr(b64.length * 3 / 4 - placeHolders) |
|
|
|
// if there are placeholders, only get up to the last complete 4 chars |
|
l = placeHolders > 0 ? b64.length - 4 : b64.length |
|
|
|
var L = 0 |
|
|
|
function push (v) { |
|
arr[L++] = v |
|
} |
|
|
|
for (i = 0, j = 0; i < l; i += 4, j += 3) { |
|
tmp = (decode(b64.charAt(i)) << 18) | (decode(b64.charAt(i + 1)) << 12) | (decode(b64.charAt(i + 2)) << 6) | decode(b64.charAt(i + 3)) |
|
push((tmp & 0xFF0000) >> 16) |
|
push((tmp & 0xFF00) >> 8) |
|
push(tmp & 0xFF) |
|
} |
|
|
|
if (placeHolders === 2) { |
|
tmp = (decode(b64.charAt(i)) << 2) | (decode(b64.charAt(i + 1)) >> 4) |
|
push(tmp & 0xFF) |
|
} else if (placeHolders === 1) { |
|
tmp = (decode(b64.charAt(i)) << 10) | (decode(b64.charAt(i + 1)) << 4) | (decode(b64.charAt(i + 2)) >> 2) |
|
push((tmp >> 8) & 0xFF) |
|
push(tmp & 0xFF) |
|
} |
|
|
|
return arr |
|
} |
|
|
|
function uint8ToBase64 (uint8) { |
|
var i, |
|
extraBytes = uint8.length % 3, // if we have 1 byte left, pad 2 bytes |
|
output = "", |
|
temp, length |
|
|
|
function encode (num) { |
|
return lookup.charAt(num) |
|
} |
|
|
|
function tripletToBase64 (num) { |
|
return encode(num >> 18 & 0x3F) + encode(num >> 12 & 0x3F) + encode(num >> 6 & 0x3F) + encode(num & 0x3F) |
|
} |
|
|
|
// go through the array every three bytes, we'll deal with trailing stuff later |
|
for (i = 0, length = uint8.length - extraBytes; i < length; i += 3) { |
|
temp = (uint8[i] << 16) + (uint8[i + 1] << 8) + (uint8[i + 2]) |
|
output += tripletToBase64(temp) |
|
} |
|
|
|
// pad the end with zeros, but make sure to not forget the extra bytes |
|
switch (extraBytes) { |
|
case 1: |
|
temp = uint8[uint8.length - 1] |
|
output += encode(temp >> 2) |
|
output += encode((temp << 4) & 0x3F) |
|
output += '==' |
|
break |
|
case 2: |
|
temp = (uint8[uint8.length - 2] << 8) + (uint8[uint8.length - 1]) |
|
output += encode(temp >> 10) |
|
output += encode((temp >> 4) & 0x3F) |
|
output += encode((temp << 2) & 0x3F) |
|
output += '=' |
|
break |
|
} |
|
|
|
return output |
|
} |
|
|
|
exports.toByteArray = b64ToByteArray |
|
exports.fromByteArray = uint8ToBase64 |
|
}(typeof exports === 'undefined' ? (this.base64js = {}) : exports)) |
|
|
|
},{}],5:[function(_dereq_,module,exports){ |
|
exports.read = function(buffer, offset, isLE, mLen, nBytes) { |
|
var e, m, |
|
eLen = nBytes * 8 - mLen - 1, |
|
eMax = (1 << eLen) - 1, |
|
eBias = eMax >> 1, |
|
nBits = -7, |
|
i = isLE ? (nBytes - 1) : 0, |
|
d = isLE ? -1 : 1, |
|
s = buffer[offset + i]; |
|
|
|
i += d; |
|
|
|
e = s & ((1 << (-nBits)) - 1); |
|
s >>= (-nBits); |
|
nBits += eLen; |
|
for (; nBits > 0; e = e * 256 + buffer[offset + i], i += d, nBits -= 8); |
|
|
|
m = e & ((1 << (-nBits)) - 1); |
|
e >>= (-nBits); |
|
nBits += mLen; |
|
for (; nBits > 0; m = m * 256 + buffer[offset + i], i += d, nBits -= 8); |
|
|
|
if (e === 0) { |
|
e = 1 - eBias; |
|
} else if (e === eMax) { |
|
return m ? NaN : ((s ? -1 : 1) * Infinity); |
|
} else { |
|
m = m + Math.pow(2, mLen); |
|
e = e - eBias; |
|
} |
|
return (s ? -1 : 1) * m * Math.pow(2, e - mLen); |
|
}; |
|
|
|
exports.write = function(buffer, value, offset, isLE, mLen, nBytes) { |
|
var e, m, c, |
|
eLen = nBytes * 8 - mLen - 1, |
|
eMax = (1 << eLen) - 1, |
|
eBias = eMax >> 1, |
|
rt = (mLen === 23 ? Math.pow(2, -24) - Math.pow(2, -77) : 0), |
|
i = isLE ? 0 : (nBytes - 1), |
|
d = isLE ? 1 : -1, |
|
s = value < 0 || (value === 0 && 1 / value < 0) ? 1 : 0; |
|
|
|
value = Math.abs(value); |
|
|
|
if (isNaN(value) || value === Infinity) { |
|
m = isNaN(value) ? 1 : 0; |
|
e = eMax; |
|
} else { |
|
e = Math.floor(Math.log(value) / Math.LN2); |
|
if (value * (c = Math.pow(2, -e)) < 1) { |
|
e--; |
|
c *= 2; |
|
} |
|
if (e + eBias >= 1) { |
|
value += rt / c; |
|
} else { |
|
value += rt * Math.pow(2, 1 - eBias); |
|
} |
|
if (value * c >= 2) { |
|
e++; |
|
c /= 2; |
|
} |
|
|
|
if (e + eBias >= eMax) { |
|
m = 0; |
|
e = eMax; |
|
} else if (e + eBias >= 1) { |
|
m = (value * c - 1) * Math.pow(2, mLen); |
|
e = e + eBias; |
|
} else { |
|
m = value * Math.pow(2, eBias - 1) * Math.pow(2, mLen); |
|
e = 0; |
|
} |
|
} |
|
|
|
for (; mLen >= 8; buffer[offset + i] = m & 0xff, i += d, m /= 256, mLen -= 8); |
|
|
|
e = (e << mLen) | m; |
|
eLen += mLen; |
|
for (; eLen > 0; buffer[offset + i] = e & 0xff, i += d, e /= 256, eLen -= 8); |
|
|
|
buffer[offset + i - d] |= s * 128; |
|
}; |
|
|
|
},{}],6:[function(_dereq_,module,exports){ |
|
|
|
/** |
|
* isArray |
|
*/ |
|
|
|
var isArray = Array.isArray; |
|
|
|
/** |
|
* toString |
|
*/ |
|
|
|
var str = Object.prototype.toString; |
|
|
|
/** |
|
* Whether or not the given `val` |
|
* is an array. |
|
* |
|
* example: |
|
* |
|
* isArray([]); |
|
* // > true |
|
* isArray(arguments); |
|
* // > false |
|
* isArray(''); |
|
* // > false |
|
* |
|
* @param {mixed} val |
|
* @return {bool} |
|
*/ |
|
|
|
module.exports = isArray || function (val) { |
|
return !! val && '[object Array]' == str.call(val); |
|
}; |
|
|
|
},{}],7:[function(_dereq_,module,exports){ |
|
(function (process){ |
|
// Copyright Joyent, Inc. and other Node contributors. |
|
// |
|
// Permission is hereby granted, free of charge, to any person obtaining a |
|
// copy of this software and associated documentation files (the |
|
// "Software"), to deal in the Software without restriction, including |
|
// without limitation the rights to use, copy, modify, merge, publish, |
|
// distribute, sublicense, and/or sell copies of the Software, and to permit |
|
// persons to whom the Software is furnished to do so, subject to the |
|
// following conditions: |
|
// |
|
// The above copyright notice and this permission notice shall be included |
|
// in all copies or substantial portions of the Software. |
|
// |
|
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
|
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
|
// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN |
|
// NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, |
|
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR |
|
// OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE |
|
// USE OR OTHER DEALINGS IN THE SOFTWARE. |
|
|
|
// resolves . and .. elements in a path array with directory names there |
|
// must be no slashes, empty elements, or device names (c:\) in the array |
|
// (so also no leading and trailing slashes - it does not distinguish |
|
// relative and absolute paths) |
|
function normalizeArray(parts, allowAboveRoot) { |
|
// if the path tries to go above the root, `up` ends up > 0 |
|
var up = 0; |
|
for (var i = parts.length - 1; i >= 0; i--) { |
|
var last = parts[i]; |
|
if (last === '.') { |
|
parts.splice(i, 1); |
|
} else if (last === '..') { |
|
parts.splice(i, 1); |
|
up++; |
|
} else if (up) { |
|
parts.splice(i, 1); |
|
up--; |
|
} |
|
} |
|
|
|
// if the path is allowed to go above the root, restore leading ..s |
|
if (allowAboveRoot) { |
|
for (; up--; up) { |
|
parts.unshift('..'); |
|
} |
|
} |
|
|
|
return parts; |
|
} |
|
|
|
// Split a filename into [root, dir, basename, ext], unix version |
|
// 'root' is just a slash, or nothing. |
|
var splitPathRe = |
|
/^(\/?|)([\s\S]*?)((?:\.{1,2}|[^\/]+?|)(\.[^.\/]*|))(?:[\/]*)$/; |
|
var splitPath = function(filename) { |
|
return splitPathRe.exec(filename).slice(1); |
|
}; |
|
|
|
// path.resolve([from ...], to) |
|
// posix version |
|
exports.resolve = function() { |
|
var resolvedPath = '', |
|
resolvedAbsolute = false; |
|
|
|
for (var i = arguments.length - 1; i >= -1 && !resolvedAbsolute; i--) { |
|
var path = (i >= 0) ? arguments[i] : process.cwd(); |
|
|
|
// Skip empty and invalid entries |
|
if (typeof path !== 'string') { |
|
throw new TypeError('Arguments to path.resolve must be strings'); |
|
} else if (!path) { |
|
continue; |
|
} |
|
|
|
resolvedPath = path + '/' + resolvedPath; |
|
resolvedAbsolute = path.charAt(0) === '/'; |
|
} |
|
|
|
// At this point the path should be resolved to a full absolute path, but |
|
// handle relative paths to be safe (might happen when process.cwd() fails) |
|
|
|
// Normalize the path |
|
resolvedPath = normalizeArray(filter(resolvedPath.split('/'), function(p) { |
|
return !!p; |
|
}), !resolvedAbsolute).join('/'); |
|
|
|
return ((resolvedAbsolute ? '/' : '') + resolvedPath) || '.'; |
|
}; |
|
|
|
// path.normalize(path) |
|
// posix version |
|
exports.normalize = function(path) { |
|
var isAbsolute = exports.isAbsolute(path), |
|
trailingSlash = substr(path, -1) === '/'; |
|
|
|
// Normalize the path |
|
path = normalizeArray(filter(path.split('/'), function(p) { |
|
return !!p; |
|
}), !isAbsolute).join('/'); |
|
|
|
if (!path && !isAbsolute) { |
|
path = '.'; |
|
} |
|
if (path && trailingSlash) { |
|
path += '/'; |
|
} |
|
|
|
return (isAbsolute ? '/' : '') + path; |
|
}; |
|
|
|
// posix version |
|
exports.isAbsolute = function(path) { |
|
return path.charAt(0) === '/'; |
|
}; |
|
|
|
// posix version |
|
exports.join = function() { |
|
var paths = Array.prototype.slice.call(arguments, 0); |
|
return exports.normalize(filter(paths, function(p, index) { |
|
if (typeof p !== 'string') { |
|
throw new TypeError('Arguments to path.join must be strings'); |
|
} |
|
return p; |
|
}).join('/')); |
|
}; |
|
|
|
|
|
// path.relative(from, to) |
|
// posix version |
|
exports.relative = function(from, to) { |
|
from = exports.resolve(from).substr(1); |
|
to = exports.resolve(to).substr(1); |
|
|
|
function trim(arr) { |
|
var start = 0; |
|
for (; start < arr.length; start++) { |
|
if (arr[start] !== '') break; |
|
} |
|
|
|
var end = arr.length - 1; |
|
for (; end >= 0; end--) { |
|
if (arr[end] !== '') break; |
|
} |
|
|
|
if (start > end) return []; |
|
return arr.slice(start, end - start + 1); |
|
} |
|
|
|
var fromParts = trim(from.split('/')); |
|
var toParts = trim(to.split('/')); |
|
|
|
var length = Math.min(fromParts.length, toParts.length); |
|
var samePartsLength = length; |
|
for (var i = 0; i < length; i++) { |
|
if (fromParts[i] !== toParts[i]) { |
|
samePartsLength = i; |
|
break; |
|
} |
|
} |
|
|
|
var outputParts = []; |
|
for (var i = samePartsLength; i < fromParts.length; i++) { |
|
outputParts.push('..'); |
|
} |
|
|
|
outputParts = outputParts.concat(toParts.slice(samePartsLength)); |
|
|
|
return outputParts.join('/'); |
|
}; |
|
|
|
exports.sep = '/'; |
|
exports.delimiter = ':'; |
|
|
|
exports.dirname = function(path) { |
|
var result = splitPath(path), |
|
root = result[0], |
|
dir = result[1]; |
|
|
|
if (!root && !dir) { |
|
// No dirname whatsoever |
|
return '.'; |
|
} |
|
|
|
if (dir) { |
|
// It has a dirname, strip trailing slash |
|
dir = dir.substr(0, dir.length - 1); |
|
} |
|
|
|
return root + dir; |
|
}; |
|
|
|
|
|
exports.basename = function(path, ext) { |
|
var f = splitPath(path)[2]; |
|
// TODO: make this comparison case-insensitive on windows? |
|
if (ext && f.substr(-1 * ext.length) === ext) { |
|
f = f.substr(0, f.length - ext.length); |
|
} |
|
return f; |
|
}; |
|
|
|
|
|
exports.extname = function(path) { |
|
return splitPath(path)[3]; |
|
}; |
|
|
|
function filter (xs, f) { |
|
if (xs.filter) return xs.filter(f); |
|
var res = []; |
|
for (var i = 0; i < xs.length; i++) { |
|
if (f(xs[i], i, xs)) res.push(xs[i]); |
|
} |
|
return res; |
|
} |
|
|
|
// String.prototype.substr - negative index don't work in IE8 |
|
var substr = 'ab'.substr(-1) === 'b' |
|
? function (str, start, len) { return str.substr(start, len) } |
|
: function (str, start, len) { |
|
if (start < 0) start = str.length + start; |
|
return str.substr(start, len); |
|
} |
|
; |
|
|
|
}).call(this,_dereq_('_process')) |
|
},{"_process":8}],8:[function(_dereq_,module,exports){ |
|
// shim for using process in browser |
|
|
|
var process = module.exports = {}; |
|
var queue = []; |
|
var draining = false; |
|
|
|
function drainQueue() { |
|
if (draining) { |
|
return; |
|
} |
|
draining = true; |
|
var currentQueue; |
|
var len = queue.length; |
|
while(len) { |
|
currentQueue = queue; |
|
queue = []; |
|
var i = -1; |
|
while (++i < len) { |
|
currentQueue[i](); |
|
} |
|
len = queue.length; |
|
} |
|
draining = false; |
|
} |
|
process.nextTick = function (fun) { |
|
queue.push(fun); |
|
if (!draining) { |
|
setTimeout(drainQueue, 0); |
|
} |
|
}; |
|
|
|
process.title = 'browser'; |
|
process.browser = true; |
|
process.env = {}; |
|
process.argv = []; |
|
process.version = ''; // empty string to avoid regexp issues |
|
process.versions = {}; |
|
|
|
function noop() {} |
|
|
|
process.on = noop; |
|
process.addListener = noop; |
|
process.once = noop; |
|
process.off = noop; |
|
process.removeListener = noop; |
|
process.removeAllListeners = noop; |
|
process.emit = noop; |
|
|
|
process.binding = function (name) { |
|
throw new Error('process.binding is not supported'); |
|
}; |
|
|
|
// TODO(shtylman) |
|
process.cwd = function () { return '/' }; |
|
process.chdir = function (dir) { |
|
throw new Error('process.chdir is not supported'); |
|
}; |
|
process.umask = function() { return 0; }; |
|
|
|
},{}],9:[function(_dereq_,module,exports){ |
|
/* |
|
Copyright (C) 2013 Ariya Hidayat <ariya.hidayat@gmail.com> |
|
Copyright (C) 2013 Thaddee Tyl <thaddee.tyl@gmail.com> |
|
Copyright (C) 2012 Ariya Hidayat <ariya.hidayat@gmail.com> |
|
Copyright (C) 2012 Mathias Bynens <mathias@qiwi.be> |
|
Copyright (C) 2012 Joost-Wim Boekesteijn <joost-wim@boekesteijn.nl> |
|
Copyright (C) 2012 Kris Kowal <kris.kowal@cixar.com> |
|
Copyright (C) 2012 Yusuke Suzuki <utatane.tea@gmail.com> |
|
Copyright (C) 2012 Arpad Borsos <arpad.borsos@googlemail.com> |
|
Copyright (C) 2011 Ariya Hidayat <ariya.hidayat@gmail.com> |
|
|
|
Redistribution and use in source and binary forms, with or without |
|
modification, are permitted provided that the following conditions are met: |
|
|
|
* Redistributions of source code must retain the above copyright |
|
notice, this list of conditions and the following disclaimer. |
|
* Redistributions in binary form must reproduce the above copyright |
|
notice, this list of conditions and the following disclaimer in the |
|
documentation and/or other materials provided with the distribution. |
|
|
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
|
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
|
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
|
ARE DISCLAIMED. IN NO EVENT SHALL <COPYRIGHT HOLDER> BE LIABLE FOR ANY |
|
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES |
|
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
|
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND |
|
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
|
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
|
THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
|
*/ |
|
|
|
(function (root, factory) { |
|
'use strict'; |
|
|
|
// Universal Module Definition (UMD) to support AMD, CommonJS/Node.js, |
|
// Rhino, and plain browser loading. |
|
|
|
/* istanbul ignore next */ |
|
if (typeof define === 'function' && define.amd) { |
|
define(['exports'], factory); |
|
} else if (typeof exports !== 'undefined') { |
|
factory(exports); |
|
} else { |
|
factory((root.esprima = {})); |
|
} |
|
}(this, function (exports) { |
|
'use strict'; |
|
|
|
var Token, |
|
TokenName, |
|
FnExprTokens, |
|
Syntax, |
|
PropertyKind, |
|
Messages, |
|
Regex, |
|
SyntaxTreeDelegate, |
|
XHTMLEntities, |
|
ClassPropertyType, |
|
source, |
|
strict, |
|
index, |
|
lineNumber, |
|
lineStart, |
|
length, |
|
delegate, |
|
lookahead, |
|
state, |
|
extra; |
|
|
|
Token = { |
|
BooleanLiteral: 1, |
|
EOF: 2, |
|
Identifier: 3, |
|
Keyword: 4, |
|
NullLiteral: 5, |
|
NumericLiteral: 6, |
|
Punctuator: 7, |
|
StringLiteral: 8, |
|
RegularExpression: 9, |
|
Template: 10, |
|
JSXIdentifier: 11, |
|
JSXText: 12 |
|
}; |
|
|
|
TokenName = {}; |
|
TokenName[Token.BooleanLiteral] = 'Boolean'; |
|
TokenName[Token.EOF] = '<end>'; |
|
TokenName[Token.Identifier] = 'Identifier'; |
|
TokenName[Token.Keyword] = 'Keyword'; |
|
TokenName[Token.NullLiteral] = 'Null'; |
|
TokenName[Token.NumericLiteral] = 'Numeric'; |
|
TokenName[Token.Punctuator] = 'Punctuator'; |
|
TokenName[Token.StringLiteral] = 'String'; |
|
TokenName[Token.JSXIdentifier] = 'JSXIdentifier'; |
|
TokenName[Token.JSXText] = 'JSXText'; |
|
TokenName[Token.RegularExpression] = 'RegularExpression'; |
|
|
|
// A function following one of those tokens is an expression. |
|
FnExprTokens = ['(', '{', '[', 'in', 'typeof', 'instanceof', 'new', |
|
'return', 'case', 'delete', 'throw', 'void', |
|
// assignment operators |
|
'=', '+=', '-=', '*=', '/=', '%=', '<<=', '>>=', '>>>=', |
|
'&=', '|=', '^=', ',', |
|
// binary/unary operators |
|
'+', '-', '*', '/', '%', '++', '--', '<<', '>>', '>>>', '&', |
|
'|', '^', '!', '~', '&&', '||', '?', ':', '===', '==', '>=', |
|
'<=', '<', '>', '!=', '!==']; |
|
|
|
Syntax = { |
|
AnyTypeAnnotation: 'AnyTypeAnnotation', |
|
ArrayExpression: 'ArrayExpression', |
|
ArrayPattern: 'ArrayPattern', |
|
ArrayTypeAnnotation: 'ArrayTypeAnnotation', |
|
ArrowFunctionExpression: 'ArrowFunctionExpression', |
|
AssignmentExpression: 'AssignmentExpression', |
|
BinaryExpression: 'BinaryExpression', |
|
BlockStatement: 'BlockStatement', |
|
BooleanTypeAnnotation: 'BooleanTypeAnnotation', |
|
BreakStatement: 'BreakStatement', |
|
CallExpression: 'CallExpression', |
|
CatchClause: 'CatchClause', |
|
ClassBody: 'ClassBody', |
|
ClassDeclaration: 'ClassDeclaration', |
|
ClassExpression: 'ClassExpression', |
|
ClassImplements: 'ClassImplements', |
|
ClassProperty: 'ClassProperty', |
|
ComprehensionBlock: 'ComprehensionBlock', |
|
ComprehensionExpression: 'ComprehensionExpression', |
|
ConditionalExpression: 'ConditionalExpression', |
|
ContinueStatement: 'ContinueStatement', |
|
DebuggerStatement: 'DebuggerStatement', |
|
DeclareClass: 'DeclareClass', |
|
DeclareFunction: 'DeclareFunction', |
|
DeclareModule: 'DeclareModule', |
|
DeclareVariable: 'DeclareVariable', |
|
DoWhileStatement: 'DoWhileStatement', |
|
EmptyStatement: 'EmptyStatement', |
|
ExportDeclaration: 'ExportDeclaration', |
|
ExportBatchSpecifier: 'ExportBatchSpecifier', |
|
ExportSpecifier: 'ExportSpecifier', |
|
ExpressionStatement: 'ExpressionStatement', |
|
ForInStatement: 'ForInStatement', |
|
ForOfStatement: 'ForOfStatement', |
|
ForStatement: 'ForStatement', |
|
FunctionDeclaration: 'FunctionDeclaration', |
|
FunctionExpression: 'FunctionExpression', |
|
FunctionTypeAnnotation: 'FunctionTypeAnnotation', |
|
FunctionTypeParam: 'FunctionTypeParam', |
|
GenericTypeAnnotation: 'GenericTypeAnnotation', |
|
Identifier: 'Identifier', |
|
IfStatement: 'IfStatement', |
|
ImportDeclaration: 'ImportDeclaration', |
|
ImportDefaultSpecifier: 'ImportDefaultSpecifier', |
|
ImportNamespaceSpecifier: 'ImportNamespaceSpecifier', |
|
ImportSpecifier: 'ImportSpecifier', |
|
InterfaceDeclaration: 'InterfaceDeclaration', |
|
InterfaceExtends: 'InterfaceExtends', |
|
IntersectionTypeAnnotation: 'IntersectionTypeAnnotation', |
|
LabeledStatement: 'LabeledStatement', |
|
Literal: 'Literal', |
|
LogicalExpression: 'LogicalExpression', |
|
MemberExpression: 'MemberExpression', |
|
MethodDefinition: 'MethodDefinition', |
|
ModuleSpecifier: 'ModuleSpecifier', |
|
NewExpression: 'NewExpression', |
|
NullableTypeAnnotation: 'NullableTypeAnnotation', |
|
NumberTypeAnnotation: 'NumberTypeAnnotation', |
|
ObjectExpression: 'ObjectExpression', |
|
ObjectPattern: 'ObjectPattern', |
|
ObjectTypeAnnotation: 'ObjectTypeAnnotation', |
|
ObjectTypeCallProperty: 'ObjectTypeCallProperty', |
|
ObjectTypeIndexer: 'ObjectTypeIndexer', |
|
ObjectTypeProperty: 'ObjectTypeProperty', |
|
Program: 'Program', |
|
Property: 'Property', |
|
QualifiedTypeIdentifier: 'QualifiedTypeIdentifier', |
|
ReturnStatement: 'ReturnStatement', |
|
SequenceExpression: 'SequenceExpression', |
|
SpreadElement: 'SpreadElement', |
|
SpreadProperty: 'SpreadProperty', |
|
StringLiteralTypeAnnotation: 'StringLiteralTypeAnnotation', |
|
StringTypeAnnotation: 'StringTypeAnnotation', |
|
SwitchCase: 'SwitchCase', |
|
SwitchStatement: 'SwitchStatement', |
|
TaggedTemplateExpression: 'TaggedTemplateExpression', |
|
TemplateElement: 'TemplateElement', |
|
TemplateLiteral: 'TemplateLiteral', |
|
ThisExpression: 'ThisExpression', |
|
ThrowStatement: 'ThrowStatement', |
|
TupleTypeAnnotation: 'TupleTypeAnnotation', |
|
TryStatement: 'TryStatement', |
|
TypeAlias: 'TypeAlias', |
|
TypeAnnotation: 'TypeAnnotation', |
|
TypeCastExpression: 'TypeCastExpression', |
|
TypeofTypeAnnotation: 'TypeofTypeAnnotation', |
|
TypeParameterDeclaration: 'TypeParameterDeclaration', |
|
TypeParameterInstantiation: 'TypeParameterInstantiation', |
|
UnaryExpression: 'UnaryExpression', |
|
UnionTypeAnnotation: 'UnionTypeAnnotation', |
|
UpdateExpression: 'UpdateExpression', |
|
VariableDeclaration: 'VariableDeclaration', |
|
VariableDeclarator: 'VariableDeclarator', |
|
VoidTypeAnnotation: 'VoidTypeAnnotation', |
|
WhileStatement: 'WhileStatement', |
|
WithStatement: 'WithStatement', |
|
JSXIdentifier: 'JSXIdentifier', |
|
JSXNamespacedName: 'JSXNamespacedName', |
|
JSXMemberExpression: 'JSXMemberExpression', |
|
JSXEmptyExpression: 'JSXEmptyExpression', |
|
JSXExpressionContainer: 'JSXExpressionContainer', |
|
JSXElement: 'JSXElement', |
|
JSXClosingElement: 'JSXClosingElement', |
|
JSXOpeningElement: 'JSXOpeningElement', |
|
JSXAttribute: 'JSXAttribute', |
|
JSXSpreadAttribute: 'JSXSpreadAttribute', |
|
JSXText: 'JSXText', |
|
YieldExpression: 'YieldExpression', |
|
AwaitExpression: 'AwaitExpression' |
|
}; |
|
|
|
PropertyKind = { |
|
Data: 1, |
|
Get: 2, |
|
Set: 4 |
|
}; |
|
|
|
ClassPropertyType = { |
|
'static': 'static', |
|
prototype: 'prototype' |
|
}; |
|
|
|
// Error messages should be identical to V8. |
|
Messages = { |
|
UnexpectedToken: 'Unexpected token %0', |
|
UnexpectedNumber: 'Unexpected number', |
|
UnexpectedString: 'Unexpected string', |
|
UnexpectedIdentifier: 'Unexpected identifier', |
|
UnexpectedReserved: 'Unexpected reserved word', |
|
UnexpectedTemplate: 'Unexpected quasi %0', |
|
UnexpectedEOS: 'Unexpected end of input', |
|
NewlineAfterThrow: 'Illegal newline after throw', |
|
InvalidRegExp: 'Invalid regular expression', |
|
UnterminatedRegExp: 'Invalid regular expression: missing /', |
|
InvalidLHSInAssignment: 'Invalid left-hand side in assignment', |
|
InvalidLHSInFormalsList: 'Invalid left-hand side in formals list', |
|
InvalidLHSInForIn: 'Invalid left-hand side in for-in', |
|
MultipleDefaultsInSwitch: 'More than one default clause in switch statement', |
|
NoCatchOrFinally: 'Missing catch or finally after try', |
|
UnknownLabel: 'Undefined label \'%0\'', |
|
Redeclaration: '%0 \'%1\' has already been declared', |
|
IllegalContinue: 'Illegal continue statement', |
|
IllegalBreak: 'Illegal break statement', |
|
IllegalDuplicateClassProperty: 'Illegal duplicate property in class definition', |
|
IllegalClassConstructorProperty: 'Illegal constructor property in class definition', |
|
IllegalReturn: 'Illegal return statement', |
|
IllegalSpread: 'Illegal spread element', |
|
StrictModeWith: 'Strict mode code may not include a with statement', |
|
StrictCatchVariable: 'Catch variable may not be eval or arguments in strict mode', |
|
StrictVarName: 'Variable name may not be eval or arguments in strict mode', |
|
StrictParamName: 'Parameter name eval or arguments is not allowed in strict mode', |
|
StrictParamDupe: 'Strict mode function may not have duplicate parameter names', |
|
ParameterAfterRestParameter: 'Rest parameter must be final parameter of an argument list', |
|
DefaultRestParameter: 'Rest parameter can not have a default value', |
|
ElementAfterSpreadElement: 'Spread must be the final element of an element list', |
|
PropertyAfterSpreadProperty: 'A rest property must be the final property of an object literal', |
|
ObjectPatternAsRestParameter: 'Invalid rest parameter', |
|
ObjectPatternAsSpread: 'Invalid spread argument', |
|
StrictFunctionName: 'Function name may not be eval or arguments in strict mode', |
|
StrictOctalLiteral: 'Octal literals are not allowed in strict mode.', |
|
StrictDelete: 'Delete of an unqualified identifier in strict mode.', |
|
StrictDuplicateProperty: 'Duplicate data property in object literal not allowed in strict mode', |
|
AccessorDataProperty: 'Object literal may not have data and accessor property with the same name', |
|
AccessorGetSet: 'Object literal may not have multiple get/set accessors with the same name', |
|
StrictLHSAssignment: 'Assignment to eval or arguments is not allowed in strict mode', |
|
StrictLHSPostfix: 'Postfix increment/decrement may not have eval or arguments operand in strict mode', |
|
StrictLHSPrefix: 'Prefix increment/decrement may not have eval or arguments operand in strict mode', |
|
StrictReservedWord: 'Use of future reserved word in strict mode', |
|
MissingFromClause: 'Missing from clause', |
|
NoAsAfterImportNamespace: 'Missing as after import *', |
|
InvalidModuleSpecifier: 'Invalid module specifier', |
|
IllegalImportDeclaration: 'Illegal import declaration', |
|
IllegalExportDeclaration: 'Illegal export declaration', |
|
NoUninitializedConst: 'Const must be initialized', |
|
ComprehensionRequiresBlock: 'Comprehension must have at least one block', |
|
ComprehensionError: 'Comprehension Error', |
|
EachNotAllowed: 'Each is not supported', |
|
InvalidJSXAttributeValue: 'JSX value should be either an expression or a quoted JSX text', |
|
ExpectedJSXClosingTag: 'Expected corresponding JSX closing tag for %0', |
|
AdjacentJSXElements: 'Adjacent JSX elements must be wrapped in an enclosing tag', |
|
ConfusedAboutFunctionType: 'Unexpected token =>. It looks like ' + |
|
'you are trying to write a function type, but you ended up ' + |
|
'writing a grouped type followed by an =>, which is a syntax ' + |
|
'error. Remember, function type parameters are named so function ' + |
|
'types look like (name1: type1, name2: type2) => returnType. You ' + |
|
'probably wrote (type1) => returnType' |
|
}; |
|
|
|
// See also tools/generate-unicode-regex.py. |
|
Regex = { |
|
NonAsciiIdentifierStart: new RegExp('[\xaa\xb5\xba\xc0-\xd6\xd8-\xf6\xf8-\u02c1\u02c6-\u02d1\u02e0-\u02e4\u02ec\u02ee\u0370-\u0374\u0376\u0377\u037a-\u037d\u0386\u0388-\u038a\u038c\u038e-\u03a1\u03a3-\u03f5\u03f7-\u0481\u048a-\u0527\u0531-\u0556\u0559\u0561-\u0587\u05d0-\u05ea\u05f0-\u05f2\u0620-\u064a\u066e\u066f\u0671-\u06d3\u06d5\u06e5\u06e6\u06ee\u06ef\u06fa-\u06fc\u06ff\u0710\u0712-\u072f\u074d-\u07a5\u07b1\u07ca-\u07ea\u07f4\u07f5\u07fa\u0800-\u0815\u081a\u0824\u0828\u0840-\u0858\u08a0\u08a2-\u08ac\u0904-\u0939\u093d\u0950\u0958-\u0961\u0971-\u0977\u0979-\u097f\u0985-\u098c\u098f\u0990\u0993-\u09a8\u09aa-\u09b0\u09b2\u09b6-\u09b9\u09bd\u09ce\u09dc\u09dd\u09df-\u09e1\u09f0\u09f1\u0a05-\u0a0a\u0a0f\u0a10\u0a13-\u0a28\u0a2a-\u0a30\u0a32\u0a33\u0a35\u0a36\u0a38\u0a39\u0a59-\u0a5c\u0a5e\u0a72-\u0a74\u0a85-\u0a8d\u0a8f-\u0a91\u0a93-\u0aa8\u0aaa-\u0ab0\u0ab2\u0ab3\u0ab5-\u0ab9\u0abd\u0ad0\u0ae0\u0ae1\u0b05-\u0b0c\u0b0f\u0b10\u0b13-\u0b28\u0b2a-\u0b30\u0b32\u0b33\u0b35-\u0b39\u0b3d\u0b5c\u0b5d\u0b5f-\u0b61\u0b71\u0b83\u0b85-\u0b8a\u0b8e-\u0b90\u0b92-\u0b95\u0b99\u0b9a\u0b9c\u0b9e\u0b9f\u0ba3\u0ba4\u0ba8-\u0baa\u0bae-\u0bb9\u0bd0\u0c05-\u0c0c\u0c0e-\u0c10\u0c12-\u0c28\u0c2a-\u0c33\u0c35-\u0c39\u0c3d\u0c58\u0c59\u0c60\u0c61\u0c85-\u0c8c\u0c8e-\u0c90\u0c92-\u0ca8\u0caa-\u0cb3\u0cb5-\u0cb9\u0cbd\u0cde\u0ce0\u0ce1\u0cf1\u0cf2\u0d05-\u0d0c\u0d0e-\u0d10\u0d12-\u0d3a\u0d3d\u0d4e\u0d60\u0d61\u0d7a-\u0d7f\u0d85-\u0d96\u0d9a-\u0db1\u0db3-\u0dbb\u0dbd\u0dc0-\u0dc6\u0e01-\u0e30\u0e32\u0e33\u0e40-\u0e46\u0e81\u0e82\u0e84\u0e87\u0e88\u0e8a\u0e8d\u0e94-\u0e97\u0e99-\u0e9f\u0ea1-\u0ea3\u0ea5\u0ea7\u0eaa\u0eab\u0ead-\u0eb0\u0eb2\u0eb3\u0ebd\u0ec0-\u0ec4\u0ec6\u0edc-\u0edf\u0f00\u0f40-\u0f47\u0f49-\u0f6c\u0f88-\u0f8c\u1000-\u102a\u103f\u1050-\u1055\u105a-\u105d\u1061\u1065\u1066\u106e-\u1070\u1075-\u1081\u108e\u10a0-\u10c5\u10c7\u10cd\u10d0-\u10fa\u10fc-\u1248\u124a-\u124d\u1250-\u1256\u1258\u125a-\u125d\u1260-\u1288\u128a-\u128d\u1290-\u12b0\u12b2-\u12b5\u12b8-\u12be\u12c0\u12c2-\u12c5\u12c8-\u12d6\u12d8-\u1310\u1312-\u1315\u1318-\u135a\u1380-\u138f\u13a0-\u13f4\u1401-\u166c\u166f-\u167f\u1681-\u169a\u16a0-\u16ea\u16ee-\u16f0\u1700-\u170c\u170e-\u1711\u1720-\u1731\u1740-\u1751\u1760-\u176c\u176e-\u1770\u1780-\u17b3\u17d7\u17dc\u1820-\u1877\u1880-\u18a8\u18aa\u18b0-\u18f5\u1900-\u191c\u1950-\u196d\u1970-\u1974\u1980-\u19ab\u19c1-\u19c7\u1a00-\u1a16\u1a20-\u1a54\u1aa7\u1b05-\u1b33\u1b45-\u1b4b\u1b83-\u1ba0\u1bae\u1baf\u1bba-\u1be5\u1c00-\u1c23\u1c4d-\u1c4f\u1c5a-\u1c7d\u1ce9-\u1cec\u1cee-\u1cf1\u1cf5\u1cf6\u1d00-\u1dbf\u1e00-\u1f15\u1f18-\u1f1d\u1f20-\u1f45\u1f48-\u1f4d\u1f50-\u1f57\u1f59\u1f5b\u1f5d\u1f5f-\u1f7d\u1f80-\u1fb4\u1fb6-\u1fbc\u1fbe\u1fc2-\u1fc4\u1fc6-\u1fcc\u1fd0-\u1fd3\u1fd6-\u1fdb\u1fe0-\u1fec\u1ff2-\u1ff4\u1ff6-\u1ffc\u2071\u207f\u2090-\u209c\u2102\u2107\u210a-\u2113\u2115\u2119-\u211d\u2124\u2126\u2128\u212a-\u212d\u212f-\u2139\u213c-\u213f\u2145-\u2149\u214e\u2160-\u2188\u2c00-\u2c2e\u2c30-\u2c5e\u2c60-\u2ce4\u2ceb-\u2cee\u2cf2\u2cf3\u2d00-\u2d25\u2d27\u2d2d\u2d30-\u2d67\u2d6f\u2d80-\u2d96\u2da0-\u2da6\u2da8-\u2dae\u2db0-\u2db6\u2db8-\u2dbe\u2dc0-\u2dc6\u2dc8-\u2dce\u2dd0-\u2dd6\u2dd8-\u2dde\u2e2f\u3005-\u3007\u3021-\u3029\u3031-\u3035\u3038-\u303c\u3041-\u3096\u309d-\u309f\u30a1-\u30fa\u30fc-\u30ff\u3105-\u312d\u3131-\u318e\u31a0-\u31ba\u31f0-\u31ff\u3400-\u4db5\u4e00-\u9fcc\ua000-\ua48c\ua4d0-\ua4fd\ua500-\ua60c\ua610-\ua61f\ua62a\ua62b\ua640-\ua66e\ua67f-\ua697\ua6a0-\ua6ef\ua717-\ua71f\ua722-\ua788\ua78b-\ua78e\ua790-\ua793\ua7a0-\ua7aa\ua7f8-\ua801\ua803-\ua805\ua807-\ua80a\ua80c-\ua822\ua840-\ua873\ua882-\ua8b3\ua8f2-\ua8f7\ua8fb\ua90a-\ua925\ua930-\ua946\ua960-\ua97c\ua984-\ua9b2\ua9cf\uaa00-\uaa28\uaa40-\uaa42\uaa44-\uaa4b\uaa60-\uaa76\uaa7a\uaa80-\uaaaf\uaab1\uaab5\uaab6\uaab9-\uaabd\uaac0\uaac2\uaadb-\uaadd\uaae0-\uaaea\uaaf2-\uaaf4\uab01-\uab06\uab09-\uab0e\uab11-\uab16\uab20-\uab26\uab28-\uab2e\uabc0-\uabe2\uac00-\ud7a3\ud7b0-\ud7c6\ud7cb-\ud7fb\uf900-\ufa6d\ufa70-\ufad9\ufb00-\ufb06\ufb13-\ufb17\ufb1d\ufb1f-\ufb28\ufb2a-\ufb36\ufb38-\ufb3c\ufb3e\ufb40\ufb41\ufb43\ufb44\ufb46-\ufbb1\ufbd3-\ufd3d\ufd50-\ufd8f\ufd92-\ufdc7\ufdf0-\ufdfb\ufe70-\ufe74\ufe76-\ufefc\uff21-\uff3a\uff41-\uff5a\uff66-\uffbe\uffc2-\uffc7\uffca-\uffcf\uffd2-\uffd7\uffda-\uffdc]'), |
|
NonAsciiIdentifierPart: new RegExp('[\xaa\xb5\xba\xc0-\xd6\xd8-\xf6\xf8-\u02c1\u02c6-\u02d1\u02e0-\u02e4\u02ec\u02ee\u0300-\u0374\u0376\u0377\u037a-\u037d\u0386\u0388-\u038a\u038c\u038e-\u03a1\u03a3-\u03f5\u03f7-\u0481\u0483-\u0487\u048a-\u0527\u0531-\u0556\u0559\u0561-\u0587\u0591-\u05bd\u05bf\u05c1\u05c2\u05c4\u05c5\u05c7\u05d0-\u05ea\u05f0-\u05f2\u0610-\u061a\u0620-\u0669\u066e-\u06d3\u06d5-\u06dc\u06df-\u06e8\u06ea-\u06fc\u06ff\u0710-\u074a\u074d-\u07b1\u07c0-\u07f5\u07fa\u0800-\u082d\u0840-\u085b\u08a0\u08a2-\u08ac\u08e4-\u08fe\u0900-\u0963\u0966-\u096f\u0971-\u0977\u0979-\u097f\u0981-\u0983\u0985-\u098c\u098f\u0990\u0993-\u09a8\u09aa-\u09b0\u09b2\u09b6-\u09b9\u09bc-\u09c4\u09c7\u09c8\u09cb-\u09ce\u09d7\u09dc\u09dd\u09df-\u09e3\u09e6-\u09f1\u0a01-\u0a03\u0a05-\u0a0a\u0a0f\u0a10\u0a13-\u0a28\u0a2a-\u0a30\u0a32\u0a33\u0a35\u0a36\u0a38\u0a39\u0a3c\u0a3e-\u0a42\u0a47\u0a48\u0a4b-\u0a4d\u0a51\u0a59-\u0a5c\u0a5e\u0a66-\u0a75\u0a81-\u0a83\u0a85-\u0a8d\u0a8f-\u0a91\u0a93-\u0aa8\u0aaa-\u0ab0\u0ab2\u0ab3\u0ab5-\u0ab9\u0abc-\u0ac5\u0ac7-\u0ac9\u0acb-\u0acd\u0ad0\u0ae0-\u0ae3\u0ae6-\u0aef\u0b01-\u0b03\u0b05-\u0b0c\u0b0f\u0b10\u0b13-\u0b28\u0b2a-\u0b30\u0b32\u0b33\u0b35-\u0b39\u0b3c-\u0b44\u0b47\u0b48\u0b4b-\u0b4d\u0b56\u0b57\u0b5c\u0b5d\u0b5f-\u0b63\u0b66-\u0b6f\u0b71\u0b82\u0b83\u0b85-\u0b8a\u0b8e-\u0b90\u0b92-\u0b95\u0b99\u0b9a\u0b9c\u0b9e\u0b9f\u0ba3\u0ba4\u0ba8-\u0baa\u0bae-\u0bb9\u0bbe-\u0bc2\u0bc6-\u0bc8\u0bca-\u0bcd\u0bd0\u0bd7\u0be6-\u0bef\u0c01-\u0c03\u0c05-\u0c0c\u0c0e-\u0c10\u0c12-\u0c28\u0c2a-\u0c33\u0c35-\u0c39\u0c3d-\u0c44\u0c46-\u0c48\u0c4a-\u0c4d\u0c55\u0c56\u0c58\u0c59\u0c60-\u0c63\u0c66-\u0c6f\u0c82\u0c83\u0c85-\u0c8c\u0c8e-\u0c90\u0c92-\u0ca8\u0caa-\u0cb3\u0cb5-\u0cb9\u0cbc-\u0cc4\u0cc6-\u0cc8\u0cca-\u0ccd\u0cd5\u0cd6\u0cde\u0ce0-\u0ce3\u0ce6-\u0cef\u0cf1\u0cf2\u0d02\u0d03\u0d05-\u0d0c\u0d0e-\u0d10\u0d12-\u0d3a\u0d3d-\u0d44\u0d46-\u0d48\u0d4a-\u0d4e\u0d57\u0d60-\u0d63\u0d66-\u0d6f\u0d7a-\u0d7f\u0d82\u0d83\u0d85-\u0d96\u0d9a-\u0db1\u0db3-\u0dbb\u0dbd\u0dc0-\u0dc6\u0dca\u0dcf-\u0dd4\u0dd6\u0dd8-\u0ddf\u0df2\u0df3\u0e01-\u0e3a\u0e40-\u0e4e\u0e50-\u0e59\u0e81\u0e82\u0e84\u0e87\u0e88\u0e8a\u0e8d\u0e94-\u0e97\u0e99-\u0e9f\u0ea1-\u0ea3\u0ea5\u0ea7\u0eaa\u0eab\u0ead-\u0eb9\u0ebb-\u0ebd\u0ec0-\u0ec4\u0ec6\u0ec8-\u0ecd\u0ed0-\u0ed9\u0edc-\u0edf\u0f00\u0f18\u0f19\u0f20-\u0f29\u0f35\u0f37\u0f39\u0f3e-\u0f47\u0f49-\u0f6c\u0f71-\u0f84\u0f86-\u0f97\u0f99-\u0fbc\u0fc6\u1000-\u1049\u1050-\u109d\u10a0-\u10c5\u10c7\u10cd\u10d0-\u10fa\u10fc-\u1248\u124a-\u124d\u1250-\u1256\u1258\u125a-\u125d\u1260-\u1288\u128a-\u128d\u1290-\u12b0\u12b2-\u12b5\u12b8-\u12be\u12c0\u12c2-\u12c5\u12c8-\u12d6\u12d8-\u1310\u1312-\u1315\u1318-\u135a\u135d-\u135f\u1380-\u138f\u13a0-\u13f4\u1401-\u166c\u166f-\u167f\u1681-\u169a\u16a0-\u16ea\u16ee-\u16f0\u1700-\u170c\u170e-\u1714\u1720-\u1734\u1740-\u1753\u1760-\u176c\u176e-\u1770\u1772\u1773\u1780-\u17d3\u17d7\u17dc\u17dd\u17e0-\u17e9\u180b-\u180d\u1810-\u1819\u1820-\u1877\u1880-\u18aa\u18b0-\u18f5\u1900-\u191c\u1920-\u192b\u1930-\u193b\u1946-\u196d\u1970-\u1974\u1980-\u19ab\u19b0-\u19c9\u19d0-\u19d9\u1a00-\u1a1b\u1a20-\u1a5e\u1a60-\u1a7c\u1a7f-\u1a89\u1a90-\u1a99\u1aa7\u1b00-\u1b4b\u1b50-\u1b59\u1b6b-\u1b73\u1b80-\u1bf3\u1c00-\u1c37\u1c40-\u1c49\u1c4d-\u1c7d\u1cd0-\u1cd2\u1cd4-\u1cf6\u1d00-\u1de6\u1dfc-\u1f15\u1f18-\u1f1d\u1f20-\u1f45\u1f48-\u1f4d\u1f50-\u1f57\u1f59\u1f5b\u1f5d\u1f5f-\u1f7d\u1f80-\u1fb4\u1fb6-\u1fbc\u1fbe\u1fc2-\u1fc4\u1fc6-\u1fcc\u1fd0-\u1fd3\u1fd6-\u1fdb\u1fe0-\u1fec\u1ff2-\u1ff4\u1ff6-\u1ffc\u200c\u200d\u203f\u2040\u2054\u2071\u207f\u2090-\u209c\u20d0-\u20dc\u20e1\u20e5-\u20f0\u2102\u2107\u210a-\u2113\u2115\u2119-\u211d\u2124\u2126\u2128\u212a-\u212d\u212f-\u2139\u213c-\u213f\u2145-\u2149\u214e\u2160-\u2188\u2c00-\u2c2e\u2c30-\u2c5e\u2c60-\u2ce4\u2ceb-\u2cf3\u2d00-\u2d25\u2d27\u2d2d\u2d30-\u2d67\u2d6f\u2d7f-\u2d96\u2da0-\u2da6\u2da8-\u2dae\u2db0-\u2db6\u2db8-\u2dbe\u2dc0-\u2dc6\u2dc8-\u2dce\u2dd0-\u2dd6\u2dd8-\u2dde\u2de0-\u2dff\u2e2f\u3005-\u3007\u3021-\u302f\u3031-\u3035\u3038-\u303c\u3041-\u3096\u3099\u309a\u309d-\u309f\u30a1-\u30fa\u30fc-\u30ff\u3105-\u312d\u3131-\u318e\u31a0-\u31ba\u31f0-\u31ff\u3400-\u4db5\u4e00-\u9fcc\ua000-\ua48c\ua4d0-\ua4fd\ua500-\ua60c\ua610-\ua62b\ua640-\ua66f\ua674-\ua67d\ua67f-\ua697\ua69f-\ua6f1\ua717-\ua71f\ua722-\ua788\ua78b-\ua78e\ua790-\ua793\ua7a0-\ua7aa\ua7f8-\ua827\ua840-\ua873\ua880-\ua8c4\ua8d0-\ua8d9\ua8e0-\ua8f7\ua8fb\ua900-\ua92d\ua930-\ua953\ua960-\ua97c\ua980-\ua9c0\ua9cf-\ua9d9\uaa00-\uaa36\uaa40-\uaa4d\uaa50-\uaa59\uaa60-\uaa76\uaa7a\uaa7b\uaa80-\uaac2\uaadb-\uaadd\uaae0-\uaaef\uaaf2-\uaaf6\uab01-\uab06\uab09-\uab0e\uab11-\uab16\uab20-\uab26\uab28-\uab2e\uabc0-\uabea\uabec\uabed\uabf0-\uabf9\uac00-\ud7a3\ud7b0-\ud7c6\ud7cb-\ud7fb\uf900-\ufa6d\ufa70-\ufad9\ufb00-\ufb06\ufb13-\ufb17\ufb1d-\ufb28\ufb2a-\ufb36\ufb38-\ufb3c\ufb3e\ufb40\ufb41\ufb43\ufb44\ufb46-\ufbb1\ufbd3-\ufd3d\ufd50-\ufd8f\ufd92-\ufdc7\ufdf0-\ufdfb\ufe00-\ufe0f\ufe20-\ufe26\ufe33\ufe34\ufe4d-\ufe4f\ufe70-\ufe74\ufe76-\ufefc\uff10-\uff19\uff21-\uff3a\uff3f\uff41-\uff5a\uff66-\uffbe\uffc2-\uffc7\uffca-\uffcf\uffd2-\uffd7\uffda-\uffdc]'), |
|
LeadingZeros: new RegExp('^0+(?!$)') |
|
}; |
|
|
|
// Ensure the condition is true, otherwise throw an error. |
|
// This is only to have a better contract semantic, i.e. another safety net |
|
// to catch a logic error. The condition shall be fulfilled in normal case. |
|
// Do NOT use this to enforce a certain condition on any user input. |
|
|
|
function assert(condition, message) { |
|
/* istanbul ignore if */ |
|
if (!condition) { |
|
throw new Error('ASSERT: ' + message); |
|
} |
|
} |
|
|
|
function StringMap() { |
|
this.$data = {}; |
|
} |
|
|
|
StringMap.prototype.get = function (key) { |
|
key = '$' + key; |
|
return this.$data[key]; |
|
}; |
|
|
|
StringMap.prototype.set = function (key, value) { |
|
key = '$' + key; |
|
this.$data[key] = value; |
|
return this; |
|
}; |
|
|
|
StringMap.prototype.has = function (key) { |
|
key = '$' + key; |
|
return Object.prototype.hasOwnProperty.call(this.$data, key); |
|
}; |
|
|
|
StringMap.prototype["delete"] = function (key) { |
|
key = '$' + key; |
|
return delete this.$data[key]; |
|
}; |
|
|
|
function isDecimalDigit(ch) { |
|
return (ch >= 48 && ch <= 57); // 0..9 |
|
} |
|
|
|
function isHexDigit(ch) { |
|
return '0123456789abcdefABCDEF'.indexOf(ch) >= 0; |
|
} |
|
|
|
function isOctalDigit(ch) { |
|
return '01234567'.indexOf(ch) >= 0; |
|
} |
|
|
|
|
|
// 7.2 White Space |
|
|
|
function isWhiteSpace(ch) { |
|
return (ch === 32) || // space |
|
(ch === 9) || // tab |
|
(ch === 0xB) || |
|
(ch === 0xC) || |
|
(ch === 0xA0) || |
|
(ch >= 0x1680 && '\u1680\u180E\u2000\u2001\u2002\u2003\u2004\u2005\u2006\u2007\u2008\u2009\u200A\u202F\u205F\u3000\uFEFF'.indexOf(String.fromCharCode(ch)) > 0); |
|
} |
|
|
|
// 7.3 Line Terminators |
|
|
|
function isLineTerminator(ch) { |
|
return (ch === 10) || (ch === 13) || (ch === 0x2028) || (ch === 0x2029); |
|
} |
|
|
|
// 7.6 Identifier Names and Identifiers |
|
|
|
function isIdentifierStart(ch) { |
|
return (ch === 36) || (ch === 95) || // $ (dollar) and _ (underscore) |
|
(ch >= 65 && ch <= 90) || // A..Z |
|
(ch >= 97 && ch <= 122) || // a..z |
|
(ch === 92) || // \ (backslash) |
|
((ch >= 0x80) && Regex.NonAsciiIdentifierStart.test(String.fromCharCode(ch))); |
|
} |
|
|
|
function isIdentifierPart(ch) { |
|
return (ch === 36) || (ch === 95) || // $ (dollar) and _ (underscore) |
|
(ch >= 65 && ch <= 90) || // A..Z |
|
(ch >= 97 && ch <= 122) || // a..z |
|
(ch >= 48 && ch <= 57) || // 0..9 |
|
(ch === 92) || // \ (backslash) |
|
((ch >= 0x80) && Regex.NonAsciiIdentifierPart.test(String.fromCharCode(ch))); |
|
} |
|
|
|
// 7.6.1.2 Future Reserved Words |
|
|
|
function isFutureReservedWord(id) { |
|
switch (id) { |
|
case 'class': |
|
case 'enum': |
|
case 'export': |
|
case 'extends': |
|
case 'import': |
|
case 'super': |
|
return true; |
|
default: |
|
return false; |
|
} |
|
} |
|
|
|
function isStrictModeReservedWord(id) { |
|
switch (id) { |
|
case 'implements': |
|
case 'interface': |
|
case 'package': |
|
case 'private': |
|
case 'protected': |
|
case 'public': |
|
case 'static': |
|
case 'yield': |
|
case 'let': |
|
return true; |
|
default: |
|
return false; |
|
} |
|
} |
|
|
|
function isRestrictedWord(id) { |
|
return id === 'eval' || id === 'arguments'; |
|
} |
|
|
|
// 7.6.1.1 Keywords |
|
|
|
function isKeyword(id) { |
|
if (strict && isStrictModeReservedWord(id)) { |
|
return true; |
|
} |
|
|
|
// 'const' is specialized as Keyword in V8. |
|
// 'yield' is only treated as a keyword in strict mode. |
|
// 'let' is for compatiblity with SpiderMonkey and ES.next. |
|
// Some others are from future reserved words. |
|
|
|
switch (id.length) { |
|
case 2: |
|
return (id === 'if') || (id === 'in') || (id === 'do'); |
|
case 3: |
|
return (id === 'var') || (id === 'for') || (id === 'new') || |
|
(id === 'try') || (id === 'let'); |
|
case 4: |
|
return (id === 'this') || (id === 'else') || (id === 'case') || |
|
(id === 'void') || (id === 'with') || (id === 'enum'); |
|
case 5: |
|
return (id === 'while') || (id === 'break') || (id === 'catch') || |
|
(id === 'throw') || (id === 'const') || |
|
(id === 'class') || (id === 'super'); |
|
case 6: |
|
return (id === 'return') || (id === 'typeof') || (id === 'delete') || |
|
(id === 'switch') || (id === 'export') || (id === 'import'); |
|
case 7: |
|
return (id === 'default') || (id === 'finally') || (id === 'extends'); |
|
case 8: |
|
return (id === 'function') || (id === 'continue') || (id === 'debugger'); |
|
case 10: |
|
return (id === 'instanceof'); |
|
default: |
|
return false; |
|
} |
|
} |
|
|
|
// 7.4 Comments |
|
|
|
function addComment(type, value, start, end, loc) { |
|
var comment; |
|
assert(typeof start === 'number', 'Comment must have valid position'); |
|
|
|
// Because the way the actual token is scanned, often the comments |
|
// (if any) are skipped twice during the lexical analysis. |
|
// Thus, we need to skip adding a comment if the comment array already |
|
// handled it. |
|
if (state.lastCommentStart >= start) { |
|
return; |
|
} |
|
state.lastCommentStart = start; |
|
|
|
comment = { |
|
type: type, |
|
value: value |
|
}; |
|
if (extra.range) { |
|
comment.range = [start, end]; |
|
} |
|
if (extra.loc) { |
|
comment.loc = loc; |
|
} |
|
extra.comments.push(comment); |
|
if (extra.attachComment) { |
|
extra.leadingComments.push(comment); |
|
extra.trailingComments.push(comment); |
|
} |
|
} |
|
|
|
function skipSingleLineComment() { |
|
var start, loc, ch, comment; |
|
|
|
start = index - 2; |
|
loc = { |
|
start: { |
|
line: lineNumber, |
|
column: index - lineStart - 2 |
|
} |
|
}; |
|
|
|
while (index < length) { |
|
ch = source.charCodeAt(index); |
|
++index; |
|
if (isLineTerminator(ch)) { |
|
if (extra.comments) { |
|
comment = source.slice(start + 2, index - 1); |
|
loc.end = { |
|
line: lineNumber, |
|
column: index - lineStart - 1 |
|
}; |
|
addComment('Line', comment, start, index - 1, loc); |
|
} |
|
if (ch === 13 && source.charCodeAt(index) === 10) { |
|
++index; |
|
} |
|
++lineNumber; |
|
lineStart = index; |
|
return; |
|
} |
|
} |
|
|
|
if (extra.comments) { |
|
comment = source.slice(start + 2, index); |
|
loc.end = { |
|
line: lineNumber, |
|
column: index - lineStart |
|
}; |
|
addComment('Line', comment, start, index, loc); |
|
} |
|
} |
|
|
|
function skipMultiLineComment() { |
|
var start, loc, ch, comment; |
|
|
|
if (extra.comments) { |
|
start = index - 2; |
|
loc = { |
|
start: { |
|
line: lineNumber, |
|
column: index - lineStart - 2 |
|
} |
|
}; |
|
} |
|
|
|
while (index < length) { |
|
ch = source.charCodeAt(index); |
|
if (isLineTerminator(ch)) { |
|
if (ch === 13 && source.charCodeAt(index + 1) === 10) { |
|
++index; |
|
} |
|
++lineNumber; |
|
++index; |
|
lineStart = index; |
|
if (index >= length) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
} else if (ch === 42) { |
|
// Block comment ends with '*/' (char #42, char #47). |
|
if (source.charCodeAt(index + 1) === 47) { |
|
++index; |
|
++index; |
|
if (extra.comments) { |
|
comment = source.slice(start + 2, index - 2); |
|
loc.end = { |
|
line: lineNumber, |
|
column: index - lineStart |
|
}; |
|
addComment('Block', comment, start, index, loc); |
|
} |
|
return; |
|
} |
|
++index; |
|
} else { |
|
++index; |
|
} |
|
} |
|
|
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
function skipComment() { |
|
var ch; |
|
|
|
while (index < length) { |
|
ch = source.charCodeAt(index); |
|
|
|
if (isWhiteSpace(ch)) { |
|
++index; |
|
} else if (isLineTerminator(ch)) { |
|
++index; |
|
if (ch === 13 && source.charCodeAt(index) === 10) { |
|
++index; |
|
} |
|
++lineNumber; |
|
lineStart = index; |
|
} else if (ch === 47) { // 47 is '/' |
|
ch = source.charCodeAt(index + 1); |
|
if (ch === 47) { |
|
++index; |
|
++index; |
|
skipSingleLineComment(); |
|
} else if (ch === 42) { // 42 is '*' |
|
++index; |
|
++index; |
|
skipMultiLineComment(); |
|
} else { |
|
break; |
|
} |
|
} else { |
|
break; |
|
} |
|
} |
|
} |
|
|
|
function scanHexEscape(prefix) { |
|
var i, len, ch, code = 0; |
|
|
|
len = (prefix === 'u') ? 4 : 2; |
|
for (i = 0; i < len; ++i) { |
|
if (index < length && isHexDigit(source[index])) { |
|
ch = source[index++]; |
|
code = code * 16 + '0123456789abcdef'.indexOf(ch.toLowerCase()); |
|
} else { |
|
return ''; |
|
} |
|
} |
|
return String.fromCharCode(code); |
|
} |
|
|
|
function scanUnicodeCodePointEscape() { |
|
var ch, code, cu1, cu2; |
|
|
|
ch = source[index]; |
|
code = 0; |
|
|
|
// At least, one hex digit is required. |
|
if (ch === '}') { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
while (index < length) { |
|
ch = source[index++]; |
|
if (!isHexDigit(ch)) { |
|
break; |
|
} |
|
code = code * 16 + '0123456789abcdef'.indexOf(ch.toLowerCase()); |
|
} |
|
|
|
if (code > 0x10FFFF || ch !== '}') { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
// UTF-16 Encoding |
|
if (code <= 0xFFFF) { |
|
return String.fromCharCode(code); |
|
} |
|
cu1 = ((code - 0x10000) >> 10) + 0xD800; |
|
cu2 = ((code - 0x10000) & 1023) + 0xDC00; |
|
return String.fromCharCode(cu1, cu2); |
|
} |
|
|
|
function getEscapedIdentifier() { |
|
var ch, id; |
|
|
|
ch = source.charCodeAt(index++); |
|
id = String.fromCharCode(ch); |
|
|
|
// '\u' (char #92, char #117) denotes an escaped character. |
|
if (ch === 92) { |
|
if (source.charCodeAt(index) !== 117) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
++index; |
|
ch = scanHexEscape('u'); |
|
if (!ch || ch === '\\' || !isIdentifierStart(ch.charCodeAt(0))) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
id = ch; |
|
} |
|
|
|
while (index < length) { |
|
ch = source.charCodeAt(index); |
|
if (!isIdentifierPart(ch)) { |
|
break; |
|
} |
|
++index; |
|
id += String.fromCharCode(ch); |
|
|
|
// '\u' (char #92, char #117) denotes an escaped character. |
|
if (ch === 92) { |
|
id = id.substr(0, id.length - 1); |
|
if (source.charCodeAt(index) !== 117) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
++index; |
|
ch = scanHexEscape('u'); |
|
if (!ch || ch === '\\' || !isIdentifierPart(ch.charCodeAt(0))) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
id += ch; |
|
} |
|
} |
|
|
|
return id; |
|
} |
|
|
|
function getIdentifier() { |
|
var start, ch; |
|
|
|
start = index++; |
|
while (index < length) { |
|
ch = source.charCodeAt(index); |
|
if (ch === 92) { |
|
// Blackslash (char #92) marks Unicode escape sequence. |
|
index = start; |
|
return getEscapedIdentifier(); |
|
} |
|
if (isIdentifierPart(ch)) { |
|
++index; |
|
} else { |
|
break; |
|
} |
|
} |
|
|
|
return source.slice(start, index); |
|
} |
|
|
|
function scanIdentifier() { |
|
var start, id, type; |
|
|
|
start = index; |
|
|
|
// Backslash (char #92) starts an escaped character. |
|
id = (source.charCodeAt(index) === 92) ? getEscapedIdentifier() : getIdentifier(); |
|
|
|
// There is no keyword or literal with only one character. |
|
// Thus, it must be an identifier. |
|
if (id.length === 1) { |
|
type = Token.Identifier; |
|
} else if (isKeyword(id)) { |
|
type = Token.Keyword; |
|
} else if (id === 'null') { |
|
type = Token.NullLiteral; |
|
} else if (id === 'true' || id === 'false') { |
|
type = Token.BooleanLiteral; |
|
} else { |
|
type = Token.Identifier; |
|
} |
|
|
|
return { |
|
type: type, |
|
value: id, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
|
|
// 7.7 Punctuators |
|
|
|
function scanPunctuator() { |
|
var start = index, |
|
code = source.charCodeAt(index), |
|
code2, |
|
ch1 = source[index], |
|
ch2, |
|
ch3, |
|
ch4; |
|
|
|
if (state.inJSXTag || state.inJSXChild) { |
|
// Don't need to check for '{' and '}' as it's already handled |
|
// correctly by default. |
|
switch (code) { |
|
case 60: // < |
|
case 62: // > |
|
++index; |
|
return { |
|
type: Token.Punctuator, |
|
value: String.fromCharCode(code), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
} |
|
|
|
switch (code) { |
|
// Check for most common single-character punctuators. |
|
case 40: // ( open bracket |
|
case 41: // ) close bracket |
|
case 59: // ; semicolon |
|
case 44: // , comma |
|
case 123: // { open curly brace |
|
case 125: // } close curly brace |
|
case 91: // [ |
|
case 93: // ] |
|
case 58: // : |
|
case 63: // ? |
|
case 126: // ~ |
|
++index; |
|
if (extra.tokenize) { |
|
if (code === 40) { |
|
extra.openParenToken = extra.tokens.length; |
|
} else if (code === 123) { |
|
extra.openCurlyToken = extra.tokens.length; |
|
} |
|
} |
|
return { |
|
type: Token.Punctuator, |
|
value: String.fromCharCode(code), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
|
|
default: |
|
code2 = source.charCodeAt(index + 1); |
|
|
|
// '=' (char #61) marks an assignment or comparison operator. |
|
if (code2 === 61) { |
|
switch (code) { |
|
case 37: // % |
|
case 38: // & |
|
case 42: // *: |
|
case 43: // + |
|
case 45: // - |
|
case 47: // / |
|
case 60: // < |
|
case 62: // > |
|
case 94: // ^ |
|
case 124: // | |
|
index += 2; |
|
return { |
|
type: Token.Punctuator, |
|
value: String.fromCharCode(code) + String.fromCharCode(code2), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
|
|
case 33: // ! |
|
case 61: // = |
|
index += 2; |
|
|
|
// !== and === |
|
if (source.charCodeAt(index) === 61) { |
|
++index; |
|
} |
|
return { |
|
type: Token.Punctuator, |
|
value: source.slice(start, index), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
default: |
|
break; |
|
} |
|
} |
|
break; |
|
} |
|
|
|
// Peek more characters. |
|
|
|
ch2 = source[index + 1]; |
|
ch3 = source[index + 2]; |
|
ch4 = source[index + 3]; |
|
|
|
// 4-character punctuator: >>>= |
|
|
|
if (ch1 === '>' && ch2 === '>' && ch3 === '>') { |
|
if (ch4 === '=') { |
|
index += 4; |
|
return { |
|
type: Token.Punctuator, |
|
value: '>>>=', |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
} |
|
|
|
// 3-character punctuators: === !== >>> <<= >>= |
|
|
|
if (ch1 === '>' && ch2 === '>' && ch3 === '>' && !state.inType) { |
|
index += 3; |
|
return { |
|
type: Token.Punctuator, |
|
value: '>>>', |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
if (ch1 === '<' && ch2 === '<' && ch3 === '=') { |
|
index += 3; |
|
return { |
|
type: Token.Punctuator, |
|
value: '<<=', |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
if (ch1 === '>' && ch2 === '>' && ch3 === '=') { |
|
index += 3; |
|
return { |
|
type: Token.Punctuator, |
|
value: '>>=', |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
if (ch1 === '.' && ch2 === '.' && ch3 === '.') { |
|
index += 3; |
|
return { |
|
type: Token.Punctuator, |
|
value: '...', |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
// Other 2-character punctuators: ++ -- << >> && || |
|
|
|
// Don't match these tokens if we're in a type, since they never can |
|
// occur and can mess up types like Map<string, Array<string>> |
|
if (ch1 === ch2 && ('+-<>&|'.indexOf(ch1) >= 0) && !state.inType) { |
|
index += 2; |
|
return { |
|
type: Token.Punctuator, |
|
value: ch1 + ch2, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
if (ch1 === '=' && ch2 === '>') { |
|
index += 2; |
|
return { |
|
type: Token.Punctuator, |
|
value: '=>', |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
if ('<>=!+-*%&|^/'.indexOf(ch1) >= 0) { |
|
++index; |
|
return { |
|
type: Token.Punctuator, |
|
value: ch1, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
if (ch1 === '.') { |
|
++index; |
|
return { |
|
type: Token.Punctuator, |
|
value: ch1, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
// 7.8.3 Numeric Literals |
|
|
|
function scanHexLiteral(start) { |
|
var number = ''; |
|
|
|
while (index < length) { |
|
if (!isHexDigit(source[index])) { |
|
break; |
|
} |
|
number += source[index++]; |
|
} |
|
|
|
if (number.length === 0) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
if (isIdentifierStart(source.charCodeAt(index))) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
return { |
|
type: Token.NumericLiteral, |
|
value: parseInt('0x' + number, 16), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanBinaryLiteral(start) { |
|
var ch, number; |
|
|
|
number = ''; |
|
|
|
while (index < length) { |
|
ch = source[index]; |
|
if (ch !== '0' && ch !== '1') { |
|
break; |
|
} |
|
number += source[index++]; |
|
} |
|
|
|
if (number.length === 0) { |
|
// only 0b or 0B |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
if (index < length) { |
|
ch = source.charCodeAt(index); |
|
/* istanbul ignore else */ |
|
if (isIdentifierStart(ch) || isDecimalDigit(ch)) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
} |
|
|
|
return { |
|
type: Token.NumericLiteral, |
|
value: parseInt(number, 2), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanOctalLiteral(prefix, start) { |
|
var number, octal; |
|
|
|
if (isOctalDigit(prefix)) { |
|
octal = true; |
|
number = '0' + source[index++]; |
|
} else { |
|
octal = false; |
|
++index; |
|
number = ''; |
|
} |
|
|
|
while (index < length) { |
|
if (!isOctalDigit(source[index])) { |
|
break; |
|
} |
|
number += source[index++]; |
|
} |
|
|
|
if (!octal && number.length === 0) { |
|
// only 0o or 0O |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
if (isIdentifierStart(source.charCodeAt(index)) || isDecimalDigit(source.charCodeAt(index))) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
return { |
|
type: Token.NumericLiteral, |
|
value: parseInt(number, 8), |
|
octal: octal, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanNumericLiteral() { |
|
var number, start, ch; |
|
|
|
ch = source[index]; |
|
assert(isDecimalDigit(ch.charCodeAt(0)) || (ch === '.'), |
|
'Numeric literal must start with a decimal digit or a decimal point'); |
|
|
|
start = index; |
|
number = ''; |
|
if (ch !== '.') { |
|
number = source[index++]; |
|
ch = source[index]; |
|
|
|
// Hex number starts with '0x'. |
|
// Octal number starts with '0'. |
|
// Octal number in ES6 starts with '0o'. |
|
// Binary number in ES6 starts with '0b'. |
|
if (number === '0') { |
|
if (ch === 'x' || ch === 'X') { |
|
++index; |
|
return scanHexLiteral(start); |
|
} |
|
if (ch === 'b' || ch === 'B') { |
|
++index; |
|
return scanBinaryLiteral(start); |
|
} |
|
if (ch === 'o' || ch === 'O' || isOctalDigit(ch)) { |
|
return scanOctalLiteral(ch, start); |
|
} |
|
// decimal number starts with '0' such as '09' is illegal. |
|
if (ch && isDecimalDigit(ch.charCodeAt(0))) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
} |
|
|
|
while (isDecimalDigit(source.charCodeAt(index))) { |
|
number += source[index++]; |
|
} |
|
ch = source[index]; |
|
} |
|
|
|
if (ch === '.') { |
|
number += source[index++]; |
|
while (isDecimalDigit(source.charCodeAt(index))) { |
|
number += source[index++]; |
|
} |
|
ch = source[index]; |
|
} |
|
|
|
if (ch === 'e' || ch === 'E') { |
|
number += source[index++]; |
|
|
|
ch = source[index]; |
|
if (ch === '+' || ch === '-') { |
|
number += source[index++]; |
|
} |
|
if (isDecimalDigit(source.charCodeAt(index))) { |
|
while (isDecimalDigit(source.charCodeAt(index))) { |
|
number += source[index++]; |
|
} |
|
} else { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
} |
|
|
|
if (isIdentifierStart(source.charCodeAt(index))) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
return { |
|
type: Token.NumericLiteral, |
|
value: parseFloat(number), |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
// 7.8.4 String Literals |
|
|
|
function scanStringLiteral() { |
|
var str = '', quote, start, ch, code, unescaped, restore, octal = false; |
|
|
|
quote = source[index]; |
|
assert((quote === '\'' || quote === '"'), |
|
'String literal must starts with a quote'); |
|
|
|
start = index; |
|
++index; |
|
|
|
while (index < length) { |
|
ch = source[index++]; |
|
|
|
if (ch === quote) { |
|
quote = ''; |
|
break; |
|
} else if (ch === '\\') { |
|
ch = source[index++]; |
|
if (!ch || !isLineTerminator(ch.charCodeAt(0))) { |
|
switch (ch) { |
|
case 'n': |
|
str += '\n'; |
|
break; |
|
case 'r': |
|
str += '\r'; |
|
break; |
|
case 't': |
|
str += '\t'; |
|
break; |
|
case 'u': |
|
case 'x': |
|
if (source[index] === '{') { |
|
++index; |
|
str += scanUnicodeCodePointEscape(); |
|
} else { |
|
restore = index; |
|
unescaped = scanHexEscape(ch); |
|
if (unescaped) { |
|
str += unescaped; |
|
} else { |
|
index = restore; |
|
str += ch; |
|
} |
|
} |
|
break; |
|
case 'b': |
|
str += '\b'; |
|
break; |
|
case 'f': |
|
str += '\f'; |
|
break; |
|
case 'v': |
|
str += '\x0B'; |
|
break; |
|
|
|
default: |
|
if (isOctalDigit(ch)) { |
|
code = '01234567'.indexOf(ch); |
|
|
|
// \0 is not octal escape sequence |
|
if (code !== 0) { |
|
octal = true; |
|
} |
|
|
|
/* istanbul ignore else */ |
|
if (index < length && isOctalDigit(source[index])) { |
|
octal = true; |
|
code = code * 8 + '01234567'.indexOf(source[index++]); |
|
|
|
// 3 digits are only allowed when string starts |
|
// with 0, 1, 2, 3 |
|
if ('0123'.indexOf(ch) >= 0 && |
|
index < length && |
|
isOctalDigit(source[index])) { |
|
code = code * 8 + '01234567'.indexOf(source[index++]); |
|
} |
|
} |
|
str += String.fromCharCode(code); |
|
} else { |
|
str += ch; |
|
} |
|
break; |
|
} |
|
} else { |
|
++lineNumber; |
|
if (ch === '\r' && source[index] === '\n') { |
|
++index; |
|
} |
|
lineStart = index; |
|
} |
|
} else if (isLineTerminator(ch.charCodeAt(0))) { |
|
break; |
|
} else { |
|
str += ch; |
|
} |
|
} |
|
|
|
if (quote !== '') { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
return { |
|
type: Token.StringLiteral, |
|
value: str, |
|
octal: octal, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanTemplate() { |
|
var cooked = '', ch, start, terminated, tail, restore, unescaped, code, octal; |
|
|
|
terminated = false; |
|
tail = false; |
|
start = index; |
|
|
|
++index; |
|
|
|
while (index < length) { |
|
ch = source[index++]; |
|
if (ch === '`') { |
|
tail = true; |
|
terminated = true; |
|
break; |
|
} else if (ch === '$') { |
|
if (source[index] === '{') { |
|
++index; |
|
terminated = true; |
|
break; |
|
} |
|
cooked += ch; |
|
} else if (ch === '\\') { |
|
ch = source[index++]; |
|
if (!isLineTerminator(ch.charCodeAt(0))) { |
|
switch (ch) { |
|
case 'n': |
|
cooked += '\n'; |
|
break; |
|
case 'r': |
|
cooked += '\r'; |
|
break; |
|
case 't': |
|
cooked += '\t'; |
|
break; |
|
case 'u': |
|
case 'x': |
|
if (source[index] === '{') { |
|
++index; |
|
cooked += scanUnicodeCodePointEscape(); |
|
} else { |
|
restore = index; |
|
unescaped = scanHexEscape(ch); |
|
if (unescaped) { |
|
cooked += unescaped; |
|
} else { |
|
index = restore; |
|
cooked += ch; |
|
} |
|
} |
|
break; |
|
case 'b': |
|
cooked += '\b'; |
|
break; |
|
case 'f': |
|
cooked += '\f'; |
|
break; |
|
case 'v': |
|
cooked += '\v'; |
|
break; |
|
|
|
default: |
|
if (isOctalDigit(ch)) { |
|
code = '01234567'.indexOf(ch); |
|
|
|
// \0 is not octal escape sequence |
|
if (code !== 0) { |
|
octal = true; |
|
} |
|
|
|
/* istanbul ignore else */ |
|
if (index < length && isOctalDigit(source[index])) { |
|
octal = true; |
|
code = code * 8 + '01234567'.indexOf(source[index++]); |
|
|
|
// 3 digits are only allowed when string starts |
|
// with 0, 1, 2, 3 |
|
if ('0123'.indexOf(ch) >= 0 && |
|
index < length && |
|
isOctalDigit(source[index])) { |
|
code = code * 8 + '01234567'.indexOf(source[index++]); |
|
} |
|
} |
|
cooked += String.fromCharCode(code); |
|
} else { |
|
cooked += ch; |
|
} |
|
break; |
|
} |
|
} else { |
|
++lineNumber; |
|
if (ch === '\r' && source[index] === '\n') { |
|
++index; |
|
} |
|
lineStart = index; |
|
} |
|
} else if (isLineTerminator(ch.charCodeAt(0))) { |
|
++lineNumber; |
|
if (ch === '\r' && source[index] === '\n') { |
|
++index; |
|
} |
|
lineStart = index; |
|
cooked += '\n'; |
|
} else { |
|
cooked += ch; |
|
} |
|
} |
|
|
|
if (!terminated) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
return { |
|
type: Token.Template, |
|
value: { |
|
cooked: cooked, |
|
raw: source.slice(start + 1, index - ((tail) ? 1 : 2)) |
|
}, |
|
tail: tail, |
|
octal: octal, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanTemplateElement(option) { |
|
var startsWith, template; |
|
|
|
lookahead = null; |
|
skipComment(); |
|
|
|
startsWith = (option.head) ? '`' : '}'; |
|
|
|
if (source[index] !== startsWith) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
template = scanTemplate(); |
|
|
|
peek(); |
|
|
|
return template; |
|
} |
|
|
|
function testRegExp(pattern, flags) { |
|
var tmp = pattern, |
|
value; |
|
|
|
if (flags.indexOf('u') >= 0) { |
|
// Replace each astral symbol and every Unicode code point |
|
// escape sequence with a single ASCII symbol to avoid throwing on |
|
// regular expressions that are only valid in combination with the |
|
// `/u` flag. |
|
// Note: replacing with the ASCII symbol `x` might cause false |
|
// negatives in unlikely scenarios. For example, `[\u{61}-b]` is a |
|
// perfectly valid pattern that is equivalent to `[a-b]`, but it |
|
// would be replaced by `[x-b]` which throws an error. |
|
tmp = tmp |
|
.replace(/\\u\{([0-9a-fA-F]+)\}/g, function ($0, $1) { |
|
if (parseInt($1, 16) <= 0x10FFFF) { |
|
return 'x'; |
|
} |
|
throwError({}, Messages.InvalidRegExp); |
|
}) |
|
.replace(/[\uD800-\uDBFF][\uDC00-\uDFFF]/g, 'x'); |
|
} |
|
|
|
// First, detect invalid regular expressions. |
|
try { |
|
value = new RegExp(tmp); |
|
} catch (e) { |
|
throwError({}, Messages.InvalidRegExp); |
|
} |
|
|
|
// Return a regular expression object for this pattern-flag pair, or |
|
// `null` in case the current environment doesn't support the flags it |
|
// uses. |
|
try { |
|
return new RegExp(pattern, flags); |
|
} catch (exception) { |
|
return null; |
|
} |
|
} |
|
|
|
function scanRegExpBody() { |
|
var ch, str, classMarker, terminated, body; |
|
|
|
ch = source[index]; |
|
assert(ch === '/', 'Regular expression literal must start with a slash'); |
|
str = source[index++]; |
|
|
|
classMarker = false; |
|
terminated = false; |
|
while (index < length) { |
|
ch = source[index++]; |
|
str += ch; |
|
if (ch === '\\') { |
|
ch = source[index++]; |
|
// ECMA-262 7.8.5 |
|
if (isLineTerminator(ch.charCodeAt(0))) { |
|
throwError({}, Messages.UnterminatedRegExp); |
|
} |
|
str += ch; |
|
} else if (isLineTerminator(ch.charCodeAt(0))) { |
|
throwError({}, Messages.UnterminatedRegExp); |
|
} else if (classMarker) { |
|
if (ch === ']') { |
|
classMarker = false; |
|
} |
|
} else { |
|
if (ch === '/') { |
|
terminated = true; |
|
break; |
|
} else if (ch === '[') { |
|
classMarker = true; |
|
} |
|
} |
|
} |
|
|
|
if (!terminated) { |
|
throwError({}, Messages.UnterminatedRegExp); |
|
} |
|
|
|
// Exclude leading and trailing slash. |
|
body = str.substr(1, str.length - 2); |
|
return { |
|
value: body, |
|
literal: str |
|
}; |
|
} |
|
|
|
function scanRegExpFlags() { |
|
var ch, str, flags, restore; |
|
|
|
str = ''; |
|
flags = ''; |
|
while (index < length) { |
|
ch = source[index]; |
|
if (!isIdentifierPart(ch.charCodeAt(0))) { |
|
break; |
|
} |
|
|
|
++index; |
|
if (ch === '\\' && index < length) { |
|
ch = source[index]; |
|
if (ch === 'u') { |
|
++index; |
|
restore = index; |
|
ch = scanHexEscape('u'); |
|
if (ch) { |
|
flags += ch; |
|
for (str += '\\u'; restore < index; ++restore) { |
|
str += source[restore]; |
|
} |
|
} else { |
|
index = restore; |
|
flags += 'u'; |
|
str += '\\u'; |
|
} |
|
throwErrorTolerant({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} else { |
|
str += '\\'; |
|
throwErrorTolerant({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
} else { |
|
flags += ch; |
|
str += ch; |
|
} |
|
} |
|
|
|
return { |
|
value: flags, |
|
literal: str |
|
}; |
|
} |
|
|
|
function scanRegExp() { |
|
var start, body, flags, value; |
|
|
|
lookahead = null; |
|
skipComment(); |
|
start = index; |
|
|
|
body = scanRegExpBody(); |
|
flags = scanRegExpFlags(); |
|
value = testRegExp(body.value, flags.value); |
|
|
|
if (extra.tokenize) { |
|
return { |
|
type: Token.RegularExpression, |
|
value: value, |
|
regex: { |
|
pattern: body.value, |
|
flags: flags.value |
|
}, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
return { |
|
literal: body.literal + flags.literal, |
|
value: value, |
|
regex: { |
|
pattern: body.value, |
|
flags: flags.value |
|
}, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function isIdentifierName(token) { |
|
return token.type === Token.Identifier || |
|
token.type === Token.Keyword || |
|
token.type === Token.BooleanLiteral || |
|
token.type === Token.NullLiteral; |
|
} |
|
|
|
function advanceSlash() { |
|
var prevToken, |
|
checkToken; |
|
// Using the following algorithm: |
|
// https://github.com/mozilla/sweet.js/wiki/design |
|
prevToken = extra.tokens[extra.tokens.length - 1]; |
|
if (!prevToken) { |
|
// Nothing before that: it cannot be a division. |
|
return scanRegExp(); |
|
} |
|
if (prevToken.type === 'Punctuator') { |
|
if (prevToken.value === ')') { |
|
checkToken = extra.tokens[extra.openParenToken - 1]; |
|
if (checkToken && |
|
checkToken.type === 'Keyword' && |
|
(checkToken.value === 'if' || |
|
checkToken.value === 'while' || |
|
checkToken.value === 'for' || |
|
checkToken.value === 'with')) { |
|
return scanRegExp(); |
|
} |
|
return scanPunctuator(); |
|
} |
|
if (prevToken.value === '}') { |
|
// Dividing a function by anything makes little sense, |
|
// but we have to check for that. |
|
if (extra.tokens[extra.openCurlyToken - 3] && |
|
extra.tokens[extra.openCurlyToken - 3].type === 'Keyword') { |
|
// Anonymous function. |
|
checkToken = extra.tokens[extra.openCurlyToken - 4]; |
|
if (!checkToken) { |
|
return scanPunctuator(); |
|
} |
|
} else if (extra.tokens[extra.openCurlyToken - 4] && |
|
extra.tokens[extra.openCurlyToken - 4].type === 'Keyword') { |
|
// Named function. |
|
checkToken = extra.tokens[extra.openCurlyToken - 5]; |
|
if (!checkToken) { |
|
return scanRegExp(); |
|
} |
|
} else { |
|
return scanPunctuator(); |
|
} |
|
// checkToken determines whether the function is |
|
// a declaration or an expression. |
|
if (FnExprTokens.indexOf(checkToken.value) >= 0) { |
|
// It is an expression. |
|
return scanPunctuator(); |
|
} |
|
// It is a declaration. |
|
return scanRegExp(); |
|
} |
|
return scanRegExp(); |
|
} |
|
if (prevToken.type === 'Keyword' && prevToken.value !== 'this') { |
|
return scanRegExp(); |
|
} |
|
return scanPunctuator(); |
|
} |
|
|
|
function advance() { |
|
var ch; |
|
|
|
if (!state.inJSXChild) { |
|
skipComment(); |
|
} |
|
|
|
if (index >= length) { |
|
return { |
|
type: Token.EOF, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [index, index] |
|
}; |
|
} |
|
|
|
if (state.inJSXChild) { |
|
return advanceJSXChild(); |
|
} |
|
|
|
ch = source.charCodeAt(index); |
|
|
|
// Very common: ( and ) and ; |
|
if (ch === 40 || ch === 41 || ch === 58) { |
|
return scanPunctuator(); |
|
} |
|
|
|
// String literal starts with single quote (#39) or double quote (#34). |
|
if (ch === 39 || ch === 34) { |
|
if (state.inJSXTag) { |
|
return scanJSXStringLiteral(); |
|
} |
|
return scanStringLiteral(); |
|
} |
|
|
|
if (state.inJSXTag && isJSXIdentifierStart(ch)) { |
|
return scanJSXIdentifier(); |
|
} |
|
|
|
if (ch === 96) { |
|
return scanTemplate(); |
|
} |
|
if (isIdentifierStart(ch)) { |
|
return scanIdentifier(); |
|
} |
|
|
|
// Dot (.) char #46 can also start a floating-point number, hence the need |
|
// to check the next character. |
|
if (ch === 46) { |
|
if (isDecimalDigit(source.charCodeAt(index + 1))) { |
|
return scanNumericLiteral(); |
|
} |
|
return scanPunctuator(); |
|
} |
|
|
|
if (isDecimalDigit(ch)) { |
|
return scanNumericLiteral(); |
|
} |
|
|
|
// Slash (/) char #47 can also start a regex. |
|
if (extra.tokenize && ch === 47) { |
|
return advanceSlash(); |
|
} |
|
|
|
return scanPunctuator(); |
|
} |
|
|
|
function lex() { |
|
var token; |
|
|
|
token = lookahead; |
|
index = token.range[1]; |
|
lineNumber = token.lineNumber; |
|
lineStart = token.lineStart; |
|
|
|
lookahead = advance(); |
|
|
|
index = token.range[1]; |
|
lineNumber = token.lineNumber; |
|
lineStart = token.lineStart; |
|
|
|
return token; |
|
} |
|
|
|
function peek() { |
|
var pos, line, start; |
|
|
|
pos = index; |
|
line = lineNumber; |
|
start = lineStart; |
|
lookahead = advance(); |
|
index = pos; |
|
lineNumber = line; |
|
lineStart = start; |
|
} |
|
|
|
function lookahead2() { |
|
var adv, pos, line, start, result; |
|
|
|
// If we are collecting the tokens, don't grab the next one yet. |
|
/* istanbul ignore next */ |
|
adv = (typeof extra.advance === 'function') ? extra.advance : advance; |
|
|
|
pos = index; |
|
line = lineNumber; |
|
start = lineStart; |
|
|
|
// Scan for the next immediate token. |
|
/* istanbul ignore if */ |
|
if (lookahead === null) { |
|
lookahead = adv(); |
|
} |
|
index = lookahead.range[1]; |
|
lineNumber = lookahead.lineNumber; |
|
lineStart = lookahead.lineStart; |
|
|
|
// Grab the token right after. |
|
result = adv(); |
|
index = pos; |
|
lineNumber = line; |
|
lineStart = start; |
|
|
|
return result; |
|
} |
|
|
|
function rewind(token) { |
|
index = token.range[0]; |
|
lineNumber = token.lineNumber; |
|
lineStart = token.lineStart; |
|
lookahead = token; |
|
} |
|
|
|
function markerCreate() { |
|
if (!extra.loc && !extra.range) { |
|
return undefined; |
|
} |
|
skipComment(); |
|
return {offset: index, line: lineNumber, col: index - lineStart}; |
|
} |
|
|
|
function markerCreatePreserveWhitespace() { |
|
if (!extra.loc && !extra.range) { |
|
return undefined; |
|
} |
|
return {offset: index, line: lineNumber, col: index - lineStart}; |
|
} |
|
|
|
function processComment(node) { |
|
var lastChild, |
|
trailingComments, |
|
bottomRight = extra.bottomRightStack, |
|
last = bottomRight[bottomRight.length - 1]; |
|
|
|
if (node.type === Syntax.Program) { |
|
/* istanbul ignore else */ |
|
if (node.body.length > 0) { |
|
return; |
|
} |
|
} |
|
|
|
if (extra.trailingComments.length > 0) { |
|
if (extra.trailingComments[0].range[0] >= node.range[1]) { |
|
trailingComments = extra.trailingComments; |
|
extra.trailingComments = []; |
|
} else { |
|
extra.trailingComments.length = 0; |
|
} |
|
} else { |
|
if (last && last.trailingComments && last.trailingComments[0].range[0] >= node.range[1]) { |
|
trailingComments = last.trailingComments; |
|
delete last.trailingComments; |
|
} |
|
} |
|
|
|
// Eating the stack. |
|
if (last) { |
|
while (last && last.range[0] >= node.range[0]) { |
|
lastChild = last; |
|
last = bottomRight.pop(); |
|
} |
|
} |
|
|
|
if (lastChild) { |
|
if (lastChild.leadingComments && lastChild.leadingComments[lastChild.leadingComments.length - 1].range[1] <= node.range[0]) { |
|
node.leadingComments = lastChild.leadingComments; |
|
delete lastChild.leadingComments; |
|
} |
|
} else if (extra.leadingComments.length > 0 && extra.leadingComments[extra.leadingComments.length - 1].range[1] <= node.range[0]) { |
|
node.leadingComments = extra.leadingComments; |
|
extra.leadingComments = []; |
|
} |
|
|
|
if (trailingComments) { |
|
node.trailingComments = trailingComments; |
|
} |
|
|
|
bottomRight.push(node); |
|
} |
|
|
|
function markerApply(marker, node) { |
|
if (extra.range) { |
|
node.range = [marker.offset, index]; |
|
} |
|
if (extra.loc) { |
|
node.loc = { |
|
start: { |
|
line: marker.line, |
|
column: marker.col |
|
}, |
|
end: { |
|
line: lineNumber, |
|
column: index - lineStart |
|
} |
|
}; |
|
node = delegate.postProcess(node); |
|
} |
|
if (extra.attachComment) { |
|
processComment(node); |
|
} |
|
return node; |
|
} |
|
|
|
SyntaxTreeDelegate = { |
|
|
|
name: 'SyntaxTree', |
|
|
|
postProcess: function (node) { |
|
return node; |
|
}, |
|
|
|
createArrayExpression: function (elements) { |
|
return { |
|
type: Syntax.ArrayExpression, |
|
elements: elements |
|
}; |
|
}, |
|
|
|
createAssignmentExpression: function (operator, left, right) { |
|
return { |
|
type: Syntax.AssignmentExpression, |
|
operator: operator, |
|
left: left, |
|
right: right |
|
}; |
|
}, |
|
|
|
createBinaryExpression: function (operator, left, right) { |
|
var type = (operator === '||' || operator === '&&') ? Syntax.LogicalExpression : |
|
Syntax.BinaryExpression; |
|
return { |
|
type: type, |
|
operator: operator, |
|
left: left, |
|
right: right |
|
}; |
|
}, |
|
|
|
createBlockStatement: function (body) { |
|
return { |
|
type: Syntax.BlockStatement, |
|
body: body |
|
}; |
|
}, |
|
|
|
createBreakStatement: function (label) { |
|
return { |
|
type: Syntax.BreakStatement, |
|
label: label |
|
}; |
|
}, |
|
|
|
createCallExpression: function (callee, args) { |
|
return { |
|
type: Syntax.CallExpression, |
|
callee: callee, |
|
'arguments': args |
|
}; |
|
}, |
|
|
|
createCatchClause: function (param, body) { |
|
return { |
|
type: Syntax.CatchClause, |
|
param: param, |
|
body: body |
|
}; |
|
}, |
|
|
|
createConditionalExpression: function (test, consequent, alternate) { |
|
return { |
|
type: Syntax.ConditionalExpression, |
|
test: test, |
|
consequent: consequent, |
|
alternate: alternate |
|
}; |
|
}, |
|
|
|
createContinueStatement: function (label) { |
|
return { |
|
type: Syntax.ContinueStatement, |
|
label: label |
|
}; |
|
}, |
|
|
|
createDebuggerStatement: function () { |
|
return { |
|
type: Syntax.DebuggerStatement |
|
}; |
|
}, |
|
|
|
createDoWhileStatement: function (body, test) { |
|
return { |
|
type: Syntax.DoWhileStatement, |
|
body: body, |
|
test: test |
|
}; |
|
}, |
|
|
|
createEmptyStatement: function () { |
|
return { |
|
type: Syntax.EmptyStatement |
|
}; |
|
}, |
|
|
|
createExpressionStatement: function (expression) { |
|
return { |
|
type: Syntax.ExpressionStatement, |
|
expression: expression |
|
}; |
|
}, |
|
|
|
createForStatement: function (init, test, update, body) { |
|
return { |
|
type: Syntax.ForStatement, |
|
init: init, |
|
test: test, |
|
update: update, |
|
body: body |
|
}; |
|
}, |
|
|
|
createForInStatement: function (left, right, body) { |
|
return { |
|
type: Syntax.ForInStatement, |
|
left: left, |
|
right: right, |
|
body: body, |
|
each: false |
|
}; |
|
}, |
|
|
|
createForOfStatement: function (left, right, body) { |
|
return { |
|
type: Syntax.ForOfStatement, |
|
left: left, |
|
right: right, |
|
body: body |
|
}; |
|
}, |
|
|
|
createFunctionDeclaration: function (id, params, defaults, body, rest, generator, expression, |
|
isAsync, returnType, typeParameters) { |
|
var funDecl = { |
|
type: Syntax.FunctionDeclaration, |
|
id: id, |
|
params: params, |
|
defaults: defaults, |
|
body: body, |
|
rest: rest, |
|
generator: generator, |
|
expression: expression, |
|
returnType: returnType, |
|
typeParameters: typeParameters |
|
}; |
|
|
|
if (isAsync) { |
|
funDecl.async = true; |
|
} |
|
|
|
return funDecl; |
|
}, |
|
|
|
createFunctionExpression: function (id, params, defaults, body, rest, generator, expression, |
|
isAsync, returnType, typeParameters) { |
|
var funExpr = { |
|
type: Syntax.FunctionExpression, |
|
id: id, |
|
params: params, |
|
defaults: defaults, |
|
body: body, |
|
rest: rest, |
|
generator: generator, |
|
expression: expression, |
|
returnType: returnType, |
|
typeParameters: typeParameters |
|
}; |
|
|
|
if (isAsync) { |
|
funExpr.async = true; |
|
} |
|
|
|
return funExpr; |
|
}, |
|
|
|
createIdentifier: function (name) { |
|
return { |
|
type: Syntax.Identifier, |
|
name: name, |
|
// Only here to initialize the shape of the object to ensure |
|
// that the 'typeAnnotation' key is ordered before others that |
|
// are added later (like 'loc' and 'range'). This just helps |
|
// keep the shape of Identifier nodes consistent with everything |
|
// else. |
|
typeAnnotation: undefined, |
|
optional: undefined |
|
}; |
|
}, |
|
|
|
createTypeAnnotation: function (typeAnnotation) { |
|
return { |
|
type: Syntax.TypeAnnotation, |
|
typeAnnotation: typeAnnotation |
|
}; |
|
}, |
|
|
|
createTypeCast: function (expression, typeAnnotation) { |
|
return { |
|
type: Syntax.TypeCastExpression, |
|
expression: expression, |
|
typeAnnotation: typeAnnotation |
|
}; |
|
}, |
|
|
|
createFunctionTypeAnnotation: function (params, returnType, rest, typeParameters) { |
|
return { |
|
type: Syntax.FunctionTypeAnnotation, |
|
params: params, |
|
returnType: returnType, |
|
rest: rest, |
|
typeParameters: typeParameters |
|
}; |
|
}, |
|
|
|
createFunctionTypeParam: function (name, typeAnnotation, optional) { |
|
return { |
|
type: Syntax.FunctionTypeParam, |
|
name: name, |
|
typeAnnotation: typeAnnotation, |
|
optional: optional |
|
}; |
|
}, |
|
|
|
createNullableTypeAnnotation: function (typeAnnotation) { |
|
return { |
|
type: Syntax.NullableTypeAnnotation, |
|
typeAnnotation: typeAnnotation |
|
}; |
|
}, |
|
|
|
createArrayTypeAnnotation: function (elementType) { |
|
return { |
|
type: Syntax.ArrayTypeAnnotation, |
|
elementType: elementType |
|
}; |
|
}, |
|
|
|
createGenericTypeAnnotation: function (id, typeParameters) { |
|
return { |
|
type: Syntax.GenericTypeAnnotation, |
|
id: id, |
|
typeParameters: typeParameters |
|
}; |
|
}, |
|
|
|
createQualifiedTypeIdentifier: function (qualification, id) { |
|
return { |
|
type: Syntax.QualifiedTypeIdentifier, |
|
qualification: qualification, |
|
id: id |
|
}; |
|
}, |
|
|
|
createTypeParameterDeclaration: function (params) { |
|
return { |
|
type: Syntax.TypeParameterDeclaration, |
|
params: params |
|
}; |
|
}, |
|
|
|
createTypeParameterInstantiation: function (params) { |
|
return { |
|
type: Syntax.TypeParameterInstantiation, |
|
params: params |
|
}; |
|
}, |
|
|
|
createAnyTypeAnnotation: function () { |
|
return { |
|
type: Syntax.AnyTypeAnnotation |
|
}; |
|
}, |
|
|
|
createBooleanTypeAnnotation: function () { |
|
return { |
|
type: Syntax.BooleanTypeAnnotation |
|
}; |
|
}, |
|
|
|
createNumberTypeAnnotation: function () { |
|
return { |
|
type: Syntax.NumberTypeAnnotation |
|
}; |
|
}, |
|
|
|
createStringTypeAnnotation: function () { |
|
return { |
|
type: Syntax.StringTypeAnnotation |
|
}; |
|
}, |
|
|
|
createStringLiteralTypeAnnotation: function (token) { |
|
return { |
|
type: Syntax.StringLiteralTypeAnnotation, |
|
value: token.value, |
|
raw: source.slice(token.range[0], token.range[1]) |
|
}; |
|
}, |
|
|
|
createVoidTypeAnnotation: function () { |
|
return { |
|
type: Syntax.VoidTypeAnnotation |
|
}; |
|
}, |
|
|
|
createTypeofTypeAnnotation: function (argument) { |
|
return { |
|
type: Syntax.TypeofTypeAnnotation, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createTupleTypeAnnotation: function (types) { |
|
return { |
|
type: Syntax.TupleTypeAnnotation, |
|
types: types |
|
}; |
|
}, |
|
|
|
createObjectTypeAnnotation: function (properties, indexers, callProperties) { |
|
return { |
|
type: Syntax.ObjectTypeAnnotation, |
|
properties: properties, |
|
indexers: indexers, |
|
callProperties: callProperties |
|
}; |
|
}, |
|
|
|
createObjectTypeIndexer: function (id, key, value, isStatic) { |
|
return { |
|
type: Syntax.ObjectTypeIndexer, |
|
id: id, |
|
key: key, |
|
value: value, |
|
"static": isStatic |
|
}; |
|
}, |
|
|
|
createObjectTypeCallProperty: function (value, isStatic) { |
|
return { |
|
type: Syntax.ObjectTypeCallProperty, |
|
value: value, |
|
"static": isStatic |
|
}; |
|
}, |
|
|
|
createObjectTypeProperty: function (key, value, optional, isStatic) { |
|
return { |
|
type: Syntax.ObjectTypeProperty, |
|
key: key, |
|
value: value, |
|
optional: optional, |
|
"static": isStatic |
|
}; |
|
}, |
|
|
|
createUnionTypeAnnotation: function (types) { |
|
return { |
|
type: Syntax.UnionTypeAnnotation, |
|
types: types |
|
}; |
|
}, |
|
|
|
createIntersectionTypeAnnotation: function (types) { |
|
return { |
|
type: Syntax.IntersectionTypeAnnotation, |
|
types: types |
|
}; |
|
}, |
|
|
|
createTypeAlias: function (id, typeParameters, right) { |
|
return { |
|
type: Syntax.TypeAlias, |
|
id: id, |
|
typeParameters: typeParameters, |
|
right: right |
|
}; |
|
}, |
|
|
|
createInterface: function (id, typeParameters, body, extended) { |
|
return { |
|
type: Syntax.InterfaceDeclaration, |
|
id: id, |
|
typeParameters: typeParameters, |
|
body: body, |
|
"extends": extended |
|
}; |
|
}, |
|
|
|
createInterfaceExtends: function (id, typeParameters) { |
|
return { |
|
type: Syntax.InterfaceExtends, |
|
id: id, |
|
typeParameters: typeParameters |
|
}; |
|
}, |
|
|
|
createDeclareFunction: function (id) { |
|
return { |
|
type: Syntax.DeclareFunction, |
|
id: id |
|
}; |
|
}, |
|
|
|
createDeclareVariable: function (id) { |
|
return { |
|
type: Syntax.DeclareVariable, |
|
id: id |
|
}; |
|
}, |
|
|
|
createDeclareModule: function (id, body) { |
|
return { |
|
type: Syntax.DeclareModule, |
|
id: id, |
|
body: body |
|
}; |
|
}, |
|
|
|
createJSXAttribute: function (name, value) { |
|
return { |
|
type: Syntax.JSXAttribute, |
|
name: name, |
|
value: value || null |
|
}; |
|
}, |
|
|
|
createJSXSpreadAttribute: function (argument) { |
|
return { |
|
type: Syntax.JSXSpreadAttribute, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createJSXIdentifier: function (name) { |
|
return { |
|
type: Syntax.JSXIdentifier, |
|
name: name |
|
}; |
|
}, |
|
|
|
createJSXNamespacedName: function (namespace, name) { |
|
return { |
|
type: Syntax.JSXNamespacedName, |
|
namespace: namespace, |
|
name: name |
|
}; |
|
}, |
|
|
|
createJSXMemberExpression: function (object, property) { |
|
return { |
|
type: Syntax.JSXMemberExpression, |
|
object: object, |
|
property: property |
|
}; |
|
}, |
|
|
|
createJSXElement: function (openingElement, closingElement, children) { |
|
return { |
|
type: Syntax.JSXElement, |
|
openingElement: openingElement, |
|
closingElement: closingElement, |
|
children: children |
|
}; |
|
}, |
|
|
|
createJSXEmptyExpression: function () { |
|
return { |
|
type: Syntax.JSXEmptyExpression |
|
}; |
|
}, |
|
|
|
createJSXExpressionContainer: function (expression) { |
|
return { |
|
type: Syntax.JSXExpressionContainer, |
|
expression: expression |
|
}; |
|
}, |
|
|
|
createJSXOpeningElement: function (name, attributes, selfClosing) { |
|
return { |
|
type: Syntax.JSXOpeningElement, |
|
name: name, |
|
selfClosing: selfClosing, |
|
attributes: attributes |
|
}; |
|
}, |
|
|
|
createJSXClosingElement: function (name) { |
|
return { |
|
type: Syntax.JSXClosingElement, |
|
name: name |
|
}; |
|
}, |
|
|
|
createIfStatement: function (test, consequent, alternate) { |
|
return { |
|
type: Syntax.IfStatement, |
|
test: test, |
|
consequent: consequent, |
|
alternate: alternate |
|
}; |
|
}, |
|
|
|
createLabeledStatement: function (label, body) { |
|
return { |
|
type: Syntax.LabeledStatement, |
|
label: label, |
|
body: body |
|
}; |
|
}, |
|
|
|
createLiteral: function (token) { |
|
var object = { |
|
type: Syntax.Literal, |
|
value: token.value, |
|
raw: source.slice(token.range[0], token.range[1]) |
|
}; |
|
if (token.regex) { |
|
object.regex = token.regex; |
|
} |
|
return object; |
|
}, |
|
|
|
createMemberExpression: function (accessor, object, property) { |
|
return { |
|
type: Syntax.MemberExpression, |
|
computed: accessor === '[', |
|
object: object, |
|
property: property |
|
}; |
|
}, |
|
|
|
createNewExpression: function (callee, args) { |
|
return { |
|
type: Syntax.NewExpression, |
|
callee: callee, |
|
'arguments': args |
|
}; |
|
}, |
|
|
|
createObjectExpression: function (properties) { |
|
return { |
|
type: Syntax.ObjectExpression, |
|
properties: properties |
|
}; |
|
}, |
|
|
|
createPostfixExpression: function (operator, argument) { |
|
return { |
|
type: Syntax.UpdateExpression, |
|
operator: operator, |
|
argument: argument, |
|
prefix: false |
|
}; |
|
}, |
|
|
|
createProgram: function (body) { |
|
return { |
|
type: Syntax.Program, |
|
body: body |
|
}; |
|
}, |
|
|
|
createProperty: function (kind, key, value, method, shorthand, computed) { |
|
return { |
|
type: Syntax.Property, |
|
key: key, |
|
value: value, |
|
kind: kind, |
|
method: method, |
|
shorthand: shorthand, |
|
computed: computed |
|
}; |
|
}, |
|
|
|
createReturnStatement: function (argument) { |
|
return { |
|
type: Syntax.ReturnStatement, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createSequenceExpression: function (expressions) { |
|
return { |
|
type: Syntax.SequenceExpression, |
|
expressions: expressions |
|
}; |
|
}, |
|
|
|
createSwitchCase: function (test, consequent) { |
|
return { |
|
type: Syntax.SwitchCase, |
|
test: test, |
|
consequent: consequent |
|
}; |
|
}, |
|
|
|
createSwitchStatement: function (discriminant, cases) { |
|
return { |
|
type: Syntax.SwitchStatement, |
|
discriminant: discriminant, |
|
cases: cases |
|
}; |
|
}, |
|
|
|
createThisExpression: function () { |
|
return { |
|
type: Syntax.ThisExpression |
|
}; |
|
}, |
|
|
|
createThrowStatement: function (argument) { |
|
return { |
|
type: Syntax.ThrowStatement, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createTryStatement: function (block, guardedHandlers, handlers, finalizer) { |
|
return { |
|
type: Syntax.TryStatement, |
|
block: block, |
|
guardedHandlers: guardedHandlers, |
|
handlers: handlers, |
|
finalizer: finalizer |
|
}; |
|
}, |
|
|
|
createUnaryExpression: function (operator, argument) { |
|
if (operator === '++' || operator === '--') { |
|
return { |
|
type: Syntax.UpdateExpression, |
|
operator: operator, |
|
argument: argument, |
|
prefix: true |
|
}; |
|
} |
|
return { |
|
type: Syntax.UnaryExpression, |
|
operator: operator, |
|
argument: argument, |
|
prefix: true |
|
}; |
|
}, |
|
|
|
createVariableDeclaration: function (declarations, kind) { |
|
return { |
|
type: Syntax.VariableDeclaration, |
|
declarations: declarations, |
|
kind: kind |
|
}; |
|
}, |
|
|
|
createVariableDeclarator: function (id, init) { |
|
return { |
|
type: Syntax.VariableDeclarator, |
|
id: id, |
|
init: init |
|
}; |
|
}, |
|
|
|
createWhileStatement: function (test, body) { |
|
return { |
|
type: Syntax.WhileStatement, |
|
test: test, |
|
body: body |
|
}; |
|
}, |
|
|
|
createWithStatement: function (object, body) { |
|
return { |
|
type: Syntax.WithStatement, |
|
object: object, |
|
body: body |
|
}; |
|
}, |
|
|
|
createTemplateElement: function (value, tail) { |
|
return { |
|
type: Syntax.TemplateElement, |
|
value: value, |
|
tail: tail |
|
}; |
|
}, |
|
|
|
createTemplateLiteral: function (quasis, expressions) { |
|
return { |
|
type: Syntax.TemplateLiteral, |
|
quasis: quasis, |
|
expressions: expressions |
|
}; |
|
}, |
|
|
|
createSpreadElement: function (argument) { |
|
return { |
|
type: Syntax.SpreadElement, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createSpreadProperty: function (argument) { |
|
return { |
|
type: Syntax.SpreadProperty, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createTaggedTemplateExpression: function (tag, quasi) { |
|
return { |
|
type: Syntax.TaggedTemplateExpression, |
|
tag: tag, |
|
quasi: quasi |
|
}; |
|
}, |
|
|
|
createArrowFunctionExpression: function (params, defaults, body, rest, expression, isAsync) { |
|
var arrowExpr = { |
|
type: Syntax.ArrowFunctionExpression, |
|
id: null, |
|
params: params, |
|
defaults: defaults, |
|
body: body, |
|
rest: rest, |
|
generator: false, |
|
expression: expression |
|
}; |
|
|
|
if (isAsync) { |
|
arrowExpr.async = true; |
|
} |
|
|
|
return arrowExpr; |
|
}, |
|
|
|
createMethodDefinition: function (propertyType, kind, key, value, computed) { |
|
return { |
|
type: Syntax.MethodDefinition, |
|
key: key, |
|
value: value, |
|
kind: kind, |
|
'static': propertyType === ClassPropertyType["static"], |
|
computed: computed |
|
}; |
|
}, |
|
|
|
createClassProperty: function (key, typeAnnotation, computed, isStatic) { |
|
return { |
|
type: Syntax.ClassProperty, |
|
key: key, |
|
typeAnnotation: typeAnnotation, |
|
computed: computed, |
|
"static": isStatic |
|
}; |
|
}, |
|
|
|
createClassBody: function (body) { |
|
return { |
|
type: Syntax.ClassBody, |
|
body: body |
|
}; |
|
}, |
|
|
|
createClassImplements: function (id, typeParameters) { |
|
return { |
|
type: Syntax.ClassImplements, |
|
id: id, |
|
typeParameters: typeParameters |
|
}; |
|
}, |
|
|
|
createClassExpression: function (id, superClass, body, typeParameters, superTypeParameters, implemented) { |
|
return { |
|
type: Syntax.ClassExpression, |
|
id: id, |
|
superClass: superClass, |
|
body: body, |
|
typeParameters: typeParameters, |
|
superTypeParameters: superTypeParameters, |
|
"implements": implemented |
|
}; |
|
}, |
|
|
|
createClassDeclaration: function (id, superClass, body, typeParameters, superTypeParameters, implemented) { |
|
return { |
|
type: Syntax.ClassDeclaration, |
|
id: id, |
|
superClass: superClass, |
|
body: body, |
|
typeParameters: typeParameters, |
|
superTypeParameters: superTypeParameters, |
|
"implements": implemented |
|
}; |
|
}, |
|
|
|
createModuleSpecifier: function (token) { |
|
return { |
|
type: Syntax.ModuleSpecifier, |
|
value: token.value, |
|
raw: source.slice(token.range[0], token.range[1]) |
|
}; |
|
}, |
|
|
|
createExportSpecifier: function (id, name) { |
|
return { |
|
type: Syntax.ExportSpecifier, |
|
id: id, |
|
name: name |
|
}; |
|
}, |
|
|
|
createExportBatchSpecifier: function () { |
|
return { |
|
type: Syntax.ExportBatchSpecifier |
|
}; |
|
}, |
|
|
|
createImportDefaultSpecifier: function (id) { |
|
return { |
|
type: Syntax.ImportDefaultSpecifier, |
|
id: id |
|
}; |
|
}, |
|
|
|
createImportNamespaceSpecifier: function (id) { |
|
return { |
|
type: Syntax.ImportNamespaceSpecifier, |
|
id: id |
|
}; |
|
}, |
|
|
|
createExportDeclaration: function (isDefault, declaration, specifiers, src) { |
|
return { |
|
type: Syntax.ExportDeclaration, |
|
'default': !!isDefault, |
|
declaration: declaration, |
|
specifiers: specifiers, |
|
source: src |
|
}; |
|
}, |
|
|
|
createImportSpecifier: function (id, name) { |
|
return { |
|
type: Syntax.ImportSpecifier, |
|
id: id, |
|
name: name |
|
}; |
|
}, |
|
|
|
createImportDeclaration: function (specifiers, src, isType) { |
|
return { |
|
type: Syntax.ImportDeclaration, |
|
specifiers: specifiers, |
|
source: src, |
|
isType: isType |
|
}; |
|
}, |
|
|
|
createYieldExpression: function (argument, dlg) { |
|
return { |
|
type: Syntax.YieldExpression, |
|
argument: argument, |
|
delegate: dlg |
|
}; |
|
}, |
|
|
|
createAwaitExpression: function (argument) { |
|
return { |
|
type: Syntax.AwaitExpression, |
|
argument: argument |
|
}; |
|
}, |
|
|
|
createComprehensionExpression: function (filter, blocks, body) { |
|
return { |
|
type: Syntax.ComprehensionExpression, |
|
filter: filter, |
|
blocks: blocks, |
|
body: body |
|
}; |
|
} |
|
|
|
}; |
|
|
|
// Return true if there is a line terminator before the next token. |
|
|
|
function peekLineTerminator() { |
|
var pos, line, start, found; |
|
|
|
pos = index; |
|
line = lineNumber; |
|
start = lineStart; |
|
skipComment(); |
|
found = lineNumber !== line; |
|
index = pos; |
|
lineNumber = line; |
|
lineStart = start; |
|
|
|
return found; |
|
} |
|
|
|
// Throw an exception |
|
|
|
function throwError(token, messageFormat) { |
|
var error, |
|
args = Array.prototype.slice.call(arguments, 2), |
|
msg = messageFormat.replace( |
|
/%(\d)/g, |
|
function (whole, idx) { |
|
assert(idx < args.length, 'Message reference must be in range'); |
|
return args[idx]; |
|
} |
|
); |
|
|
|
if (typeof token.lineNumber === 'number') { |
|
error = new Error('Line ' + token.lineNumber + ': ' + msg); |
|
error.index = token.range[0]; |
|
error.lineNumber = token.lineNumber; |
|
error.column = token.range[0] - lineStart + 1; |
|
} else { |
|
error = new Error('Line ' + lineNumber + ': ' + msg); |
|
error.index = index; |
|
error.lineNumber = lineNumber; |
|
error.column = index - lineStart + 1; |
|
} |
|
|
|
error.description = msg; |
|
throw error; |
|
} |
|
|
|
function throwErrorTolerant() { |
|
try { |
|
throwError.apply(null, arguments); |
|
} catch (e) { |
|
if (extra.errors) { |
|
extra.errors.push(e); |
|
} else { |
|
throw e; |
|
} |
|
} |
|
} |
|
|
|
|
|
// Throw an exception because of the token. |
|
|
|
function throwUnexpected(token) { |
|
if (token.type === Token.EOF) { |
|
throwError(token, Messages.UnexpectedEOS); |
|
} |
|
|
|
if (token.type === Token.NumericLiteral) { |
|
throwError(token, Messages.UnexpectedNumber); |
|
} |
|
|
|
if (token.type === Token.StringLiteral || token.type === Token.JSXText) { |
|
throwError(token, Messages.UnexpectedString); |
|
} |
|
|
|
if (token.type === Token.Identifier) { |
|
throwError(token, Messages.UnexpectedIdentifier); |
|
} |
|
|
|
if (token.type === Token.Keyword) { |
|
if (isFutureReservedWord(token.value)) { |
|
throwError(token, Messages.UnexpectedReserved); |
|
} else if (strict && isStrictModeReservedWord(token.value)) { |
|
throwErrorTolerant(token, Messages.StrictReservedWord); |
|
return; |
|
} |
|
throwError(token, Messages.UnexpectedToken, token.value); |
|
} |
|
|
|
if (token.type === Token.Template) { |
|
throwError(token, Messages.UnexpectedTemplate, token.value.raw); |
|
} |
|
|
|
// BooleanLiteral, NullLiteral, or Punctuator. |
|
throwError(token, Messages.UnexpectedToken, token.value); |
|
} |
|
|
|
// Expect the next token to match the specified punctuator. |
|
// If not, an exception will be thrown. |
|
|
|
function expect(value) { |
|
var token = lex(); |
|
if (token.type !== Token.Punctuator || token.value !== value) { |
|
throwUnexpected(token); |
|
} |
|
} |
|
|
|
// Expect the next token to match the specified keyword. |
|
// If not, an exception will be thrown. |
|
|
|
function expectKeyword(keyword, contextual) { |
|
var token = lex(); |
|
if (token.type !== (contextual ? Token.Identifier : Token.Keyword) || |
|
token.value !== keyword) { |
|
throwUnexpected(token); |
|
} |
|
} |
|
|
|
// Expect the next token to match the specified contextual keyword. |
|
// If not, an exception will be thrown. |
|
|
|
function expectContextualKeyword(keyword) { |
|
return expectKeyword(keyword, true); |
|
} |
|
|
|
// Return true if the next token matches the specified punctuator. |
|
|
|
function match(value) { |
|
return lookahead.type === Token.Punctuator && lookahead.value === value; |
|
} |
|
|
|
// Return true if the next token matches the specified keyword |
|
|
|
function matchKeyword(keyword, contextual) { |
|
var expectedType = contextual ? Token.Identifier : Token.Keyword; |
|
return lookahead.type === expectedType && lookahead.value === keyword; |
|
} |
|
|
|
// Return true if the next token matches the specified contextual keyword |
|
|
|
function matchContextualKeyword(keyword) { |
|
return matchKeyword(keyword, true); |
|
} |
|
|
|
// Return true if the next token is an assignment operator |
|
|
|
function matchAssign() { |
|
var op; |
|
|
|
if (lookahead.type !== Token.Punctuator) { |
|
return false; |
|
} |
|
op = lookahead.value; |
|
return op === '=' || |
|
op === '*=' || |
|
op === '/=' || |
|
op === '%=' || |
|
op === '+=' || |
|
op === '-=' || |
|
op === '<<=' || |
|
op === '>>=' || |
|
op === '>>>=' || |
|
op === '&=' || |
|
op === '^=' || |
|
op === '|='; |
|
} |
|
|
|
// Note that 'yield' is treated as a keyword in strict mode, but a |
|
// contextual keyword (identifier) in non-strict mode, so we need to |
|
// use matchKeyword('yield', false) and matchKeyword('yield', true) |
|
// (i.e. matchContextualKeyword) appropriately. |
|
function matchYield() { |
|
return state.yieldAllowed && matchKeyword('yield', !strict); |
|
} |
|
|
|
function matchAsync() { |
|
var backtrackToken = lookahead, matches = false; |
|
|
|
if (matchContextualKeyword('async')) { |
|
lex(); // Make sure peekLineTerminator() starts after 'async'. |
|
matches = !peekLineTerminator(); |
|
rewind(backtrackToken); // Revert the lex(). |
|
} |
|
|
|
return matches; |
|
} |
|
|
|
function matchAwait() { |
|
return state.awaitAllowed && matchContextualKeyword('await'); |
|
} |
|
|
|
function consumeSemicolon() { |
|
var line, oldIndex = index, oldLineNumber = lineNumber, |
|
oldLineStart = lineStart, oldLookahead = lookahead; |
|
|
|
// Catch the very common case first: immediately a semicolon (char #59). |
|
if (source.charCodeAt(index) === 59) { |
|
lex(); |
|
return; |
|
} |
|
|
|
line = lineNumber; |
|
skipComment(); |
|
if (lineNumber !== line) { |
|
index = oldIndex; |
|
lineNumber = oldLineNumber; |
|
lineStart = oldLineStart; |
|
lookahead = oldLookahead; |
|
return; |
|
} |
|
|
|
if (match(';')) { |
|
lex(); |
|
return; |
|
} |
|
|
|
if (lookahead.type !== Token.EOF && !match('}')) { |
|
throwUnexpected(lookahead); |
|
} |
|
} |
|
|
|
// Return true if provided expression is LeftHandSideExpression |
|
|
|
function isLeftHandSide(expr) { |
|
return expr.type === Syntax.Identifier || expr.type === Syntax.MemberExpression; |
|
} |
|
|
|
function isAssignableLeftHandSide(expr) { |
|
return isLeftHandSide(expr) || expr.type === Syntax.ObjectPattern || expr.type === Syntax.ArrayPattern; |
|
} |
|
|
|
// 11.1.4 Array Initialiser |
|
|
|
function parseArrayInitialiser() { |
|
var elements = [], blocks = [], filter = null, tmp, possiblecomprehension = true, |
|
marker = markerCreate(); |
|
|
|
expect('['); |
|
while (!match(']')) { |
|
if (lookahead.value === 'for' && |
|
lookahead.type === Token.Keyword) { |
|
if (!possiblecomprehension) { |
|
throwError({}, Messages.ComprehensionError); |
|
} |
|
matchKeyword('for'); |
|
tmp = parseForStatement({ignoreBody: true}); |
|
tmp.of = tmp.type === Syntax.ForOfStatement; |
|
tmp.type = Syntax.ComprehensionBlock; |
|
if (tmp.left.kind) { // can't be let or const |
|
throwError({}, Messages.ComprehensionError); |
|
} |
|
blocks.push(tmp); |
|
} else if (lookahead.value === 'if' && |
|
lookahead.type === Token.Keyword) { |
|
if (!possiblecomprehension) { |
|
throwError({}, Messages.ComprehensionError); |
|
} |
|
expectKeyword('if'); |
|
expect('('); |
|
filter = parseExpression(); |
|
expect(')'); |
|
} else if (lookahead.value === ',' && |
|
lookahead.type === Token.Punctuator) { |
|
possiblecomprehension = false; // no longer allowed. |
|
lex(); |
|
elements.push(null); |
|
} else { |
|
tmp = parseSpreadOrAssignmentExpression(); |
|
elements.push(tmp); |
|
if (tmp && tmp.type === Syntax.SpreadElement) { |
|
if (!match(']')) { |
|
throwError({}, Messages.ElementAfterSpreadElement); |
|
} |
|
} else if (!(match(']') || matchKeyword('for') || matchKeyword('if'))) { |
|
expect(','); // this lexes. |
|
possiblecomprehension = false; |
|
} |
|
} |
|
} |
|
|
|
expect(']'); |
|
|
|
if (filter && !blocks.length) { |
|
throwError({}, Messages.ComprehensionRequiresBlock); |
|
} |
|
|
|
if (blocks.length) { |
|
if (elements.length !== 1) { |
|
throwError({}, Messages.ComprehensionError); |
|
} |
|
return markerApply(marker, delegate.createComprehensionExpression(filter, blocks, elements[0])); |
|
} |
|
return markerApply(marker, delegate.createArrayExpression(elements)); |
|
} |
|
|
|
// 11.1.5 Object Initialiser |
|
|
|
function parsePropertyFunction(options) { |
|
var previousStrict, previousYieldAllowed, previousAwaitAllowed, |
|
params, defaults, body, marker = markerCreate(); |
|
|
|
previousStrict = strict; |
|
previousYieldAllowed = state.yieldAllowed; |
|
state.yieldAllowed = options.generator; |
|
previousAwaitAllowed = state.awaitAllowed; |
|
state.awaitAllowed = options.async; |
|
params = options.params || []; |
|
defaults = options.defaults || []; |
|
|
|
body = parseConciseBody(); |
|
if (options.name && strict && isRestrictedWord(params[0].name)) { |
|
throwErrorTolerant(options.name, Messages.StrictParamName); |
|
} |
|
strict = previousStrict; |
|
state.yieldAllowed = previousYieldAllowed; |
|
state.awaitAllowed = previousAwaitAllowed; |
|
|
|
return markerApply(marker, delegate.createFunctionExpression( |
|
null, |
|
params, |
|
defaults, |
|
body, |
|
options.rest || null, |
|
options.generator, |
|
body.type !== Syntax.BlockStatement, |
|
options.async, |
|
options.returnType, |
|
options.typeParameters |
|
)); |
|
} |
|
|
|
|
|
function parsePropertyMethodFunction(options) { |
|
var previousStrict, tmp, method; |
|
|
|
previousStrict = strict; |
|
strict = true; |
|
|
|
tmp = parseParams(); |
|
|
|
if (tmp.stricted) { |
|
throwErrorTolerant(tmp.stricted, tmp.message); |
|
} |
|
|
|
method = parsePropertyFunction({ |
|
params: tmp.params, |
|
defaults: tmp.defaults, |
|
rest: tmp.rest, |
|
generator: options.generator, |
|
async: options.async, |
|
returnType: tmp.returnType, |
|
typeParameters: options.typeParameters |
|
}); |
|
|
|
strict = previousStrict; |
|
|
|
return method; |
|
} |
|
|
|
|
|
function parseObjectPropertyKey() { |
|
var marker = markerCreate(), |
|
token = lex(), |
|
propertyKey, |
|
result; |
|
|
|
// Note: This function is called only from parseObjectProperty(), where |
|
// EOF and Punctuator tokens are already filtered out. |
|
|
|
if (token.type === Token.StringLiteral || token.type === Token.NumericLiteral) { |
|
if (strict && token.octal) { |
|
throwErrorTolerant(token, Messages.StrictOctalLiteral); |
|
} |
|
return markerApply(marker, delegate.createLiteral(token)); |
|
} |
|
|
|
if (token.type === Token.Punctuator && token.value === '[') { |
|
// For computed properties we should skip the [ and ], and |
|
// capture in marker only the assignment expression itself. |
|
marker = markerCreate(); |
|
propertyKey = parseAssignmentExpression(); |
|
result = markerApply(marker, propertyKey); |
|
expect(']'); |
|
return result; |
|
} |
|
|
|
return markerApply(marker, delegate.createIdentifier(token.value)); |
|
} |
|
|
|
function parseObjectProperty() { |
|
var token, key, id, param, computed, |
|
marker = markerCreate(), returnType, typeParameters; |
|
|
|
token = lookahead; |
|
computed = (token.value === '[' && token.type === Token.Punctuator); |
|
|
|
if (token.type === Token.Identifier || computed || matchAsync()) { |
|
id = parseObjectPropertyKey(); |
|
|
|
if (match(':')) { |
|
lex(); |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createProperty( |
|
'init', |
|
id, |
|
parseAssignmentExpression(), |
|
false, |
|
false, |
|
computed |
|
) |
|
); |
|
} |
|
|
|
if (match('(') || match('<')) { |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
return markerApply( |
|
marker, |
|
delegate.createProperty( |
|
'init', |
|
id, |
|
parsePropertyMethodFunction({ |
|
generator: false, |
|
async: false, |
|
typeParameters: typeParameters |
|
}), |
|
true, |
|
false, |
|
computed |
|
) |
|
); |
|
} |
|
|
|
// Property Assignment: Getter and Setter. |
|
|
|
if (token.value === 'get') { |
|
computed = (lookahead.value === '['); |
|
key = parseObjectPropertyKey(); |
|
|
|
expect('('); |
|
expect(')'); |
|
if (match(':')) { |
|
returnType = parseTypeAnnotation(); |
|
} |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createProperty( |
|
'get', |
|
key, |
|
parsePropertyFunction({ |
|
generator: false, |
|
async: false, |
|
returnType: returnType |
|
}), |
|
false, |
|
false, |
|
computed |
|
) |
|
); |
|
} |
|
|
|
if (token.value === 'set') { |
|
computed = (lookahead.value === '['); |
|
key = parseObjectPropertyKey(); |
|
|
|
expect('('); |
|
token = lookahead; |
|
param = [ parseTypeAnnotatableIdentifier() ]; |
|
expect(')'); |
|
if (match(':')) { |
|
returnType = parseTypeAnnotation(); |
|
} |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createProperty( |
|
'set', |
|
key, |
|
parsePropertyFunction({ |
|
params: param, |
|
generator: false, |
|
async: false, |
|
name: token, |
|
returnType: returnType |
|
}), |
|
false, |
|
false, |
|
computed |
|
) |
|
); |
|
} |
|
|
|
if (token.value === 'async') { |
|
computed = (lookahead.value === '['); |
|
key = parseObjectPropertyKey(); |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createProperty( |
|
'init', |
|
key, |
|
parsePropertyMethodFunction({ |
|
generator: false, |
|
async: true, |
|
typeParameters: typeParameters |
|
}), |
|
true, |
|
false, |
|
computed |
|
) |
|
); |
|
} |
|
|
|
if (computed) { |
|
// Computed properties can only be used with full notation. |
|
throwUnexpected(lookahead); |
|
} |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createProperty('init', id, id, false, true, false) |
|
); |
|
} |
|
|
|
if (token.type === Token.EOF || token.type === Token.Punctuator) { |
|
if (!match('*')) { |
|
throwUnexpected(token); |
|
} |
|
lex(); |
|
|
|
computed = (lookahead.type === Token.Punctuator && lookahead.value === '['); |
|
|
|
id = parseObjectPropertyKey(); |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
if (!match('(')) { |
|
throwUnexpected(lex()); |
|
} |
|
|
|
return markerApply(marker, delegate.createProperty( |
|
'init', |
|
id, |
|
parsePropertyMethodFunction({ |
|
generator: true, |
|
typeParameters: typeParameters |
|
}), |
|
true, |
|
false, |
|
computed |
|
)); |
|
} |
|
key = parseObjectPropertyKey(); |
|
if (match(':')) { |
|
lex(); |
|
return markerApply(marker, delegate.createProperty('init', key, parseAssignmentExpression(), false, false, false)); |
|
} |
|
if (match('(') || match('<')) { |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
return markerApply(marker, delegate.createProperty( |
|
'init', |
|
key, |
|
parsePropertyMethodFunction({ |
|
generator: false, |
|
typeParameters: typeParameters |
|
}), |
|
true, |
|
false, |
|
false |
|
)); |
|
} |
|
throwUnexpected(lex()); |
|
} |
|
|
|
function parseObjectSpreadProperty() { |
|
var marker = markerCreate(); |
|
expect('...'); |
|
return markerApply(marker, delegate.createSpreadProperty(parseAssignmentExpression())); |
|
} |
|
|
|
function getFieldName(key) { |
|
var toString = String; |
|
if (key.type === Syntax.Identifier) { |
|
return key.name; |
|
} |
|
return toString(key.value); |
|
} |
|
|
|
function parseObjectInitialiser() { |
|
var properties = [], property, name, kind, storedKind, map = new StringMap(), |
|
marker = markerCreate(), toString = String; |
|
|
|
expect('{'); |
|
|
|
while (!match('}')) { |
|
if (match('...')) { |
|
property = parseObjectSpreadProperty(); |
|
} else { |
|
property = parseObjectProperty(); |
|
|
|
if (property.key.type === Syntax.Identifier) { |
|
name = property.key.name; |
|
} else { |
|
name = toString(property.key.value); |
|
} |
|
kind = (property.kind === 'init') ? PropertyKind.Data : (property.kind === 'get') ? PropertyKind.Get : PropertyKind.Set; |
|
|
|
if (map.has(name)) { |
|
storedKind = map.get(name); |
|
if (storedKind === PropertyKind.Data) { |
|
if (strict && kind === PropertyKind.Data) { |
|
throwErrorTolerant({}, Messages.StrictDuplicateProperty); |
|
} else if (kind !== PropertyKind.Data) { |
|
throwErrorTolerant({}, Messages.AccessorDataProperty); |
|
} |
|
} else { |
|
if (kind === PropertyKind.Data) { |
|
throwErrorTolerant({}, Messages.AccessorDataProperty); |
|
} else if (storedKind & kind) { |
|
throwErrorTolerant({}, Messages.AccessorGetSet); |
|
} |
|
} |
|
map.set(name, storedKind | kind); |
|
} else { |
|
map.set(name, kind); |
|
} |
|
} |
|
|
|
properties.push(property); |
|
|
|
if (!match('}')) { |
|
expect(','); |
|
} |
|
} |
|
|
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createObjectExpression(properties)); |
|
} |
|
|
|
function parseTemplateElement(option) { |
|
var marker = markerCreate(), |
|
token = scanTemplateElement(option); |
|
if (strict && token.octal) { |
|
throwError(token, Messages.StrictOctalLiteral); |
|
} |
|
return markerApply(marker, delegate.createTemplateElement({ raw: token.value.raw, cooked: token.value.cooked }, token.tail)); |
|
} |
|
|
|
function parseTemplateLiteral() { |
|
var quasi, quasis, expressions, marker = markerCreate(); |
|
|
|
quasi = parseTemplateElement({ head: true }); |
|
quasis = [ quasi ]; |
|
expressions = []; |
|
|
|
while (!quasi.tail) { |
|
expressions.push(parseExpression()); |
|
quasi = parseTemplateElement({ head: false }); |
|
quasis.push(quasi); |
|
} |
|
|
|
return markerApply(marker, delegate.createTemplateLiteral(quasis, expressions)); |
|
} |
|
|
|
// 11.1.6 The Grouping Operator |
|
|
|
function parseGroupExpression() { |
|
var expr, marker, typeAnnotation; |
|
|
|
expect('('); |
|
|
|
++state.parenthesizedCount; |
|
|
|
marker = markerCreate(); |
|
|
|
expr = parseExpression(); |
|
|
|
if (match(':')) { |
|
typeAnnotation = parseTypeAnnotation(); |
|
expr = markerApply(marker, delegate.createTypeCast( |
|
expr, |
|
typeAnnotation |
|
)); |
|
} |
|
|
|
expect(')'); |
|
|
|
return expr; |
|
} |
|
|
|
function matchAsyncFuncExprOrDecl() { |
|
var token; |
|
|
|
if (matchAsync()) { |
|
token = lookahead2(); |
|
if (token.type === Token.Keyword && token.value === 'function') { |
|
return true; |
|
} |
|
} |
|
|
|
return false; |
|
} |
|
|
|
// 11.1 Primary Expressions |
|
|
|
function parsePrimaryExpression() { |
|
var marker, type, token, expr; |
|
|
|
type = lookahead.type; |
|
|
|
if (type === Token.Identifier) { |
|
marker = markerCreate(); |
|
return markerApply(marker, delegate.createIdentifier(lex().value)); |
|
} |
|
|
|
if (type === Token.StringLiteral || type === Token.NumericLiteral) { |
|
if (strict && lookahead.octal) { |
|
throwErrorTolerant(lookahead, Messages.StrictOctalLiteral); |
|
} |
|
marker = markerCreate(); |
|
return markerApply(marker, delegate.createLiteral(lex())); |
|
} |
|
|
|
if (type === Token.Keyword) { |
|
if (matchKeyword('this')) { |
|
marker = markerCreate(); |
|
lex(); |
|
return markerApply(marker, delegate.createThisExpression()); |
|
} |
|
|
|
if (matchKeyword('function')) { |
|
return parseFunctionExpression(); |
|
} |
|
|
|
if (matchKeyword('class')) { |
|
return parseClassExpression(); |
|
} |
|
|
|
if (matchKeyword('super')) { |
|
marker = markerCreate(); |
|
lex(); |
|
return markerApply(marker, delegate.createIdentifier('super')); |
|
} |
|
} |
|
|
|
if (type === Token.BooleanLiteral) { |
|
marker = markerCreate(); |
|
token = lex(); |
|
token.value = (token.value === 'true'); |
|
return markerApply(marker, delegate.createLiteral(token)); |
|
} |
|
|
|
if (type === Token.NullLiteral) { |
|
marker = markerCreate(); |
|
token = lex(); |
|
token.value = null; |
|
return markerApply(marker, delegate.createLiteral(token)); |
|
} |
|
|
|
if (match('[')) { |
|
return parseArrayInitialiser(); |
|
} |
|
|
|
if (match('{')) { |
|
return parseObjectInitialiser(); |
|
} |
|
|
|
if (match('(')) { |
|
return parseGroupExpression(); |
|
} |
|
|
|
if (match('/') || match('/=')) { |
|
marker = markerCreate(); |
|
expr = delegate.createLiteral(scanRegExp()); |
|
peek(); |
|
return markerApply(marker, expr); |
|
} |
|
|
|
if (type === Token.Template) { |
|
return parseTemplateLiteral(); |
|
} |
|
|
|
if (match('<')) { |
|
return parseJSXElement(); |
|
} |
|
|
|
throwUnexpected(lex()); |
|
} |
|
|
|
// 11.2 Left-Hand-Side Expressions |
|
|
|
function parseArguments() { |
|
var args = [], arg; |
|
|
|
expect('('); |
|
|
|
if (!match(')')) { |
|
while (index < length) { |
|
arg = parseSpreadOrAssignmentExpression(); |
|
args.push(arg); |
|
|
|
if (match(')')) { |
|
break; |
|
} else if (arg.type === Syntax.SpreadElement) { |
|
throwError({}, Messages.ElementAfterSpreadElement); |
|
} |
|
|
|
expect(','); |
|
} |
|
} |
|
|
|
expect(')'); |
|
|
|
return args; |
|
} |
|
|
|
function parseSpreadOrAssignmentExpression() { |
|
if (match('...')) { |
|
var marker = markerCreate(); |
|
lex(); |
|
return markerApply(marker, delegate.createSpreadElement(parseAssignmentExpression())); |
|
} |
|
return parseAssignmentExpression(); |
|
} |
|
|
|
function parseNonComputedProperty() { |
|
var marker = markerCreate(), |
|
token = lex(); |
|
|
|
if (!isIdentifierName(token)) { |
|
throwUnexpected(token); |
|
} |
|
|
|
return markerApply(marker, delegate.createIdentifier(token.value)); |
|
} |
|
|
|
function parseNonComputedMember() { |
|
expect('.'); |
|
|
|
return parseNonComputedProperty(); |
|
} |
|
|
|
function parseComputedMember() { |
|
var expr; |
|
|
|
expect('['); |
|
|
|
expr = parseExpression(); |
|
|
|
expect(']'); |
|
|
|
return expr; |
|
} |
|
|
|
function parseNewExpression() { |
|
var callee, args, marker = markerCreate(); |
|
|
|
expectKeyword('new'); |
|
callee = parseLeftHandSideExpression(); |
|
args = match('(') ? parseArguments() : []; |
|
|
|
return markerApply(marker, delegate.createNewExpression(callee, args)); |
|
} |
|
|
|
function parseLeftHandSideExpressionAllowCall() { |
|
var expr, args, marker = markerCreate(); |
|
|
|
expr = matchKeyword('new') ? parseNewExpression() : parsePrimaryExpression(); |
|
|
|
while (match('.') || match('[') || match('(') || lookahead.type === Token.Template) { |
|
if (match('(')) { |
|
args = parseArguments(); |
|
expr = markerApply(marker, delegate.createCallExpression(expr, args)); |
|
} else if (match('[')) { |
|
expr = markerApply(marker, delegate.createMemberExpression('[', expr, parseComputedMember())); |
|
} else if (match('.')) { |
|
expr = markerApply(marker, delegate.createMemberExpression('.', expr, parseNonComputedMember())); |
|
} else { |
|
expr = markerApply(marker, delegate.createTaggedTemplateExpression(expr, parseTemplateLiteral())); |
|
} |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
function parseLeftHandSideExpression() { |
|
var expr, marker = markerCreate(); |
|
|
|
expr = matchKeyword('new') ? parseNewExpression() : parsePrimaryExpression(); |
|
|
|
while (match('.') || match('[') || lookahead.type === Token.Template) { |
|
if (match('[')) { |
|
expr = markerApply(marker, delegate.createMemberExpression('[', expr, parseComputedMember())); |
|
} else if (match('.')) { |
|
expr = markerApply(marker, delegate.createMemberExpression('.', expr, parseNonComputedMember())); |
|
} else { |
|
expr = markerApply(marker, delegate.createTaggedTemplateExpression(expr, parseTemplateLiteral())); |
|
} |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
// 11.3 Postfix Expressions |
|
|
|
function parsePostfixExpression() { |
|
var marker = markerCreate(), |
|
expr = parseLeftHandSideExpressionAllowCall(), |
|
token; |
|
|
|
if (lookahead.type !== Token.Punctuator) { |
|
return expr; |
|
} |
|
|
|
if ((match('++') || match('--')) && !peekLineTerminator()) { |
|
// 11.3.1, 11.3.2 |
|
if (strict && expr.type === Syntax.Identifier && isRestrictedWord(expr.name)) { |
|
throwErrorTolerant({}, Messages.StrictLHSPostfix); |
|
} |
|
|
|
if (!isLeftHandSide(expr)) { |
|
throwError({}, Messages.InvalidLHSInAssignment); |
|
} |
|
|
|
token = lex(); |
|
expr = markerApply(marker, delegate.createPostfixExpression(token.value, expr)); |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
// 11.4 Unary Operators |
|
|
|
function parseUnaryExpression() { |
|
var marker, token, expr; |
|
|
|
if (lookahead.type !== Token.Punctuator && lookahead.type !== Token.Keyword) { |
|
return parsePostfixExpression(); |
|
} |
|
|
|
if (match('++') || match('--')) { |
|
marker = markerCreate(); |
|
token = lex(); |
|
expr = parseUnaryExpression(); |
|
// 11.4.4, 11.4.5 |
|
if (strict && expr.type === Syntax.Identifier && isRestrictedWord(expr.name)) { |
|
throwErrorTolerant({}, Messages.StrictLHSPrefix); |
|
} |
|
|
|
if (!isLeftHandSide(expr)) { |
|
throwError({}, Messages.InvalidLHSInAssignment); |
|
} |
|
|
|
return markerApply(marker, delegate.createUnaryExpression(token.value, expr)); |
|
} |
|
|
|
if (match('+') || match('-') || match('~') || match('!')) { |
|
marker = markerCreate(); |
|
token = lex(); |
|
expr = parseUnaryExpression(); |
|
return markerApply(marker, delegate.createUnaryExpression(token.value, expr)); |
|
} |
|
|
|
if (matchKeyword('delete') || matchKeyword('void') || matchKeyword('typeof')) { |
|
marker = markerCreate(); |
|
token = lex(); |
|
expr = parseUnaryExpression(); |
|
expr = markerApply(marker, delegate.createUnaryExpression(token.value, expr)); |
|
if (strict && expr.operator === 'delete' && expr.argument.type === Syntax.Identifier) { |
|
throwErrorTolerant({}, Messages.StrictDelete); |
|
} |
|
return expr; |
|
} |
|
|
|
return parsePostfixExpression(); |
|
} |
|
|
|
function binaryPrecedence(token, allowIn) { |
|
var prec = 0; |
|
|
|
if (token.type !== Token.Punctuator && token.type !== Token.Keyword) { |
|
return 0; |
|
} |
|
|
|
switch (token.value) { |
|
case '||': |
|
prec = 1; |
|
break; |
|
|
|
case '&&': |
|
prec = 2; |
|
break; |
|
|
|
case '|': |
|
prec = 3; |
|
break; |
|
|
|
case '^': |
|
prec = 4; |
|
break; |
|
|
|
case '&': |
|
prec = 5; |
|
break; |
|
|
|
case '==': |
|
case '!=': |
|
case '===': |
|
case '!==': |
|
prec = 6; |
|
break; |
|
|
|
case '<': |
|
case '>': |
|
case '<=': |
|
case '>=': |
|
case 'instanceof': |
|
prec = 7; |
|
break; |
|
|
|
case 'in': |
|
prec = allowIn ? 7 : 0; |
|
break; |
|
|
|
case '<<': |
|
case '>>': |
|
case '>>>': |
|
prec = 8; |
|
break; |
|
|
|
case '+': |
|
case '-': |
|
prec = 9; |
|
break; |
|
|
|
case '*': |
|
case '/': |
|
case '%': |
|
prec = 11; |
|
break; |
|
|
|
default: |
|
break; |
|
} |
|
|
|
return prec; |
|
} |
|
|
|
// 11.5 Multiplicative Operators |
|
// 11.6 Additive Operators |
|
// 11.7 Bitwise Shift Operators |
|
// 11.8 Relational Operators |
|
// 11.9 Equality Operators |
|
// 11.10 Binary Bitwise Operators |
|
// 11.11 Binary Logical Operators |
|
|
|
function parseBinaryExpression() { |
|
var expr, token, prec, previousAllowIn, stack, right, operator, left, i, |
|
marker, markers; |
|
|
|
previousAllowIn = state.allowIn; |
|
state.allowIn = true; |
|
|
|
marker = markerCreate(); |
|
left = parseUnaryExpression(); |
|
|
|
token = lookahead; |
|
prec = binaryPrecedence(token, previousAllowIn); |
|
if (prec === 0) { |
|
return left; |
|
} |
|
token.prec = prec; |
|
lex(); |
|
|
|
markers = [marker, markerCreate()]; |
|
right = parseUnaryExpression(); |
|
|
|
stack = [left, token, right]; |
|
|
|
while ((prec = binaryPrecedence(lookahead, previousAllowIn)) > 0) { |
|
|
|
// Reduce: make a binary expression from the three topmost entries. |
|
while ((stack.length > 2) && (prec <= stack[stack.length - 2].prec)) { |
|
right = stack.pop(); |
|
operator = stack.pop().value; |
|
left = stack.pop(); |
|
expr = delegate.createBinaryExpression(operator, left, right); |
|
markers.pop(); |
|
marker = markers.pop(); |
|
markerApply(marker, expr); |
|
stack.push(expr); |
|
markers.push(marker); |
|
} |
|
|
|
// Shift. |
|
token = lex(); |
|
token.prec = prec; |
|
stack.push(token); |
|
markers.push(markerCreate()); |
|
expr = parseUnaryExpression(); |
|
stack.push(expr); |
|
} |
|
|
|
state.allowIn = previousAllowIn; |
|
|
|
// Final reduce to clean-up the stack. |
|
i = stack.length - 1; |
|
expr = stack[i]; |
|
markers.pop(); |
|
while (i > 1) { |
|
expr = delegate.createBinaryExpression(stack[i - 1].value, stack[i - 2], expr); |
|
i -= 2; |
|
marker = markers.pop(); |
|
markerApply(marker, expr); |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
|
|
// 11.12 Conditional Operator |
|
|
|
function parseConditionalExpression() { |
|
var expr, previousAllowIn, consequent, alternate, marker = markerCreate(); |
|
expr = parseBinaryExpression(); |
|
|
|
if (match('?')) { |
|
lex(); |
|
previousAllowIn = state.allowIn; |
|
state.allowIn = true; |
|
consequent = parseAssignmentExpression(); |
|
state.allowIn = previousAllowIn; |
|
expect(':'); |
|
alternate = parseAssignmentExpression(); |
|
|
|
expr = markerApply(marker, delegate.createConditionalExpression(expr, consequent, alternate)); |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
// 11.13 Assignment Operators |
|
|
|
// 12.14.5 AssignmentPattern |
|
|
|
function reinterpretAsAssignmentBindingPattern(expr) { |
|
var i, len, property, element; |
|
|
|
if (expr.type === Syntax.ObjectExpression) { |
|
expr.type = Syntax.ObjectPattern; |
|
for (i = 0, len = expr.properties.length; i < len; i += 1) { |
|
property = expr.properties[i]; |
|
if (property.type === Syntax.SpreadProperty) { |
|
if (i < len - 1) { |
|
throwError({}, Messages.PropertyAfterSpreadProperty); |
|
} |
|
reinterpretAsAssignmentBindingPattern(property.argument); |
|
} else { |
|
if (property.kind !== 'init') { |
|
throwError({}, Messages.InvalidLHSInAssignment); |
|
} |
|
reinterpretAsAssignmentBindingPattern(property.value); |
|
} |
|
} |
|
} else if (expr.type === Syntax.ArrayExpression) { |
|
expr.type = Syntax.ArrayPattern; |
|
for (i = 0, len = expr.elements.length; i < len; i += 1) { |
|
element = expr.elements[i]; |
|
/* istanbul ignore else */ |
|
if (element) { |
|
reinterpretAsAssignmentBindingPattern(element); |
|
} |
|
} |
|
} else if (expr.type === Syntax.Identifier) { |
|
if (isRestrictedWord(expr.name)) { |
|
throwError({}, Messages.InvalidLHSInAssignment); |
|
} |
|
} else if (expr.type === Syntax.SpreadElement) { |
|
reinterpretAsAssignmentBindingPattern(expr.argument); |
|
if (expr.argument.type === Syntax.ObjectPattern) { |
|
throwError({}, Messages.ObjectPatternAsSpread); |
|
} |
|
} else { |
|
/* istanbul ignore else */ |
|
if (expr.type !== Syntax.MemberExpression && expr.type !== Syntax.CallExpression && expr.type !== Syntax.NewExpression) { |
|
throwError({}, Messages.InvalidLHSInAssignment); |
|
} |
|
} |
|
} |
|
|
|
// 13.2.3 BindingPattern |
|
|
|
function reinterpretAsDestructuredParameter(options, expr) { |
|
var i, len, property, element; |
|
|
|
if (expr.type === Syntax.ObjectExpression) { |
|
expr.type = Syntax.ObjectPattern; |
|
for (i = 0, len = expr.properties.length; i < len; i += 1) { |
|
property = expr.properties[i]; |
|
if (property.type === Syntax.SpreadProperty) { |
|
if (i < len - 1) { |
|
throwError({}, Messages.PropertyAfterSpreadProperty); |
|
} |
|
reinterpretAsDestructuredParameter(options, property.argument); |
|
} else { |
|
if (property.kind !== 'init') { |
|
throwError({}, Messages.InvalidLHSInFormalsList); |
|
} |
|
reinterpretAsDestructuredParameter(options, property.value); |
|
} |
|
} |
|
} else if (expr.type === Syntax.ArrayExpression) { |
|
expr.type = Syntax.ArrayPattern; |
|
for (i = 0, len = expr.elements.length; i < len; i += 1) { |
|
element = expr.elements[i]; |
|
if (element) { |
|
reinterpretAsDestructuredParameter(options, element); |
|
} |
|
} |
|
} else if (expr.type === Syntax.Identifier) { |
|
validateParam(options, expr, expr.name); |
|
} else if (expr.type === Syntax.SpreadElement) { |
|
// BindingRestElement only allows BindingIdentifier |
|
if (expr.argument.type !== Syntax.Identifier) { |
|
throwError({}, Messages.InvalidLHSInFormalsList); |
|
} |
|
validateParam(options, expr.argument, expr.argument.name); |
|
} else { |
|
throwError({}, Messages.InvalidLHSInFormalsList); |
|
} |
|
} |
|
|
|
function reinterpretAsCoverFormalsList(expressions) { |
|
var i, len, param, params, defaults, defaultCount, options, rest; |
|
|
|
params = []; |
|
defaults = []; |
|
defaultCount = 0; |
|
rest = null; |
|
options = { |
|
paramSet: new StringMap() |
|
}; |
|
|
|
for (i = 0, len = expressions.length; i < len; i += 1) { |
|
param = expressions[i]; |
|
if (param.type === Syntax.Identifier) { |
|
params.push(param); |
|
defaults.push(null); |
|
validateParam(options, param, param.name); |
|
} else if (param.type === Syntax.ObjectExpression || param.type === Syntax.ArrayExpression) { |
|
reinterpretAsDestructuredParameter(options, param); |
|
params.push(param); |
|
defaults.push(null); |
|
} else if (param.type === Syntax.SpreadElement) { |
|
assert(i === len - 1, 'It is guaranteed that SpreadElement is last element by parseExpression'); |
|
if (param.argument.type !== Syntax.Identifier) { |
|
throwError({}, Messages.InvalidLHSInFormalsList); |
|
} |
|
reinterpretAsDestructuredParameter(options, param.argument); |
|
rest = param.argument; |
|
} else if (param.type === Syntax.AssignmentExpression) { |
|
params.push(param.left); |
|
defaults.push(param.right); |
|
++defaultCount; |
|
validateParam(options, param.left, param.left.name); |
|
} else { |
|
return null; |
|
} |
|
} |
|
|
|
if (options.message === Messages.StrictParamDupe) { |
|
throwError( |
|
strict ? options.stricted : options.firstRestricted, |
|
options.message |
|
); |
|
} |
|
|
|
if (defaultCount === 0) { |
|
defaults = []; |
|
} |
|
|
|
return { |
|
params: params, |
|
defaults: defaults, |
|
rest: rest, |
|
stricted: options.stricted, |
|
firstRestricted: options.firstRestricted, |
|
message: options.message |
|
}; |
|
} |
|
|
|
function parseArrowFunctionExpression(options, marker) { |
|
var previousStrict, previousYieldAllowed, previousAwaitAllowed, body; |
|
|
|
expect('=>'); |
|
|
|
previousStrict = strict; |
|
previousYieldAllowed = state.yieldAllowed; |
|
state.yieldAllowed = false; |
|
previousAwaitAllowed = state.awaitAllowed; |
|
state.awaitAllowed = !!options.async; |
|
body = parseConciseBody(); |
|
|
|
if (strict && options.firstRestricted) { |
|
throwError(options.firstRestricted, options.message); |
|
} |
|
if (strict && options.stricted) { |
|
throwErrorTolerant(options.stricted, options.message); |
|
} |
|
|
|
strict = previousStrict; |
|
state.yieldAllowed = previousYieldAllowed; |
|
state.awaitAllowed = previousAwaitAllowed; |
|
|
|
return markerApply(marker, delegate.createArrowFunctionExpression( |
|
options.params, |
|
options.defaults, |
|
body, |
|
options.rest, |
|
body.type !== Syntax.BlockStatement, |
|
!!options.async |
|
)); |
|
} |
|
|
|
function parseAssignmentExpression() { |
|
var marker, expr, token, params, oldParenthesizedCount, |
|
startsWithParen = false, backtrackToken = lookahead, |
|
possiblyAsync = false; |
|
|
|
if (matchYield()) { |
|
return parseYieldExpression(); |
|
} |
|
|
|
if (matchAwait()) { |
|
return parseAwaitExpression(); |
|
} |
|
|
|
oldParenthesizedCount = state.parenthesizedCount; |
|
|
|
marker = markerCreate(); |
|
|
|
if (matchAsyncFuncExprOrDecl()) { |
|
return parseFunctionExpression(); |
|
} |
|
|
|
if (matchAsync()) { |
|
// We can't be completely sure that this 'async' token is |
|
// actually a contextual keyword modifying a function |
|
// expression, so we might have to un-lex() it later by |
|
// calling rewind(backtrackToken). |
|
possiblyAsync = true; |
|
lex(); |
|
} |
|
|
|
if (match('(')) { |
|
token = lookahead2(); |
|
if ((token.type === Token.Punctuator && token.value === ')') || token.value === '...') { |
|
params = parseParams(); |
|
if (!match('=>')) { |
|
throwUnexpected(lex()); |
|
} |
|
params.async = possiblyAsync; |
|
return parseArrowFunctionExpression(params, marker); |
|
} |
|
startsWithParen = true; |
|
} |
|
|
|
token = lookahead; |
|
|
|
// If the 'async' keyword is not followed by a '(' character or an |
|
// identifier, then it can't be an arrow function modifier, and we |
|
// should interpret it as a normal identifer. |
|
if (possiblyAsync && !match('(') && token.type !== Token.Identifier) { |
|
possiblyAsync = false; |
|
rewind(backtrackToken); |
|
} |
|
|
|
expr = parseConditionalExpression(); |
|
|
|
if (match('=>') && |
|
(state.parenthesizedCount === oldParenthesizedCount || |
|
state.parenthesizedCount === (oldParenthesizedCount + 1))) { |
|
if (expr.type === Syntax.Identifier) { |
|
params = reinterpretAsCoverFormalsList([ expr ]); |
|
} else if (expr.type === Syntax.AssignmentExpression || |
|
expr.type === Syntax.ArrayExpression || |
|
expr.type === Syntax.ObjectExpression) { |
|
if (!startsWithParen) { |
|
throwUnexpected(lex()); |
|
} |
|
params = reinterpretAsCoverFormalsList([ expr ]); |
|
} else if (expr.type === Syntax.SequenceExpression) { |
|
params = reinterpretAsCoverFormalsList(expr.expressions); |
|
} |
|
if (params) { |
|
params.async = possiblyAsync; |
|
return parseArrowFunctionExpression(params, marker); |
|
} |
|
} |
|
|
|
// If we haven't returned by now, then the 'async' keyword was not |
|
// a function modifier, and we should rewind and interpret it as a |
|
// normal identifier. |
|
if (possiblyAsync) { |
|
possiblyAsync = false; |
|
rewind(backtrackToken); |
|
expr = parseConditionalExpression(); |
|
} |
|
|
|
if (matchAssign()) { |
|
// 11.13.1 |
|
if (strict && expr.type === Syntax.Identifier && isRestrictedWord(expr.name)) { |
|
throwErrorTolerant(token, Messages.StrictLHSAssignment); |
|
} |
|
|
|
// ES.next draf 11.13 Runtime Semantics step 1 |
|
if (match('=') && (expr.type === Syntax.ObjectExpression || expr.type === Syntax.ArrayExpression)) { |
|
reinterpretAsAssignmentBindingPattern(expr); |
|
} else if (!isLeftHandSide(expr)) { |
|
throwError({}, Messages.InvalidLHSInAssignment); |
|
} |
|
|
|
expr = markerApply(marker, delegate.createAssignmentExpression(lex().value, expr, parseAssignmentExpression())); |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
// 11.14 Comma Operator |
|
|
|
function parseExpression() { |
|
var marker, expr, expressions, sequence, spreadFound; |
|
|
|
marker = markerCreate(); |
|
expr = parseAssignmentExpression(); |
|
expressions = [ expr ]; |
|
|
|
if (match(',')) { |
|
while (index < length) { |
|
if (!match(',')) { |
|
break; |
|
} |
|
|
|
lex(); |
|
expr = parseSpreadOrAssignmentExpression(); |
|
expressions.push(expr); |
|
|
|
if (expr.type === Syntax.SpreadElement) { |
|
spreadFound = true; |
|
if (!match(')')) { |
|
throwError({}, Messages.ElementAfterSpreadElement); |
|
} |
|
break; |
|
} |
|
} |
|
|
|
sequence = markerApply(marker, delegate.createSequenceExpression(expressions)); |
|
} |
|
|
|
if (spreadFound && lookahead2().value !== '=>') { |
|
throwError({}, Messages.IllegalSpread); |
|
} |
|
|
|
return sequence || expr; |
|
} |
|
|
|
// 12.1 Block |
|
|
|
function parseStatementList() { |
|
var list = [], |
|
statement; |
|
|
|
while (index < length) { |
|
if (match('}')) { |
|
break; |
|
} |
|
statement = parseSourceElement(); |
|
if (typeof statement === 'undefined') { |
|
break; |
|
} |
|
list.push(statement); |
|
} |
|
|
|
return list; |
|
} |
|
|
|
function parseBlock() { |
|
var block, marker = markerCreate(); |
|
|
|
expect('{'); |
|
|
|
block = parseStatementList(); |
|
|
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createBlockStatement(block)); |
|
} |
|
|
|
// 12.2 Variable Statement |
|
|
|
function parseTypeParameterDeclaration() { |
|
var marker = markerCreate(), paramTypes = []; |
|
|
|
expect('<'); |
|
while (!match('>')) { |
|
paramTypes.push(parseTypeAnnotatableIdentifier()); |
|
if (!match('>')) { |
|
expect(','); |
|
} |
|
} |
|
expect('>'); |
|
|
|
return markerApply(marker, delegate.createTypeParameterDeclaration( |
|
paramTypes |
|
)); |
|
} |
|
|
|
function parseTypeParameterInstantiation() { |
|
var marker = markerCreate(), oldInType = state.inType, paramTypes = []; |
|
|
|
state.inType = true; |
|
|
|
expect('<'); |
|
while (!match('>')) { |
|
paramTypes.push(parseType()); |
|
if (!match('>')) { |
|
expect(','); |
|
} |
|
} |
|
expect('>'); |
|
|
|
state.inType = oldInType; |
|
|
|
return markerApply(marker, delegate.createTypeParameterInstantiation( |
|
paramTypes |
|
)); |
|
} |
|
|
|
function parseObjectTypeIndexer(marker, isStatic) { |
|
var id, key, value; |
|
|
|
expect('['); |
|
id = parseObjectPropertyKey(); |
|
expect(':'); |
|
key = parseType(); |
|
expect(']'); |
|
expect(':'); |
|
value = parseType(); |
|
|
|
return markerApply(marker, delegate.createObjectTypeIndexer( |
|
id, |
|
key, |
|
value, |
|
isStatic |
|
)); |
|
} |
|
|
|
function parseObjectTypeMethodish(marker) { |
|
var params = [], rest = null, returnType, typeParameters = null; |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
expect('('); |
|
while (lookahead.type === Token.Identifier) { |
|
params.push(parseFunctionTypeParam()); |
|
if (!match(')')) { |
|
expect(','); |
|
} |
|
} |
|
|
|
if (match('...')) { |
|
lex(); |
|
rest = parseFunctionTypeParam(); |
|
} |
|
expect(')'); |
|
expect(':'); |
|
returnType = parseType(); |
|
|
|
return markerApply(marker, delegate.createFunctionTypeAnnotation( |
|
params, |
|
returnType, |
|
rest, |
|
typeParameters |
|
)); |
|
} |
|
|
|
function parseObjectTypeMethod(marker, isStatic, key) { |
|
var optional = false, value; |
|
value = parseObjectTypeMethodish(marker); |
|
|
|
return markerApply(marker, delegate.createObjectTypeProperty( |
|
key, |
|
value, |
|
optional, |
|
isStatic |
|
)); |
|
} |
|
|
|
function parseObjectTypeCallProperty(marker, isStatic) { |
|
var valueMarker = markerCreate(); |
|
return markerApply(marker, delegate.createObjectTypeCallProperty( |
|
parseObjectTypeMethodish(valueMarker), |
|
isStatic |
|
)); |
|
} |
|
|
|
function parseObjectType(allowStatic) { |
|
var callProperties = [], indexers = [], marker, optional = false, |
|
properties = [], propertyKey, propertyTypeAnnotation, |
|
token, isStatic, matchStatic; |
|
|
|
expect('{'); |
|
|
|
while (!match('}')) { |
|
marker = markerCreate(); |
|
matchStatic = |
|
strict |
|
? matchKeyword('static') |
|
: matchContextualKeyword('static'); |
|
|
|
if (allowStatic && matchStatic) { |
|
token = lex(); |
|
isStatic = true; |
|
} |
|
|
|
if (match('[')) { |
|
indexers.push(parseObjectTypeIndexer(marker, isStatic)); |
|
} else if (match('(') || match('<')) { |
|
callProperties.push(parseObjectTypeCallProperty(marker, allowStatic)); |
|
} else { |
|
if (isStatic && match(':')) { |
|
propertyKey = markerApply(marker, delegate.createIdentifier(token)); |
|
throwErrorTolerant(token, Messages.StrictReservedWord); |
|
} else { |
|
propertyKey = parseObjectPropertyKey(); |
|
} |
|
if (match('<') || match('(')) { |
|
// This is a method property |
|
properties.push(parseObjectTypeMethod(marker, isStatic, propertyKey)); |
|
} else { |
|
if (match('?')) { |
|
lex(); |
|
optional = true; |
|
} |
|
expect(':'); |
|
propertyTypeAnnotation = parseType(); |
|
properties.push(markerApply(marker, delegate.createObjectTypeProperty( |
|
propertyKey, |
|
propertyTypeAnnotation, |
|
optional, |
|
isStatic |
|
))); |
|
} |
|
} |
|
|
|
if (match(';')) { |
|
lex(); |
|
} else if (!match('}')) { |
|
throwUnexpected(lookahead); |
|
} |
|
} |
|
|
|
expect('}'); |
|
|
|
return delegate.createObjectTypeAnnotation( |
|
properties, |
|
indexers, |
|
callProperties |
|
); |
|
} |
|
|
|
function parseGenericType() { |
|
var marker = markerCreate(), |
|
typeParameters = null, typeIdentifier; |
|
|
|
typeIdentifier = parseVariableIdentifier(); |
|
|
|
while (match('.')) { |
|
expect('.'); |
|
typeIdentifier = markerApply(marker, delegate.createQualifiedTypeIdentifier( |
|
typeIdentifier, |
|
parseVariableIdentifier() |
|
)); |
|
} |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterInstantiation(); |
|
} |
|
|
|
return markerApply(marker, delegate.createGenericTypeAnnotation( |
|
typeIdentifier, |
|
typeParameters |
|
)); |
|
} |
|
|
|
function parseVoidType() { |
|
var marker = markerCreate(); |
|
expectKeyword('void'); |
|
return markerApply(marker, delegate.createVoidTypeAnnotation()); |
|
} |
|
|
|
function parseTypeofType() { |
|
var argument, marker = markerCreate(); |
|
expectKeyword('typeof'); |
|
argument = parsePrimaryType(); |
|
return markerApply(marker, delegate.createTypeofTypeAnnotation( |
|
argument |
|
)); |
|
} |
|
|
|
function parseTupleType() { |
|
var marker = markerCreate(), types = []; |
|
expect('['); |
|
// We allow trailing commas |
|
while (index < length && !match(']')) { |
|
types.push(parseType()); |
|
if (match(']')) { |
|
break; |
|
} |
|
expect(','); |
|
} |
|
expect(']'); |
|
return markerApply(marker, delegate.createTupleTypeAnnotation( |
|
types |
|
)); |
|
} |
|
|
|
function parseFunctionTypeParam() { |
|
var marker = markerCreate(), name, optional = false, typeAnnotation; |
|
name = parseVariableIdentifier(); |
|
if (match('?')) { |
|
lex(); |
|
optional = true; |
|
} |
|
expect(':'); |
|
typeAnnotation = parseType(); |
|
return markerApply(marker, delegate.createFunctionTypeParam( |
|
name, |
|
typeAnnotation, |
|
optional |
|
)); |
|
} |
|
|
|
function parseFunctionTypeParams() { |
|
var ret = { params: [], rest: null }; |
|
while (lookahead.type === Token.Identifier) { |
|
ret.params.push(parseFunctionTypeParam()); |
|
if (!match(')')) { |
|
expect(','); |
|
} |
|
} |
|
|
|
if (match('...')) { |
|
lex(); |
|
ret.rest = parseFunctionTypeParam(); |
|
} |
|
return ret; |
|
} |
|
|
|
// The parsing of types roughly parallels the parsing of expressions, and |
|
// primary types are kind of like primary expressions...they're the |
|
// primitives with which other types are constructed. |
|
function parsePrimaryType() { |
|
var params = null, returnType = null, |
|
marker = markerCreate(), rest = null, tmp, |
|
typeParameters, token, type, isGroupedType = false; |
|
|
|
switch (lookahead.type) { |
|
case Token.Identifier: |
|
switch (lookahead.value) { |
|
case 'any': |
|
lex(); |
|
return markerApply(marker, delegate.createAnyTypeAnnotation()); |
|
case 'bool': // fallthrough |
|
case 'boolean': |
|
lex(); |
|
return markerApply(marker, delegate.createBooleanTypeAnnotation()); |
|
case 'number': |
|
lex(); |
|
return markerApply(marker, delegate.createNumberTypeAnnotation()); |
|
case 'string': |
|
lex(); |
|
return markerApply(marker, delegate.createStringTypeAnnotation()); |
|
} |
|
return markerApply(marker, parseGenericType()); |
|
case Token.Punctuator: |
|
switch (lookahead.value) { |
|
case '{': |
|
return markerApply(marker, parseObjectType()); |
|
case '[': |
|
return parseTupleType(); |
|
case '<': |
|
typeParameters = parseTypeParameterDeclaration(); |
|
expect('('); |
|
tmp = parseFunctionTypeParams(); |
|
params = tmp.params; |
|
rest = tmp.rest; |
|
expect(')'); |
|
|
|
expect('=>'); |
|
|
|
returnType = parseType(); |
|
|
|
return markerApply(marker, delegate.createFunctionTypeAnnotation( |
|
params, |
|
returnType, |
|
rest, |
|
typeParameters |
|
)); |
|
case '(': |
|
lex(); |
|
// Check to see if this is actually a grouped type |
|
if (!match(')') && !match('...')) { |
|
if (lookahead.type === Token.Identifier) { |
|
token = lookahead2(); |
|
isGroupedType = token.value !== '?' && token.value !== ':'; |
|
} else { |
|
isGroupedType = true; |
|
} |
|
} |
|
|
|
if (isGroupedType) { |
|
type = parseType(); |
|
expect(')'); |
|
|
|
// If we see a => next then someone was probably confused about |
|
// function types, so we can provide a better error message |
|
if (match('=>')) { |
|
throwError({}, Messages.ConfusedAboutFunctionType); |
|
} |
|
|
|
return type; |
|
} |
|
|
|
tmp = parseFunctionTypeParams(); |
|
params = tmp.params; |
|
rest = tmp.rest; |
|
|
|
expect(')'); |
|
|
|
expect('=>'); |
|
|
|
returnType = parseType(); |
|
|
|
return markerApply(marker, delegate.createFunctionTypeAnnotation( |
|
params, |
|
returnType, |
|
rest, |
|
null /* typeParameters */ |
|
)); |
|
} |
|
break; |
|
case Token.Keyword: |
|
switch (lookahead.value) { |
|
case 'void': |
|
return markerApply(marker, parseVoidType()); |
|
case 'typeof': |
|
return markerApply(marker, parseTypeofType()); |
|
} |
|
break; |
|
case Token.StringLiteral: |
|
token = lex(); |
|
if (token.octal) { |
|
throwError(token, Messages.StrictOctalLiteral); |
|
} |
|
return markerApply(marker, delegate.createStringLiteralTypeAnnotation( |
|
token |
|
)); |
|
} |
|
|
|
throwUnexpected(lookahead); |
|
} |
|
|
|
function parsePostfixType() { |
|
var marker = markerCreate(), t = parsePrimaryType(); |
|
if (match('[')) { |
|
expect('['); |
|
expect(']'); |
|
return markerApply(marker, delegate.createArrayTypeAnnotation(t)); |
|
} |
|
return t; |
|
} |
|
|
|
function parsePrefixType() { |
|
var marker = markerCreate(); |
|
if (match('?')) { |
|
lex(); |
|
return markerApply(marker, delegate.createNullableTypeAnnotation( |
|
parsePrefixType() |
|
)); |
|
} |
|
return parsePostfixType(); |
|
} |
|
|
|
|
|
function parseIntersectionType() { |
|
var marker = markerCreate(), type, types; |
|
type = parsePrefixType(); |
|
types = [type]; |
|
while (match('&')) { |
|
lex(); |
|
types.push(parsePrefixType()); |
|
} |
|
|
|
return types.length === 1 ? |
|
type : |
|
markerApply(marker, delegate.createIntersectionTypeAnnotation( |
|
types |
|
)); |
|
} |
|
|
|
function parseUnionType() { |
|
var marker = markerCreate(), type, types; |
|
type = parseIntersectionType(); |
|
types = [type]; |
|
while (match('|')) { |
|
lex(); |
|
types.push(parseIntersectionType()); |
|
} |
|
return types.length === 1 ? |
|
type : |
|
markerApply(marker, delegate.createUnionTypeAnnotation( |
|
types |
|
)); |
|
} |
|
|
|
function parseType() { |
|
var oldInType = state.inType, type; |
|
state.inType = true; |
|
|
|
type = parseUnionType(); |
|
|
|
state.inType = oldInType; |
|
return type; |
|
} |
|
|
|
function parseTypeAnnotation() { |
|
var marker = markerCreate(), type; |
|
|
|
expect(':'); |
|
type = parseType(); |
|
|
|
return markerApply(marker, delegate.createTypeAnnotation(type)); |
|
} |
|
|
|
function parseVariableIdentifier() { |
|
var marker = markerCreate(), |
|
token = lex(); |
|
|
|
if (token.type !== Token.Identifier) { |
|
throwUnexpected(token); |
|
} |
|
|
|
return markerApply(marker, delegate.createIdentifier(token.value)); |
|
} |
|
|
|
function parseTypeAnnotatableIdentifier(requireTypeAnnotation, canBeOptionalParam) { |
|
var marker = markerCreate(), |
|
ident = parseVariableIdentifier(), |
|
isOptionalParam = false; |
|
|
|
if (canBeOptionalParam && match('?')) { |
|
expect('?'); |
|
isOptionalParam = true; |
|
} |
|
|
|
if (requireTypeAnnotation || match(':')) { |
|
ident.typeAnnotation = parseTypeAnnotation(); |
|
ident = markerApply(marker, ident); |
|
} |
|
|
|
if (isOptionalParam) { |
|
ident.optional = true; |
|
ident = markerApply(marker, ident); |
|
} |
|
|
|
return ident; |
|
} |
|
|
|
function parseVariableDeclaration(kind) { |
|
var id, |
|
marker = markerCreate(), |
|
init = null, |
|
typeAnnotationMarker = markerCreate(); |
|
if (match('{')) { |
|
id = parseObjectInitialiser(); |
|
reinterpretAsAssignmentBindingPattern(id); |
|
if (match(':')) { |
|
id.typeAnnotation = parseTypeAnnotation(); |
|
markerApply(typeAnnotationMarker, id); |
|
} |
|
} else if (match('[')) { |
|
id = parseArrayInitialiser(); |
|
reinterpretAsAssignmentBindingPattern(id); |
|
if (match(':')) { |
|
id.typeAnnotation = parseTypeAnnotation(); |
|
markerApply(typeAnnotationMarker, id); |
|
} |
|
} else { |
|
/* istanbul ignore next */ |
|
id = state.allowKeyword ? parseNonComputedProperty() : parseTypeAnnotatableIdentifier(); |
|
// 12.2.1 |
|
if (strict && isRestrictedWord(id.name)) { |
|
throwErrorTolerant({}, Messages.StrictVarName); |
|
} |
|
} |
|
|
|
if (kind === 'const') { |
|
if (!match('=')) { |
|
throwError({}, Messages.NoUninitializedConst); |
|
} |
|
expect('='); |
|
init = parseAssignmentExpression(); |
|
} else if (match('=')) { |
|
lex(); |
|
init = parseAssignmentExpression(); |
|
} |
|
|
|
return markerApply(marker, delegate.createVariableDeclarator(id, init)); |
|
} |
|
|
|
function parseVariableDeclarationList(kind) { |
|
var list = []; |
|
|
|
do { |
|
list.push(parseVariableDeclaration(kind)); |
|
if (!match(',')) { |
|
break; |
|
} |
|
lex(); |
|
} while (index < length); |
|
|
|
return list; |
|
} |
|
|
|
function parseVariableStatement() { |
|
var declarations, marker = markerCreate(); |
|
|
|
expectKeyword('var'); |
|
|
|
declarations = parseVariableDeclarationList(); |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createVariableDeclaration(declarations, 'var')); |
|
} |
|
|
|
// kind may be `const` or `let` |
|
// Both are experimental and not in the specification yet. |
|
// see http://wiki.ecmascript.org/doku.php?id=harmony:const |
|
// and http://wiki.ecmascript.org/doku.php?id=harmony:let |
|
function parseConstLetDeclaration(kind) { |
|
var declarations, marker = markerCreate(); |
|
|
|
expectKeyword(kind); |
|
|
|
declarations = parseVariableDeclarationList(kind); |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createVariableDeclaration(declarations, kind)); |
|
} |
|
|
|
// people.mozilla.org/~jorendorff/es6-draft.html |
|
|
|
function parseModuleSpecifier() { |
|
var marker = markerCreate(), |
|
specifier; |
|
|
|
if (lookahead.type !== Token.StringLiteral) { |
|
throwError({}, Messages.InvalidModuleSpecifier); |
|
} |
|
specifier = delegate.createModuleSpecifier(lookahead); |
|
lex(); |
|
return markerApply(marker, specifier); |
|
} |
|
|
|
function parseExportBatchSpecifier() { |
|
var marker = markerCreate(); |
|
expect('*'); |
|
return markerApply(marker, delegate.createExportBatchSpecifier()); |
|
} |
|
|
|
function parseExportSpecifier() { |
|
var id, name = null, marker = markerCreate(), from; |
|
if (matchKeyword('default')) { |
|
lex(); |
|
id = markerApply(marker, delegate.createIdentifier('default')); |
|
// export {default} from "something"; |
|
} else { |
|
id = parseVariableIdentifier(); |
|
} |
|
if (matchContextualKeyword('as')) { |
|
lex(); |
|
name = parseNonComputedProperty(); |
|
} |
|
|
|
return markerApply(marker, delegate.createExportSpecifier(id, name)); |
|
} |
|
|
|
function parseExportDeclaration() { |
|
var declaration = null, |
|
possibleIdentifierToken, sourceElement, |
|
isExportFromIdentifier, |
|
src = null, specifiers = [], |
|
marker = markerCreate(); |
|
|
|
expectKeyword('export'); |
|
|
|
if (matchKeyword('default')) { |
|
// covers: |
|
// export default ... |
|
lex(); |
|
if (matchKeyword('function') || matchKeyword('class')) { |
|
possibleIdentifierToken = lookahead2(); |
|
if (isIdentifierName(possibleIdentifierToken)) { |
|
// covers: |
|
// export default function foo () {} |
|
// export default class foo {} |
|
sourceElement = parseSourceElement(); |
|
return markerApply(marker, delegate.createExportDeclaration(true, sourceElement, [sourceElement.id], null)); |
|
} |
|
// covers: |
|
// export default function () {} |
|
// export default class {} |
|
switch (lookahead.value) { |
|
case 'class': |
|
return markerApply(marker, delegate.createExportDeclaration(true, parseClassExpression(), [], null)); |
|
case 'function': |
|
return markerApply(marker, delegate.createExportDeclaration(true, parseFunctionExpression(), [], null)); |
|
} |
|
} |
|
|
|
if (matchContextualKeyword('from')) { |
|
throwError({}, Messages.UnexpectedToken, lookahead.value); |
|
} |
|
|
|
// covers: |
|
// export default {}; |
|
// export default []; |
|
if (match('{')) { |
|
declaration = parseObjectInitialiser(); |
|
} else if (match('[')) { |
|
declaration = parseArrayInitialiser(); |
|
} else { |
|
declaration = parseAssignmentExpression(); |
|
} |
|
consumeSemicolon(); |
|
return markerApply(marker, delegate.createExportDeclaration(true, declaration, [], null)); |
|
} |
|
|
|
// non-default export |
|
if (lookahead.type === Token.Keyword || matchContextualKeyword('type')) { |
|
// covers: |
|
// export var f = 1; |
|
switch (lookahead.value) { |
|
case 'type': |
|
case 'let': |
|
case 'const': |
|
case 'var': |
|
case 'class': |
|
case 'function': |
|
return markerApply(marker, delegate.createExportDeclaration(false, parseSourceElement(), specifiers, null)); |
|
} |
|
} |
|
|
|
if (match('*')) { |
|
// covers: |
|
// export * from "foo"; |
|
specifiers.push(parseExportBatchSpecifier()); |
|
|
|
if (!matchContextualKeyword('from')) { |
|
throwError({}, lookahead.value ? |
|
Messages.UnexpectedToken : Messages.MissingFromClause, lookahead.value); |
|
} |
|
lex(); |
|
src = parseModuleSpecifier(); |
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createExportDeclaration(false, null, specifiers, src)); |
|
} |
|
|
|
expect('{'); |
|
if (!match('}')) { |
|
do { |
|
isExportFromIdentifier = isExportFromIdentifier || matchKeyword('default'); |
|
specifiers.push(parseExportSpecifier()); |
|
} while (match(',') && lex()); |
|
} |
|
expect('}'); |
|
|
|
if (matchContextualKeyword('from')) { |
|
// covering: |
|
// export {default} from "foo"; |
|
// export {foo} from "foo"; |
|
lex(); |
|
src = parseModuleSpecifier(); |
|
consumeSemicolon(); |
|
} else if (isExportFromIdentifier) { |
|
// covering: |
|
// export {default}; // missing fromClause |
|
throwError({}, lookahead.value ? |
|
Messages.UnexpectedToken : Messages.MissingFromClause, lookahead.value); |
|
} else { |
|
// cover |
|
// export {foo}; |
|
consumeSemicolon(); |
|
} |
|
return markerApply(marker, delegate.createExportDeclaration(false, declaration, specifiers, src)); |
|
} |
|
|
|
|
|
function parseImportSpecifier() { |
|
// import {<foo as bar>} ...; |
|
var id, name = null, marker = markerCreate(); |
|
|
|
id = parseNonComputedProperty(); |
|
if (matchContextualKeyword('as')) { |
|
lex(); |
|
name = parseVariableIdentifier(); |
|
} |
|
|
|
return markerApply(marker, delegate.createImportSpecifier(id, name)); |
|
} |
|
|
|
function parseNamedImports() { |
|
var specifiers = []; |
|
// {foo, bar as bas} |
|
expect('{'); |
|
if (!match('}')) { |
|
do { |
|
specifiers.push(parseImportSpecifier()); |
|
} while (match(',') && lex()); |
|
} |
|
expect('}'); |
|
return specifiers; |
|
} |
|
|
|
function parseImportDefaultSpecifier() { |
|
// import <foo> ...; |
|
var id, marker = markerCreate(); |
|
|
|
id = parseNonComputedProperty(); |
|
|
|
return markerApply(marker, delegate.createImportDefaultSpecifier(id)); |
|
} |
|
|
|
function parseImportNamespaceSpecifier() { |
|
// import <* as foo> ...; |
|
var id, marker = markerCreate(); |
|
|
|
expect('*'); |
|
if (!matchContextualKeyword('as')) { |
|
throwError({}, Messages.NoAsAfterImportNamespace); |
|
} |
|
lex(); |
|
id = parseNonComputedProperty(); |
|
|
|
return markerApply(marker, delegate.createImportNamespaceSpecifier(id)); |
|
} |
|
|
|
function parseImportDeclaration() { |
|
var specifiers, src, marker = markerCreate(), isType = false, token2; |
|
|
|
expectKeyword('import'); |
|
|
|
if (matchContextualKeyword('type')) { |
|
token2 = lookahead2(); |
|
if ((token2.type === Token.Identifier && token2.value !== 'from') || |
|
(token2.type === Token.Punctuator && |
|
(token2.value === '{' || token2.value === '*'))) { |
|
isType = true; |
|
lex(); |
|
} |
|
} |
|
|
|
specifiers = []; |
|
|
|
if (lookahead.type === Token.StringLiteral) { |
|
// covers: |
|
// import "foo"; |
|
src = parseModuleSpecifier(); |
|
consumeSemicolon(); |
|
return markerApply(marker, delegate.createImportDeclaration(specifiers, src, isType)); |
|
} |
|
|
|
if (!matchKeyword('default') && isIdentifierName(lookahead)) { |
|
// covers: |
|
// import foo |
|
// import foo, ... |
|
specifiers.push(parseImportDefaultSpecifier()); |
|
if (match(',')) { |
|
lex(); |
|
} |
|
} |
|
if (match('*')) { |
|
// covers: |
|
// import foo, * as foo |
|
// import * as foo |
|
specifiers.push(parseImportNamespaceSpecifier()); |
|
} else if (match('{')) { |
|
// covers: |
|
// import foo, {bar} |
|
// import {bar} |
|
specifiers = specifiers.concat(parseNamedImports()); |
|
} |
|
|
|
if (!matchContextualKeyword('from')) { |
|
throwError({}, lookahead.value ? |
|
Messages.UnexpectedToken : Messages.MissingFromClause, lookahead.value); |
|
} |
|
lex(); |
|
src = parseModuleSpecifier(); |
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createImportDeclaration(specifiers, src, isType)); |
|
} |
|
|
|
// 12.3 Empty Statement |
|
|
|
function parseEmptyStatement() { |
|
var marker = markerCreate(); |
|
expect(';'); |
|
return markerApply(marker, delegate.createEmptyStatement()); |
|
} |
|
|
|
// 12.4 Expression Statement |
|
|
|
function parseExpressionStatement() { |
|
var marker = markerCreate(), expr = parseExpression(); |
|
consumeSemicolon(); |
|
return markerApply(marker, delegate.createExpressionStatement(expr)); |
|
} |
|
|
|
// 12.5 If statement |
|
|
|
function parseIfStatement() { |
|
var test, consequent, alternate, marker = markerCreate(); |
|
|
|
expectKeyword('if'); |
|
|
|
expect('('); |
|
|
|
test = parseExpression(); |
|
|
|
expect(')'); |
|
|
|
consequent = parseStatement(); |
|
|
|
if (matchKeyword('else')) { |
|
lex(); |
|
alternate = parseStatement(); |
|
} else { |
|
alternate = null; |
|
} |
|
|
|
return markerApply(marker, delegate.createIfStatement(test, consequent, alternate)); |
|
} |
|
|
|
// 12.6 Iteration Statements |
|
|
|
function parseDoWhileStatement() { |
|
var body, test, oldInIteration, marker = markerCreate(); |
|
|
|
expectKeyword('do'); |
|
|
|
oldInIteration = state.inIteration; |
|
state.inIteration = true; |
|
|
|
body = parseStatement(); |
|
|
|
state.inIteration = oldInIteration; |
|
|
|
expectKeyword('while'); |
|
|
|
expect('('); |
|
|
|
test = parseExpression(); |
|
|
|
expect(')'); |
|
|
|
if (match(';')) { |
|
lex(); |
|
} |
|
|
|
return markerApply(marker, delegate.createDoWhileStatement(body, test)); |
|
} |
|
|
|
function parseWhileStatement() { |
|
var test, body, oldInIteration, marker = markerCreate(); |
|
|
|
expectKeyword('while'); |
|
|
|
expect('('); |
|
|
|
test = parseExpression(); |
|
|
|
expect(')'); |
|
|
|
oldInIteration = state.inIteration; |
|
state.inIteration = true; |
|
|
|
body = parseStatement(); |
|
|
|
state.inIteration = oldInIteration; |
|
|
|
return markerApply(marker, delegate.createWhileStatement(test, body)); |
|
} |
|
|
|
function parseForVariableDeclaration() { |
|
var marker = markerCreate(), |
|
token = lex(), |
|
declarations = parseVariableDeclarationList(); |
|
|
|
return markerApply(marker, delegate.createVariableDeclaration(declarations, token.value)); |
|
} |
|
|
|
function parseForStatement(opts) { |
|
var init, test, update, left, right, body, operator, oldInIteration, |
|
marker = markerCreate(); |
|
init = test = update = null; |
|
expectKeyword('for'); |
|
|
|
// http://wiki.ecmascript.org/doku.php?id=proposals:iterators_and_generators&s=each |
|
if (matchContextualKeyword('each')) { |
|
throwError({}, Messages.EachNotAllowed); |
|
} |
|
|
|
expect('('); |
|
|
|
if (match(';')) { |
|
lex(); |
|
} else { |
|
if (matchKeyword('var') || matchKeyword('let') || matchKeyword('const')) { |
|
state.allowIn = false; |
|
init = parseForVariableDeclaration(); |
|
state.allowIn = true; |
|
|
|
if (init.declarations.length === 1) { |
|
if (matchKeyword('in') || matchContextualKeyword('of')) { |
|
operator = lookahead; |
|
if (!((operator.value === 'in' || init.kind !== 'var') && init.declarations[0].init)) { |
|
lex(); |
|
left = init; |
|
right = parseExpression(); |
|
init = null; |
|
} |
|
} |
|
} |
|
} else { |
|
state.allowIn = false; |
|
init = parseExpression(); |
|
state.allowIn = true; |
|
|
|
if (matchContextualKeyword('of')) { |
|
operator = lex(); |
|
left = init; |
|
right = parseExpression(); |
|
init = null; |
|
} else if (matchKeyword('in')) { |
|
// LeftHandSideExpression |
|
if (!isAssignableLeftHandSide(init)) { |
|
throwError({}, Messages.InvalidLHSInForIn); |
|
} |
|
operator = lex(); |
|
left = init; |
|
right = parseExpression(); |
|
init = null; |
|
} |
|
} |
|
|
|
if (typeof left === 'undefined') { |
|
expect(';'); |
|
} |
|
} |
|
|
|
if (typeof left === 'undefined') { |
|
|
|
if (!match(';')) { |
|
test = parseExpression(); |
|
} |
|
expect(';'); |
|
|
|
if (!match(')')) { |
|
update = parseExpression(); |
|
} |
|
} |
|
|
|
expect(')'); |
|
|
|
oldInIteration = state.inIteration; |
|
state.inIteration = true; |
|
|
|
if (!(opts !== undefined && opts.ignoreBody)) { |
|
body = parseStatement(); |
|
} |
|
|
|
state.inIteration = oldInIteration; |
|
|
|
if (typeof left === 'undefined') { |
|
return markerApply(marker, delegate.createForStatement(init, test, update, body)); |
|
} |
|
|
|
if (operator.value === 'in') { |
|
return markerApply(marker, delegate.createForInStatement(left, right, body)); |
|
} |
|
return markerApply(marker, delegate.createForOfStatement(left, right, body)); |
|
} |
|
|
|
// 12.7 The continue statement |
|
|
|
function parseContinueStatement() { |
|
var label = null, marker = markerCreate(); |
|
|
|
expectKeyword('continue'); |
|
|
|
// Optimize the most common form: 'continue;'. |
|
if (source.charCodeAt(index) === 59) { |
|
lex(); |
|
|
|
if (!state.inIteration) { |
|
throwError({}, Messages.IllegalContinue); |
|
} |
|
|
|
return markerApply(marker, delegate.createContinueStatement(null)); |
|
} |
|
|
|
if (peekLineTerminator()) { |
|
if (!state.inIteration) { |
|
throwError({}, Messages.IllegalContinue); |
|
} |
|
|
|
return markerApply(marker, delegate.createContinueStatement(null)); |
|
} |
|
|
|
if (lookahead.type === Token.Identifier) { |
|
label = parseVariableIdentifier(); |
|
|
|
if (!state.labelSet.has(label.name)) { |
|
throwError({}, Messages.UnknownLabel, label.name); |
|
} |
|
} |
|
|
|
consumeSemicolon(); |
|
|
|
if (label === null && !state.inIteration) { |
|
throwError({}, Messages.IllegalContinue); |
|
} |
|
|
|
return markerApply(marker, delegate.createContinueStatement(label)); |
|
} |
|
|
|
// 12.8 The break statement |
|
|
|
function parseBreakStatement() { |
|
var label = null, marker = markerCreate(); |
|
|
|
expectKeyword('break'); |
|
|
|
// Catch the very common case first: immediately a semicolon (char #59). |
|
if (source.charCodeAt(index) === 59) { |
|
lex(); |
|
|
|
if (!(state.inIteration || state.inSwitch)) { |
|
throwError({}, Messages.IllegalBreak); |
|
} |
|
|
|
return markerApply(marker, delegate.createBreakStatement(null)); |
|
} |
|
|
|
if (peekLineTerminator()) { |
|
if (!(state.inIteration || state.inSwitch)) { |
|
throwError({}, Messages.IllegalBreak); |
|
} |
|
|
|
return markerApply(marker, delegate.createBreakStatement(null)); |
|
} |
|
|
|
if (lookahead.type === Token.Identifier) { |
|
label = parseVariableIdentifier(); |
|
|
|
if (!state.labelSet.has(label.name)) { |
|
throwError({}, Messages.UnknownLabel, label.name); |
|
} |
|
} |
|
|
|
consumeSemicolon(); |
|
|
|
if (label === null && !(state.inIteration || state.inSwitch)) { |
|
throwError({}, Messages.IllegalBreak); |
|
} |
|
|
|
return markerApply(marker, delegate.createBreakStatement(label)); |
|
} |
|
|
|
// 12.9 The return statement |
|
|
|
function parseReturnStatement() { |
|
var argument = null, marker = markerCreate(); |
|
|
|
expectKeyword('return'); |
|
|
|
if (!state.inFunctionBody) { |
|
throwErrorTolerant({}, Messages.IllegalReturn); |
|
} |
|
|
|
// 'return' followed by a space and an identifier is very common. |
|
if (source.charCodeAt(index) === 32) { |
|
if (isIdentifierStart(source.charCodeAt(index + 1))) { |
|
argument = parseExpression(); |
|
consumeSemicolon(); |
|
return markerApply(marker, delegate.createReturnStatement(argument)); |
|
} |
|
} |
|
|
|
if (peekLineTerminator()) { |
|
return markerApply(marker, delegate.createReturnStatement(null)); |
|
} |
|
|
|
if (!match(';')) { |
|
if (!match('}') && lookahead.type !== Token.EOF) { |
|
argument = parseExpression(); |
|
} |
|
} |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createReturnStatement(argument)); |
|
} |
|
|
|
// 12.10 The with statement |
|
|
|
function parseWithStatement() { |
|
var object, body, marker = markerCreate(); |
|
|
|
if (strict) { |
|
throwErrorTolerant({}, Messages.StrictModeWith); |
|
} |
|
|
|
expectKeyword('with'); |
|
|
|
expect('('); |
|
|
|
object = parseExpression(); |
|
|
|
expect(')'); |
|
|
|
body = parseStatement(); |
|
|
|
return markerApply(marker, delegate.createWithStatement(object, body)); |
|
} |
|
|
|
// 12.10 The swith statement |
|
|
|
function parseSwitchCase() { |
|
var test, |
|
consequent = [], |
|
sourceElement, |
|
marker = markerCreate(); |
|
|
|
if (matchKeyword('default')) { |
|
lex(); |
|
test = null; |
|
} else { |
|
expectKeyword('case'); |
|
test = parseExpression(); |
|
} |
|
expect(':'); |
|
|
|
while (index < length) { |
|
if (match('}') || matchKeyword('default') || matchKeyword('case')) { |
|
break; |
|
} |
|
sourceElement = parseSourceElement(); |
|
if (typeof sourceElement === 'undefined') { |
|
break; |
|
} |
|
consequent.push(sourceElement); |
|
} |
|
|
|
return markerApply(marker, delegate.createSwitchCase(test, consequent)); |
|
} |
|
|
|
function parseSwitchStatement() { |
|
var discriminant, cases, clause, oldInSwitch, defaultFound, marker = markerCreate(); |
|
|
|
expectKeyword('switch'); |
|
|
|
expect('('); |
|
|
|
discriminant = parseExpression(); |
|
|
|
expect(')'); |
|
|
|
expect('{'); |
|
|
|
cases = []; |
|
|
|
if (match('}')) { |
|
lex(); |
|
return markerApply(marker, delegate.createSwitchStatement(discriminant, cases)); |
|
} |
|
|
|
oldInSwitch = state.inSwitch; |
|
state.inSwitch = true; |
|
defaultFound = false; |
|
|
|
while (index < length) { |
|
if (match('}')) { |
|
break; |
|
} |
|
clause = parseSwitchCase(); |
|
if (clause.test === null) { |
|
if (defaultFound) { |
|
throwError({}, Messages.MultipleDefaultsInSwitch); |
|
} |
|
defaultFound = true; |
|
} |
|
cases.push(clause); |
|
} |
|
|
|
state.inSwitch = oldInSwitch; |
|
|
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createSwitchStatement(discriminant, cases)); |
|
} |
|
|
|
// 12.13 The throw statement |
|
|
|
function parseThrowStatement() { |
|
var argument, marker = markerCreate(); |
|
|
|
expectKeyword('throw'); |
|
|
|
if (peekLineTerminator()) { |
|
throwError({}, Messages.NewlineAfterThrow); |
|
} |
|
|
|
argument = parseExpression(); |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createThrowStatement(argument)); |
|
} |
|
|
|
// 12.14 The try statement |
|
|
|
function parseCatchClause() { |
|
var param, body, marker = markerCreate(); |
|
|
|
expectKeyword('catch'); |
|
|
|
expect('('); |
|
if (match(')')) { |
|
throwUnexpected(lookahead); |
|
} |
|
|
|
param = parseExpression(); |
|
// 12.14.1 |
|
if (strict && param.type === Syntax.Identifier && isRestrictedWord(param.name)) { |
|
throwErrorTolerant({}, Messages.StrictCatchVariable); |
|
} |
|
|
|
expect(')'); |
|
body = parseBlock(); |
|
return markerApply(marker, delegate.createCatchClause(param, body)); |
|
} |
|
|
|
function parseTryStatement() { |
|
var block, handlers = [], finalizer = null, marker = markerCreate(); |
|
|
|
expectKeyword('try'); |
|
|
|
block = parseBlock(); |
|
|
|
if (matchKeyword('catch')) { |
|
handlers.push(parseCatchClause()); |
|
} |
|
|
|
if (matchKeyword('finally')) { |
|
lex(); |
|
finalizer = parseBlock(); |
|
} |
|
|
|
if (handlers.length === 0 && !finalizer) { |
|
throwError({}, Messages.NoCatchOrFinally); |
|
} |
|
|
|
return markerApply(marker, delegate.createTryStatement(block, [], handlers, finalizer)); |
|
} |
|
|
|
// 12.15 The debugger statement |
|
|
|
function parseDebuggerStatement() { |
|
var marker = markerCreate(); |
|
expectKeyword('debugger'); |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createDebuggerStatement()); |
|
} |
|
|
|
// 12 Statements |
|
|
|
function parseStatement() { |
|
var type = lookahead.type, |
|
marker, |
|
expr, |
|
labeledBody; |
|
|
|
if (type === Token.EOF) { |
|
throwUnexpected(lookahead); |
|
} |
|
|
|
if (type === Token.Punctuator) { |
|
switch (lookahead.value) { |
|
case ';': |
|
return parseEmptyStatement(); |
|
case '{': |
|
return parseBlock(); |
|
case '(': |
|
return parseExpressionStatement(); |
|
default: |
|
break; |
|
} |
|
} |
|
|
|
if (type === Token.Keyword) { |
|
switch (lookahead.value) { |
|
case 'break': |
|
return parseBreakStatement(); |
|
case 'continue': |
|
return parseContinueStatement(); |
|
case 'debugger': |
|
return parseDebuggerStatement(); |
|
case 'do': |
|
return parseDoWhileStatement(); |
|
case 'for': |
|
return parseForStatement(); |
|
case 'function': |
|
return parseFunctionDeclaration(); |
|
case 'class': |
|
return parseClassDeclaration(); |
|
case 'if': |
|
return parseIfStatement(); |
|
case 'return': |
|
return parseReturnStatement(); |
|
case 'switch': |
|
return parseSwitchStatement(); |
|
case 'throw': |
|
return parseThrowStatement(); |
|
case 'try': |
|
return parseTryStatement(); |
|
case 'var': |
|
return parseVariableStatement(); |
|
case 'while': |
|
return parseWhileStatement(); |
|
case 'with': |
|
return parseWithStatement(); |
|
default: |
|
break; |
|
} |
|
} |
|
|
|
if (matchAsyncFuncExprOrDecl()) { |
|
return parseFunctionDeclaration(); |
|
} |
|
|
|
marker = markerCreate(); |
|
expr = parseExpression(); |
|
|
|
// 12.12 Labelled Statements |
|
if ((expr.type === Syntax.Identifier) && match(':')) { |
|
lex(); |
|
|
|
if (state.labelSet.has(expr.name)) { |
|
throwError({}, Messages.Redeclaration, 'Label', expr.name); |
|
} |
|
|
|
state.labelSet.set(expr.name, true); |
|
labeledBody = parseStatement(); |
|
state.labelSet["delete"](expr.name); |
|
return markerApply(marker, delegate.createLabeledStatement(expr, labeledBody)); |
|
} |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createExpressionStatement(expr)); |
|
} |
|
|
|
// 13 Function Definition |
|
|
|
function parseConciseBody() { |
|
if (match('{')) { |
|
return parseFunctionSourceElements(); |
|
} |
|
return parseAssignmentExpression(); |
|
} |
|
|
|
function parseFunctionSourceElements() { |
|
var sourceElement, sourceElements = [], token, directive, firstRestricted, |
|
oldLabelSet, oldInIteration, oldInSwitch, oldInFunctionBody, oldParenthesizedCount, |
|
marker = markerCreate(); |
|
|
|
expect('{'); |
|
|
|
while (index < length) { |
|
if (lookahead.type !== Token.StringLiteral) { |
|
break; |
|
} |
|
token = lookahead; |
|
|
|
sourceElement = parseSourceElement(); |
|
sourceElements.push(sourceElement); |
|
if (sourceElement.expression.type !== Syntax.Literal) { |
|
// this is not directive |
|
break; |
|
} |
|
directive = source.slice(token.range[0] + 1, token.range[1] - 1); |
|
if (directive === 'use strict') { |
|
strict = true; |
|
if (firstRestricted) { |
|
throwErrorTolerant(firstRestricted, Messages.StrictOctalLiteral); |
|
} |
|
} else { |
|
if (!firstRestricted && token.octal) { |
|
firstRestricted = token; |
|
} |
|
} |
|
} |
|
|
|
oldLabelSet = state.labelSet; |
|
oldInIteration = state.inIteration; |
|
oldInSwitch = state.inSwitch; |
|
oldInFunctionBody = state.inFunctionBody; |
|
oldParenthesizedCount = state.parenthesizedCount; |
|
|
|
state.labelSet = new StringMap(); |
|
state.inIteration = false; |
|
state.inSwitch = false; |
|
state.inFunctionBody = true; |
|
state.parenthesizedCount = 0; |
|
|
|
while (index < length) { |
|
if (match('}')) { |
|
break; |
|
} |
|
sourceElement = parseSourceElement(); |
|
if (typeof sourceElement === 'undefined') { |
|
break; |
|
} |
|
sourceElements.push(sourceElement); |
|
} |
|
|
|
expect('}'); |
|
|
|
state.labelSet = oldLabelSet; |
|
state.inIteration = oldInIteration; |
|
state.inSwitch = oldInSwitch; |
|
state.inFunctionBody = oldInFunctionBody; |
|
state.parenthesizedCount = oldParenthesizedCount; |
|
|
|
return markerApply(marker, delegate.createBlockStatement(sourceElements)); |
|
} |
|
|
|
function validateParam(options, param, name) { |
|
if (strict) { |
|
if (isRestrictedWord(name)) { |
|
options.stricted = param; |
|
options.message = Messages.StrictParamName; |
|
} |
|
if (options.paramSet.has(name)) { |
|
options.stricted = param; |
|
options.message = Messages.StrictParamDupe; |
|
} |
|
} else if (!options.firstRestricted) { |
|
if (isRestrictedWord(name)) { |
|
options.firstRestricted = param; |
|
options.message = Messages.StrictParamName; |
|
} else if (isStrictModeReservedWord(name)) { |
|
options.firstRestricted = param; |
|
options.message = Messages.StrictReservedWord; |
|
} else if (options.paramSet.has(name)) { |
|
options.firstRestricted = param; |
|
options.message = Messages.StrictParamDupe; |
|
} |
|
} |
|
options.paramSet.set(name, true); |
|
} |
|
|
|
function parseParam(options) { |
|
var marker, token, rest, param, def; |
|
|
|
token = lookahead; |
|
if (token.value === '...') { |
|
token = lex(); |
|
rest = true; |
|
} |
|
|
|
if (match('[')) { |
|
marker = markerCreate(); |
|
param = parseArrayInitialiser(); |
|
reinterpretAsDestructuredParameter(options, param); |
|
if (match(':')) { |
|
param.typeAnnotation = parseTypeAnnotation(); |
|
markerApply(marker, param); |
|
} |
|
} else if (match('{')) { |
|
marker = markerCreate(); |
|
if (rest) { |
|
throwError({}, Messages.ObjectPatternAsRestParameter); |
|
} |
|
param = parseObjectInitialiser(); |
|
reinterpretAsDestructuredParameter(options, param); |
|
if (match(':')) { |
|
param.typeAnnotation = parseTypeAnnotation(); |
|
markerApply(marker, param); |
|
} |
|
} else { |
|
param = |
|
rest |
|
? parseTypeAnnotatableIdentifier( |
|
false, /* requireTypeAnnotation */ |
|
false /* canBeOptionalParam */ |
|
) |
|
: parseTypeAnnotatableIdentifier( |
|
false, /* requireTypeAnnotation */ |
|
true /* canBeOptionalParam */ |
|
); |
|
|
|
validateParam(options, token, token.value); |
|
} |
|
|
|
if (match('=')) { |
|
if (rest) { |
|
throwErrorTolerant(lookahead, Messages.DefaultRestParameter); |
|
} |
|
lex(); |
|
def = parseAssignmentExpression(); |
|
++options.defaultCount; |
|
} |
|
|
|
if (rest) { |
|
if (!match(')')) { |
|
throwError({}, Messages.ParameterAfterRestParameter); |
|
} |
|
options.rest = param; |
|
return false; |
|
} |
|
|
|
options.params.push(param); |
|
options.defaults.push(def); |
|
return !match(')'); |
|
} |
|
|
|
function parseParams(firstRestricted) { |
|
var options, marker = markerCreate(); |
|
|
|
options = { |
|
params: [], |
|
defaultCount: 0, |
|
defaults: [], |
|
rest: null, |
|
firstRestricted: firstRestricted |
|
}; |
|
|
|
expect('('); |
|
|
|
if (!match(')')) { |
|
options.paramSet = new StringMap(); |
|
while (index < length) { |
|
if (!parseParam(options)) { |
|
break; |
|
} |
|
expect(','); |
|
} |
|
} |
|
|
|
expect(')'); |
|
|
|
if (options.defaultCount === 0) { |
|
options.defaults = []; |
|
} |
|
|
|
if (match(':')) { |
|
options.returnType = parseTypeAnnotation(); |
|
} |
|
|
|
return markerApply(marker, options); |
|
} |
|
|
|
function parseFunctionDeclaration() { |
|
var id, body, token, tmp, firstRestricted, message, generator, isAsync, |
|
previousStrict, previousYieldAllowed, previousAwaitAllowed, |
|
marker = markerCreate(), typeParameters; |
|
|
|
isAsync = false; |
|
if (matchAsync()) { |
|
lex(); |
|
isAsync = true; |
|
} |
|
|
|
expectKeyword('function'); |
|
|
|
generator = false; |
|
if (match('*')) { |
|
lex(); |
|
generator = true; |
|
} |
|
|
|
token = lookahead; |
|
|
|
id = parseVariableIdentifier(); |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
if (strict) { |
|
if (isRestrictedWord(token.value)) { |
|
throwErrorTolerant(token, Messages.StrictFunctionName); |
|
} |
|
} else { |
|
if (isRestrictedWord(token.value)) { |
|
firstRestricted = token; |
|
message = Messages.StrictFunctionName; |
|
} else if (isStrictModeReservedWord(token.value)) { |
|
firstRestricted = token; |
|
message = Messages.StrictReservedWord; |
|
} |
|
} |
|
|
|
tmp = parseParams(firstRestricted); |
|
firstRestricted = tmp.firstRestricted; |
|
if (tmp.message) { |
|
message = tmp.message; |
|
} |
|
|
|
previousStrict = strict; |
|
previousYieldAllowed = state.yieldAllowed; |
|
state.yieldAllowed = generator; |
|
previousAwaitAllowed = state.awaitAllowed; |
|
state.awaitAllowed = isAsync; |
|
|
|
body = parseFunctionSourceElements(); |
|
|
|
if (strict && firstRestricted) { |
|
throwError(firstRestricted, message); |
|
} |
|
if (strict && tmp.stricted) { |
|
throwErrorTolerant(tmp.stricted, message); |
|
} |
|
strict = previousStrict; |
|
state.yieldAllowed = previousYieldAllowed; |
|
state.awaitAllowed = previousAwaitAllowed; |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createFunctionDeclaration( |
|
id, |
|
tmp.params, |
|
tmp.defaults, |
|
body, |
|
tmp.rest, |
|
generator, |
|
false, |
|
isAsync, |
|
tmp.returnType, |
|
typeParameters |
|
) |
|
); |
|
} |
|
|
|
function parseFunctionExpression() { |
|
var token, id = null, firstRestricted, message, tmp, body, generator, isAsync, |
|
previousStrict, previousYieldAllowed, previousAwaitAllowed, |
|
marker = markerCreate(), typeParameters; |
|
|
|
isAsync = false; |
|
if (matchAsync()) { |
|
lex(); |
|
isAsync = true; |
|
} |
|
|
|
expectKeyword('function'); |
|
|
|
generator = false; |
|
|
|
if (match('*')) { |
|
lex(); |
|
generator = true; |
|
} |
|
|
|
if (!match('(')) { |
|
if (!match('<')) { |
|
token = lookahead; |
|
id = parseVariableIdentifier(); |
|
|
|
if (strict) { |
|
if (isRestrictedWord(token.value)) { |
|
throwErrorTolerant(token, Messages.StrictFunctionName); |
|
} |
|
} else { |
|
if (isRestrictedWord(token.value)) { |
|
firstRestricted = token; |
|
message = Messages.StrictFunctionName; |
|
} else if (isStrictModeReservedWord(token.value)) { |
|
firstRestricted = token; |
|
message = Messages.StrictReservedWord; |
|
} |
|
} |
|
} |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
} |
|
|
|
tmp = parseParams(firstRestricted); |
|
firstRestricted = tmp.firstRestricted; |
|
if (tmp.message) { |
|
message = tmp.message; |
|
} |
|
|
|
previousStrict = strict; |
|
previousYieldAllowed = state.yieldAllowed; |
|
state.yieldAllowed = generator; |
|
previousAwaitAllowed = state.awaitAllowed; |
|
state.awaitAllowed = isAsync; |
|
|
|
body = parseFunctionSourceElements(); |
|
|
|
if (strict && firstRestricted) { |
|
throwError(firstRestricted, message); |
|
} |
|
if (strict && tmp.stricted) { |
|
throwErrorTolerant(tmp.stricted, message); |
|
} |
|
strict = previousStrict; |
|
state.yieldAllowed = previousYieldAllowed; |
|
state.awaitAllowed = previousAwaitAllowed; |
|
|
|
return markerApply( |
|
marker, |
|
delegate.createFunctionExpression( |
|
id, |
|
tmp.params, |
|
tmp.defaults, |
|
body, |
|
tmp.rest, |
|
generator, |
|
false, |
|
isAsync, |
|
tmp.returnType, |
|
typeParameters |
|
) |
|
); |
|
} |
|
|
|
function parseYieldExpression() { |
|
var delegateFlag, expr, marker = markerCreate(); |
|
|
|
expectKeyword('yield', !strict); |
|
|
|
delegateFlag = false; |
|
if (match('*')) { |
|
lex(); |
|
delegateFlag = true; |
|
} |
|
|
|
expr = parseAssignmentExpression(); |
|
|
|
return markerApply(marker, delegate.createYieldExpression(expr, delegateFlag)); |
|
} |
|
|
|
function parseAwaitExpression() { |
|
var expr, marker = markerCreate(); |
|
expectContextualKeyword('await'); |
|
expr = parseAssignmentExpression(); |
|
return markerApply(marker, delegate.createAwaitExpression(expr)); |
|
} |
|
|
|
// 14 Functions and classes |
|
|
|
// 14.1 Functions is defined above (13 in ES5) |
|
// 14.2 Arrow Functions Definitions is defined in (7.3 assignments) |
|
|
|
// 14.3 Method Definitions |
|
// 14.3.7 |
|
function specialMethod(methodDefinition) { |
|
return methodDefinition.kind === 'get' || |
|
methodDefinition.kind === 'set' || |
|
methodDefinition.value.generator; |
|
} |
|
|
|
function parseMethodDefinition(key, isStatic, generator, computed) { |
|
var token, param, propType, |
|
isAsync, typeParameters, tokenValue, returnType; |
|
|
|
propType = isStatic ? ClassPropertyType["static"] : ClassPropertyType.prototype; |
|
|
|
if (generator) { |
|
return delegate.createMethodDefinition( |
|
propType, |
|
'', |
|
key, |
|
parsePropertyMethodFunction({ generator: true }), |
|
computed |
|
); |
|
} |
|
|
|
tokenValue = key.type === 'Identifier' && key.name; |
|
|
|
if (tokenValue === 'get' && !match('(')) { |
|
key = parseObjectPropertyKey(); |
|
|
|
expect('('); |
|
expect(')'); |
|
if (match(':')) { |
|
returnType = parseTypeAnnotation(); |
|
} |
|
return delegate.createMethodDefinition( |
|
propType, |
|
'get', |
|
key, |
|
parsePropertyFunction({ generator: false, returnType: returnType }), |
|
computed |
|
); |
|
} |
|
if (tokenValue === 'set' && !match('(')) { |
|
key = parseObjectPropertyKey(); |
|
|
|
expect('('); |
|
token = lookahead; |
|
param = [ parseTypeAnnotatableIdentifier() ]; |
|
expect(')'); |
|
if (match(':')) { |
|
returnType = parseTypeAnnotation(); |
|
} |
|
return delegate.createMethodDefinition( |
|
propType, |
|
'set', |
|
key, |
|
parsePropertyFunction({ |
|
params: param, |
|
generator: false, |
|
name: token, |
|
returnType: returnType |
|
}), |
|
computed |
|
); |
|
} |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
isAsync = tokenValue === 'async' && !match('('); |
|
if (isAsync) { |
|
key = parseObjectPropertyKey(); |
|
} |
|
|
|
return delegate.createMethodDefinition( |
|
propType, |
|
'', |
|
key, |
|
parsePropertyMethodFunction({ |
|
generator: false, |
|
async: isAsync, |
|
typeParameters: typeParameters |
|
}), |
|
computed |
|
); |
|
} |
|
|
|
function parseClassProperty(key, computed, isStatic) { |
|
var typeAnnotation; |
|
|
|
typeAnnotation = parseTypeAnnotation(); |
|
expect(';'); |
|
|
|
return delegate.createClassProperty( |
|
key, |
|
typeAnnotation, |
|
computed, |
|
isStatic |
|
); |
|
} |
|
|
|
function parseClassElement() { |
|
var computed = false, generator = false, key, marker = markerCreate(), |
|
isStatic = false, possiblyOpenBracketToken; |
|
if (match(';')) { |
|
lex(); |
|
return undefined; |
|
} |
|
|
|
if (lookahead.value === 'static') { |
|
lex(); |
|
isStatic = true; |
|
} |
|
|
|
if (match('*')) { |
|
lex(); |
|
generator = true; |
|
} |
|
|
|
possiblyOpenBracketToken = lookahead; |
|
if (matchContextualKeyword('get') || matchContextualKeyword('set')) { |
|
possiblyOpenBracketToken = lookahead2(); |
|
} |
|
|
|
if (possiblyOpenBracketToken.type === Token.Punctuator |
|
&& possiblyOpenBracketToken.value === '[') { |
|
computed = true; |
|
} |
|
|
|
key = parseObjectPropertyKey(); |
|
|
|
if (!generator && lookahead.value === ':') { |
|
return markerApply(marker, parseClassProperty(key, computed, isStatic)); |
|
} |
|
|
|
return markerApply(marker, parseMethodDefinition( |
|
key, |
|
isStatic, |
|
generator, |
|
computed |
|
)); |
|
} |
|
|
|
function parseClassBody() { |
|
var classElement, classElements = [], existingProps = {}, |
|
marker = markerCreate(), propName, propType; |
|
|
|
existingProps[ClassPropertyType["static"]] = new StringMap(); |
|
existingProps[ClassPropertyType.prototype] = new StringMap(); |
|
|
|
expect('{'); |
|
|
|
while (index < length) { |
|
if (match('}')) { |
|
break; |
|
} |
|
classElement = parseClassElement(existingProps); |
|
|
|
if (typeof classElement !== 'undefined') { |
|
classElements.push(classElement); |
|
|
|
propName = !classElement.computed && getFieldName(classElement.key); |
|
if (propName !== false) { |
|
propType = classElement["static"] ? |
|
ClassPropertyType["static"] : |
|
ClassPropertyType.prototype; |
|
|
|
if (classElement.type === Syntax.MethodDefinition) { |
|
if (propName === 'constructor' && !classElement["static"]) { |
|
if (specialMethod(classElement)) { |
|
throwError(classElement, Messages.IllegalClassConstructorProperty); |
|
} |
|
if (existingProps[ClassPropertyType.prototype].has('constructor')) { |
|
throwError(classElement.key, Messages.IllegalDuplicateClassProperty); |
|
} |
|
} |
|
existingProps[propType].set(propName, true); |
|
} |
|
} |
|
} |
|
} |
|
|
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createClassBody(classElements)); |
|
} |
|
|
|
function parseClassImplements() { |
|
var id, implemented = [], marker, typeParameters; |
|
if (strict) { |
|
expectKeyword('implements'); |
|
} else { |
|
expectContextualKeyword('implements'); |
|
} |
|
while (index < length) { |
|
marker = markerCreate(); |
|
id = parseVariableIdentifier(); |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterInstantiation(); |
|
} else { |
|
typeParameters = null; |
|
} |
|
implemented.push(markerApply(marker, delegate.createClassImplements( |
|
id, |
|
typeParameters |
|
))); |
|
if (!match(',')) { |
|
break; |
|
} |
|
expect(','); |
|
} |
|
return implemented; |
|
} |
|
|
|
function parseClassExpression() { |
|
var id, implemented, previousYieldAllowed, superClass = null, |
|
superTypeParameters, marker = markerCreate(), typeParameters, |
|
matchImplements; |
|
|
|
expectKeyword('class'); |
|
|
|
matchImplements = |
|
strict |
|
? matchKeyword('implements') |
|
: matchContextualKeyword('implements'); |
|
|
|
if (!matchKeyword('extends') && !matchImplements && !match('{')) { |
|
id = parseVariableIdentifier(); |
|
} |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
if (matchKeyword('extends')) { |
|
expectKeyword('extends'); |
|
previousYieldAllowed = state.yieldAllowed; |
|
state.yieldAllowed = false; |
|
superClass = parseLeftHandSideExpressionAllowCall(); |
|
if (match('<')) { |
|
superTypeParameters = parseTypeParameterInstantiation(); |
|
} |
|
state.yieldAllowed = previousYieldAllowed; |
|
} |
|
|
|
if (strict ? matchKeyword('implements') : matchContextualKeyword('implements')) { |
|
implemented = parseClassImplements(); |
|
} |
|
|
|
return markerApply(marker, delegate.createClassExpression( |
|
id, |
|
superClass, |
|
parseClassBody(), |
|
typeParameters, |
|
superTypeParameters, |
|
implemented |
|
)); |
|
} |
|
|
|
function parseClassDeclaration() { |
|
var id, implemented, previousYieldAllowed, superClass = null, |
|
superTypeParameters, marker = markerCreate(), typeParameters; |
|
|
|
expectKeyword('class'); |
|
|
|
id = parseVariableIdentifier(); |
|
|
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
if (matchKeyword('extends')) { |
|
expectKeyword('extends'); |
|
previousYieldAllowed = state.yieldAllowed; |
|
state.yieldAllowed = false; |
|
superClass = parseLeftHandSideExpressionAllowCall(); |
|
if (match('<')) { |
|
superTypeParameters = parseTypeParameterInstantiation(); |
|
} |
|
state.yieldAllowed = previousYieldAllowed; |
|
} |
|
|
|
if (strict ? matchKeyword('implements') : matchContextualKeyword('implements')) { |
|
implemented = parseClassImplements(); |
|
} |
|
|
|
return markerApply(marker, delegate.createClassDeclaration( |
|
id, |
|
superClass, |
|
parseClassBody(), |
|
typeParameters, |
|
superTypeParameters, |
|
implemented |
|
)); |
|
} |
|
|
|
// 15 Program |
|
|
|
function parseSourceElement() { |
|
var token; |
|
if (lookahead.type === Token.Keyword) { |
|
switch (lookahead.value) { |
|
case 'const': |
|
case 'let': |
|
return parseConstLetDeclaration(lookahead.value); |
|
case 'function': |
|
return parseFunctionDeclaration(); |
|
case 'export': |
|
throwErrorTolerant({}, Messages.IllegalExportDeclaration); |
|
return parseExportDeclaration(); |
|
case 'import': |
|
throwErrorTolerant({}, Messages.IllegalImportDeclaration); |
|
return parseImportDeclaration(); |
|
case 'interface': |
|
if (lookahead2().type === Token.Identifier) { |
|
return parseInterface(); |
|
} |
|
return parseStatement(); |
|
default: |
|
return parseStatement(); |
|
} |
|
} |
|
|
|
if (matchContextualKeyword('type') |
|
&& lookahead2().type === Token.Identifier) { |
|
return parseTypeAlias(); |
|
} |
|
|
|
if (matchContextualKeyword('interface') |
|
&& lookahead2().type === Token.Identifier) { |
|
return parseInterface(); |
|
} |
|
|
|
if (matchContextualKeyword('declare')) { |
|
token = lookahead2(); |
|
if (token.type === Token.Keyword) { |
|
switch (token.value) { |
|
case 'class': |
|
return parseDeclareClass(); |
|
case 'function': |
|
return parseDeclareFunction(); |
|
case 'var': |
|
return parseDeclareVariable(); |
|
} |
|
} else if (token.type === Token.Identifier |
|
&& token.value === 'module') { |
|
return parseDeclareModule(); |
|
} |
|
} |
|
|
|
if (lookahead.type !== Token.EOF) { |
|
return parseStatement(); |
|
} |
|
} |
|
|
|
function parseProgramElement() { |
|
var isModule = extra.sourceType === 'module' || extra.sourceType === 'nonStrictModule'; |
|
|
|
if (isModule && lookahead.type === Token.Keyword) { |
|
switch (lookahead.value) { |
|
case 'export': |
|
return parseExportDeclaration(); |
|
case 'import': |
|
return parseImportDeclaration(); |
|
} |
|
} |
|
|
|
return parseSourceElement(); |
|
} |
|
|
|
function parseProgramElements() { |
|
var sourceElement, sourceElements = [], token, directive, firstRestricted; |
|
|
|
while (index < length) { |
|
token = lookahead; |
|
if (token.type !== Token.StringLiteral) { |
|
break; |
|
} |
|
|
|
sourceElement = parseProgramElement(); |
|
sourceElements.push(sourceElement); |
|
if (sourceElement.expression.type !== Syntax.Literal) { |
|
// this is not directive |
|
break; |
|
} |
|
directive = source.slice(token.range[0] + 1, token.range[1] - 1); |
|
if (directive === 'use strict') { |
|
strict = true; |
|
if (firstRestricted) { |
|
throwErrorTolerant(firstRestricted, Messages.StrictOctalLiteral); |
|
} |
|
} else { |
|
if (!firstRestricted && token.octal) { |
|
firstRestricted = token; |
|
} |
|
} |
|
} |
|
|
|
while (index < length) { |
|
sourceElement = parseProgramElement(); |
|
if (typeof sourceElement === 'undefined') { |
|
break; |
|
} |
|
sourceElements.push(sourceElement); |
|
} |
|
return sourceElements; |
|
} |
|
|
|
function parseProgram() { |
|
var body, marker = markerCreate(); |
|
strict = extra.sourceType === 'module'; |
|
peek(); |
|
body = parseProgramElements(); |
|
return markerApply(marker, delegate.createProgram(body)); |
|
} |
|
|
|
// 16 JSX |
|
|
|
XHTMLEntities = { |
|
quot: '\u0022', |
|
amp: '&', |
|
apos: '\u0027', |
|
lt: '<', |
|
gt: '>', |
|
nbsp: '\u00A0', |
|
iexcl: '\u00A1', |
|
cent: '\u00A2', |
|
pound: '\u00A3', |
|
curren: '\u00A4', |
|
yen: '\u00A5', |
|
brvbar: '\u00A6', |
|
sect: '\u00A7', |
|
uml: '\u00A8', |
|
copy: '\u00A9', |
|
ordf: '\u00AA', |
|
laquo: '\u00AB', |
|
not: '\u00AC', |
|
shy: '\u00AD', |
|
reg: '\u00AE', |
|
macr: '\u00AF', |
|
deg: '\u00B0', |
|
plusmn: '\u00B1', |
|
sup2: '\u00B2', |
|
sup3: '\u00B3', |
|
acute: '\u00B4', |
|
micro: '\u00B5', |
|
para: '\u00B6', |
|
middot: '\u00B7', |
|
cedil: '\u00B8', |
|
sup1: '\u00B9', |
|
ordm: '\u00BA', |
|
raquo: '\u00BB', |
|
frac14: '\u00BC', |
|
frac12: '\u00BD', |
|
frac34: '\u00BE', |
|
iquest: '\u00BF', |
|
Agrave: '\u00C0', |
|
Aacute: '\u00C1', |
|
Acirc: '\u00C2', |
|
Atilde: '\u00C3', |
|
Auml: '\u00C4', |
|
Aring: '\u00C5', |
|
AElig: '\u00C6', |
|
Ccedil: '\u00C7', |
|
Egrave: '\u00C8', |
|
Eacute: '\u00C9', |
|
Ecirc: '\u00CA', |
|
Euml: '\u00CB', |
|
Igrave: '\u00CC', |
|
Iacute: '\u00CD', |
|
Icirc: '\u00CE', |
|
Iuml: '\u00CF', |
|
ETH: '\u00D0', |
|
Ntilde: '\u00D1', |
|
Ograve: '\u00D2', |
|
Oacute: '\u00D3', |
|
Ocirc: '\u00D4', |
|
Otilde: '\u00D5', |
|
Ouml: '\u00D6', |
|
times: '\u00D7', |
|
Oslash: '\u00D8', |
|
Ugrave: '\u00D9', |
|
Uacute: '\u00DA', |
|
Ucirc: '\u00DB', |
|
Uuml: '\u00DC', |
|
Yacute: '\u00DD', |
|
THORN: '\u00DE', |
|
szlig: '\u00DF', |
|
agrave: '\u00E0', |
|
aacute: '\u00E1', |
|
acirc: '\u00E2', |
|
atilde: '\u00E3', |
|
auml: '\u00E4', |
|
aring: '\u00E5', |
|
aelig: '\u00E6', |
|
ccedil: '\u00E7', |
|
egrave: '\u00E8', |
|
eacute: '\u00E9', |
|
ecirc: '\u00EA', |
|
euml: '\u00EB', |
|
igrave: '\u00EC', |
|
iacute: '\u00ED', |
|
icirc: '\u00EE', |
|
iuml: '\u00EF', |
|
eth: '\u00F0', |
|
ntilde: '\u00F1', |
|
ograve: '\u00F2', |
|
oacute: '\u00F3', |
|
ocirc: '\u00F4', |
|
otilde: '\u00F5', |
|
ouml: '\u00F6', |
|
divide: '\u00F7', |
|
oslash: '\u00F8', |
|
ugrave: '\u00F9', |
|
uacute: '\u00FA', |
|
ucirc: '\u00FB', |
|
uuml: '\u00FC', |
|
yacute: '\u00FD', |
|
thorn: '\u00FE', |
|
yuml: '\u00FF', |
|
OElig: '\u0152', |
|
oelig: '\u0153', |
|
Scaron: '\u0160', |
|
scaron: '\u0161', |
|
Yuml: '\u0178', |
|
fnof: '\u0192', |
|
circ: '\u02C6', |
|
tilde: '\u02DC', |
|
Alpha: '\u0391', |
|
Beta: '\u0392', |
|
Gamma: '\u0393', |
|
Delta: '\u0394', |
|
Epsilon: '\u0395', |
|
Zeta: '\u0396', |
|
Eta: '\u0397', |
|
Theta: '\u0398', |
|
Iota: '\u0399', |
|
Kappa: '\u039A', |
|
Lambda: '\u039B', |
|
Mu: '\u039C', |
|
Nu: '\u039D', |
|
Xi: '\u039E', |
|
Omicron: '\u039F', |
|
Pi: '\u03A0', |
|
Rho: '\u03A1', |
|
Sigma: '\u03A3', |
|
Tau: '\u03A4', |
|
Upsilon: '\u03A5', |
|
Phi: '\u03A6', |
|
Chi: '\u03A7', |
|
Psi: '\u03A8', |
|
Omega: '\u03A9', |
|
alpha: '\u03B1', |
|
beta: '\u03B2', |
|
gamma: '\u03B3', |
|
delta: '\u03B4', |
|
epsilon: '\u03B5', |
|
zeta: '\u03B6', |
|
eta: '\u03B7', |
|
theta: '\u03B8', |
|
iota: '\u03B9', |
|
kappa: '\u03BA', |
|
lambda: '\u03BB', |
|
mu: '\u03BC', |
|
nu: '\u03BD', |
|
xi: '\u03BE', |
|
omicron: '\u03BF', |
|
pi: '\u03C0', |
|
rho: '\u03C1', |
|
sigmaf: '\u03C2', |
|
sigma: '\u03C3', |
|
tau: '\u03C4', |
|
upsilon: '\u03C5', |
|
phi: '\u03C6', |
|
chi: '\u03C7', |
|
psi: '\u03C8', |
|
omega: '\u03C9', |
|
thetasym: '\u03D1', |
|
upsih: '\u03D2', |
|
piv: '\u03D6', |
|
ensp: '\u2002', |
|
emsp: '\u2003', |
|
thinsp: '\u2009', |
|
zwnj: '\u200C', |
|
zwj: '\u200D', |
|
lrm: '\u200E', |
|
rlm: '\u200F', |
|
ndash: '\u2013', |
|
mdash: '\u2014', |
|
lsquo: '\u2018', |
|
rsquo: '\u2019', |
|
sbquo: '\u201A', |
|
ldquo: '\u201C', |
|
rdquo: '\u201D', |
|
bdquo: '\u201E', |
|
dagger: '\u2020', |
|
Dagger: '\u2021', |
|
bull: '\u2022', |
|
hellip: '\u2026', |
|
permil: '\u2030', |
|
prime: '\u2032', |
|
Prime: '\u2033', |
|
lsaquo: '\u2039', |
|
rsaquo: '\u203A', |
|
oline: '\u203E', |
|
frasl: '\u2044', |
|
euro: '\u20AC', |
|
image: '\u2111', |
|
weierp: '\u2118', |
|
real: '\u211C', |
|
trade: '\u2122', |
|
alefsym: '\u2135', |
|
larr: '\u2190', |
|
uarr: '\u2191', |
|
rarr: '\u2192', |
|
darr: '\u2193', |
|
harr: '\u2194', |
|
crarr: '\u21B5', |
|
lArr: '\u21D0', |
|
uArr: '\u21D1', |
|
rArr: '\u21D2', |
|
dArr: '\u21D3', |
|
hArr: '\u21D4', |
|
forall: '\u2200', |
|
part: '\u2202', |
|
exist: '\u2203', |
|
empty: '\u2205', |
|
nabla: '\u2207', |
|
isin: '\u2208', |
|
notin: '\u2209', |
|
ni: '\u220B', |
|
prod: '\u220F', |
|
sum: '\u2211', |
|
minus: '\u2212', |
|
lowast: '\u2217', |
|
radic: '\u221A', |
|
prop: '\u221D', |
|
infin: '\u221E', |
|
ang: '\u2220', |
|
and: '\u2227', |
|
or: '\u2228', |
|
cap: '\u2229', |
|
cup: '\u222A', |
|
'int': '\u222B', |
|
there4: '\u2234', |
|
sim: '\u223C', |
|
cong: '\u2245', |
|
asymp: '\u2248', |
|
ne: '\u2260', |
|
equiv: '\u2261', |
|
le: '\u2264', |
|
ge: '\u2265', |
|
sub: '\u2282', |
|
sup: '\u2283', |
|
nsub: '\u2284', |
|
sube: '\u2286', |
|
supe: '\u2287', |
|
oplus: '\u2295', |
|
otimes: '\u2297', |
|
perp: '\u22A5', |
|
sdot: '\u22C5', |
|
lceil: '\u2308', |
|
rceil: '\u2309', |
|
lfloor: '\u230A', |
|
rfloor: '\u230B', |
|
lang: '\u2329', |
|
rang: '\u232A', |
|
loz: '\u25CA', |
|
spades: '\u2660', |
|
clubs: '\u2663', |
|
hearts: '\u2665', |
|
diams: '\u2666' |
|
}; |
|
|
|
function getQualifiedJSXName(object) { |
|
if (object.type === Syntax.JSXIdentifier) { |
|
return object.name; |
|
} |
|
if (object.type === Syntax.JSXNamespacedName) { |
|
return object.namespace.name + ':' + object.name.name; |
|
} |
|
/* istanbul ignore else */ |
|
if (object.type === Syntax.JSXMemberExpression) { |
|
return ( |
|
getQualifiedJSXName(object.object) + '.' + |
|
getQualifiedJSXName(object.property) |
|
); |
|
} |
|
/* istanbul ignore next */ |
|
throwUnexpected(object); |
|
} |
|
|
|
function isJSXIdentifierStart(ch) { |
|
// exclude backslash (\) |
|
return (ch !== 92) && isIdentifierStart(ch); |
|
} |
|
|
|
function isJSXIdentifierPart(ch) { |
|
// exclude backslash (\) and add hyphen (-) |
|
return (ch !== 92) && (ch === 45 || isIdentifierPart(ch)); |
|
} |
|
|
|
function scanJSXIdentifier() { |
|
var ch, start, value = ''; |
|
|
|
start = index; |
|
while (index < length) { |
|
ch = source.charCodeAt(index); |
|
if (!isJSXIdentifierPart(ch)) { |
|
break; |
|
} |
|
value += source[index++]; |
|
} |
|
|
|
return { |
|
type: Token.JSXIdentifier, |
|
value: value, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanJSXEntity() { |
|
var ch, str = '', start = index, count = 0, code; |
|
ch = source[index]; |
|
assert(ch === '&', 'Entity must start with an ampersand'); |
|
index++; |
|
while (index < length && count++ < 10) { |
|
ch = source[index++]; |
|
if (ch === ';') { |
|
break; |
|
} |
|
str += ch; |
|
} |
|
|
|
// Well-formed entity (ending was found). |
|
if (ch === ';') { |
|
// Numeric entity. |
|
if (str[0] === '#') { |
|
if (str[1] === 'x') { |
|
code = +('0' + str.substr(1)); |
|
} else { |
|
// Removing leading zeros in order to avoid treating as octal in old browsers. |
|
code = +str.substr(1).replace(Regex.LeadingZeros, ''); |
|
} |
|
|
|
if (!isNaN(code)) { |
|
return String.fromCharCode(code); |
|
} |
|
/* istanbul ignore else */ |
|
} else if (XHTMLEntities[str]) { |
|
return XHTMLEntities[str]; |
|
} |
|
} |
|
|
|
// Treat non-entity sequences as regular text. |
|
index = start + 1; |
|
return '&'; |
|
} |
|
|
|
function scanJSXText(stopChars) { |
|
var ch, str = '', start; |
|
start = index; |
|
while (index < length) { |
|
ch = source[index]; |
|
if (stopChars.indexOf(ch) !== -1) { |
|
break; |
|
} |
|
if (ch === '&') { |
|
str += scanJSXEntity(); |
|
} else { |
|
index++; |
|
if (ch === '\r' && source[index] === '\n') { |
|
str += ch; |
|
ch = source[index]; |
|
index++; |
|
} |
|
if (isLineTerminator(ch.charCodeAt(0))) { |
|
++lineNumber; |
|
lineStart = index; |
|
} |
|
str += ch; |
|
} |
|
} |
|
return { |
|
type: Token.JSXText, |
|
value: str, |
|
lineNumber: lineNumber, |
|
lineStart: lineStart, |
|
range: [start, index] |
|
}; |
|
} |
|
|
|
function scanJSXStringLiteral() { |
|
var innerToken, quote, start; |
|
|
|
quote = source[index]; |
|
assert((quote === '\'' || quote === '"'), |
|
'String literal must starts with a quote'); |
|
|
|
start = index; |
|
++index; |
|
|
|
innerToken = scanJSXText([quote]); |
|
|
|
if (quote !== source[index]) { |
|
throwError({}, Messages.UnexpectedToken, 'ILLEGAL'); |
|
} |
|
|
|
++index; |
|
|
|
innerToken.range = [start, index]; |
|
|
|
return innerToken; |
|
} |
|
|
|
/** |
|
* Between JSX opening and closing tags (e.g. <foo>HERE</foo>), anything that |
|
* is not another JSX tag and is not an expression wrapped by {} is text. |
|
*/ |
|
function advanceJSXChild() { |
|
var ch = source.charCodeAt(index); |
|
|
|
// '<' 60, '>' 62, '{' 123, '}' 125 |
|
if (ch !== 60 && ch !== 62 && ch !== 123 && ch !== 125) { |
|
return scanJSXText(['<', '>', '{', '}']); |
|
} |
|
|
|
return scanPunctuator(); |
|
} |
|
|
|
function parseJSXIdentifier() { |
|
var token, marker = markerCreate(); |
|
|
|
if (lookahead.type !== Token.JSXIdentifier) { |
|
throwUnexpected(lookahead); |
|
} |
|
|
|
token = lex(); |
|
return markerApply(marker, delegate.createJSXIdentifier(token.value)); |
|
} |
|
|
|
function parseJSXNamespacedName() { |
|
var namespace, name, marker = markerCreate(); |
|
|
|
namespace = parseJSXIdentifier(); |
|
expect(':'); |
|
name = parseJSXIdentifier(); |
|
|
|
return markerApply(marker, delegate.createJSXNamespacedName(namespace, name)); |
|
} |
|
|
|
function parseJSXMemberExpression() { |
|
var marker = markerCreate(), |
|
expr = parseJSXIdentifier(); |
|
|
|
while (match('.')) { |
|
lex(); |
|
expr = markerApply(marker, delegate.createJSXMemberExpression(expr, parseJSXIdentifier())); |
|
} |
|
|
|
return expr; |
|
} |
|
|
|
function parseJSXElementName() { |
|
if (lookahead2().value === ':') { |
|
return parseJSXNamespacedName(); |
|
} |
|
if (lookahead2().value === '.') { |
|
return parseJSXMemberExpression(); |
|
} |
|
|
|
return parseJSXIdentifier(); |
|
} |
|
|
|
function parseJSXAttributeName() { |
|
if (lookahead2().value === ':') { |
|
return parseJSXNamespacedName(); |
|
} |
|
|
|
return parseJSXIdentifier(); |
|
} |
|
|
|
function parseJSXAttributeValue() { |
|
var value, marker; |
|
if (match('{')) { |
|
value = parseJSXExpressionContainer(); |
|
if (value.expression.type === Syntax.JSXEmptyExpression) { |
|
throwError( |
|
value, |
|
'JSX attributes must only be assigned a non-empty ' + |
|
'expression' |
|
); |
|
} |
|
} else if (match('<')) { |
|
value = parseJSXElement(); |
|
} else if (lookahead.type === Token.JSXText) { |
|
marker = markerCreate(); |
|
value = markerApply(marker, delegate.createLiteral(lex())); |
|
} else { |
|
throwError({}, Messages.InvalidJSXAttributeValue); |
|
} |
|
return value; |
|
} |
|
|
|
function parseJSXEmptyExpression() { |
|
var marker = markerCreatePreserveWhitespace(); |
|
while (source.charAt(index) !== '}') { |
|
index++; |
|
} |
|
return markerApply(marker, delegate.createJSXEmptyExpression()); |
|
} |
|
|
|
function parseJSXExpressionContainer() { |
|
var expression, origInJSXChild, origInJSXTag, marker = markerCreate(); |
|
|
|
origInJSXChild = state.inJSXChild; |
|
origInJSXTag = state.inJSXTag; |
|
state.inJSXChild = false; |
|
state.inJSXTag = false; |
|
|
|
expect('{'); |
|
|
|
if (match('}')) { |
|
expression = parseJSXEmptyExpression(); |
|
} else { |
|
expression = parseExpression(); |
|
} |
|
|
|
state.inJSXChild = origInJSXChild; |
|
state.inJSXTag = origInJSXTag; |
|
|
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createJSXExpressionContainer(expression)); |
|
} |
|
|
|
function parseJSXSpreadAttribute() { |
|
var expression, origInJSXChild, origInJSXTag, marker = markerCreate(); |
|
|
|
origInJSXChild = state.inJSXChild; |
|
origInJSXTag = state.inJSXTag; |
|
state.inJSXChild = false; |
|
state.inJSXTag = false; |
|
|
|
expect('{'); |
|
expect('...'); |
|
|
|
expression = parseAssignmentExpression(); |
|
|
|
state.inJSXChild = origInJSXChild; |
|
state.inJSXTag = origInJSXTag; |
|
|
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createJSXSpreadAttribute(expression)); |
|
} |
|
|
|
function parseJSXAttribute() { |
|
var name, marker; |
|
|
|
if (match('{')) { |
|
return parseJSXSpreadAttribute(); |
|
} |
|
|
|
marker = markerCreate(); |
|
|
|
name = parseJSXAttributeName(); |
|
|
|
// HTML empty attribute |
|
if (match('=')) { |
|
lex(); |
|
return markerApply(marker, delegate.createJSXAttribute(name, parseJSXAttributeValue())); |
|
} |
|
|
|
return markerApply(marker, delegate.createJSXAttribute(name)); |
|
} |
|
|
|
function parseJSXChild() { |
|
var token, marker; |
|
if (match('{')) { |
|
token = parseJSXExpressionContainer(); |
|
} else if (lookahead.type === Token.JSXText) { |
|
marker = markerCreatePreserveWhitespace(); |
|
token = markerApply(marker, delegate.createLiteral(lex())); |
|
} else if (match('<')) { |
|
token = parseJSXElement(); |
|
} else { |
|
throwUnexpected(lookahead); |
|
} |
|
return token; |
|
} |
|
|
|
function parseJSXClosingElement() { |
|
var name, origInJSXChild, origInJSXTag, marker = markerCreate(); |
|
origInJSXChild = state.inJSXChild; |
|
origInJSXTag = state.inJSXTag; |
|
state.inJSXChild = false; |
|
state.inJSXTag = true; |
|
expect('<'); |
|
expect('/'); |
|
name = parseJSXElementName(); |
|
// Because advance() (called by lex() called by expect()) expects there |
|
// to be a valid token after >, it needs to know whether to look for a |
|
// standard JS token or an JSX text node |
|
state.inJSXChild = origInJSXChild; |
|
state.inJSXTag = origInJSXTag; |
|
expect('>'); |
|
return markerApply(marker, delegate.createJSXClosingElement(name)); |
|
} |
|
|
|
function parseJSXOpeningElement() { |
|
var name, attributes = [], selfClosing = false, origInJSXChild, origInJSXTag, marker = markerCreate(); |
|
|
|
origInJSXChild = state.inJSXChild; |
|
origInJSXTag = state.inJSXTag; |
|
state.inJSXChild = false; |
|
state.inJSXTag = true; |
|
|
|
expect('<'); |
|
|
|
name = parseJSXElementName(); |
|
|
|
while (index < length && |
|
lookahead.value !== '/' && |
|
lookahead.value !== '>') { |
|
attributes.push(parseJSXAttribute()); |
|
} |
|
|
|
state.inJSXTag = origInJSXTag; |
|
|
|
if (lookahead.value === '/') { |
|
expect('/'); |
|
// Because advance() (called by lex() called by expect()) expects |
|
// there to be a valid token after >, it needs to know whether to |
|
// look for a standard JS token or an JSX text node |
|
state.inJSXChild = origInJSXChild; |
|
expect('>'); |
|
selfClosing = true; |
|
} else { |
|
state.inJSXChild = true; |
|
expect('>'); |
|
} |
|
return markerApply(marker, delegate.createJSXOpeningElement(name, attributes, selfClosing)); |
|
} |
|
|
|
function parseJSXElement() { |
|
var openingElement, closingElement = null, children = [], origInJSXChild, origInJSXTag, marker = markerCreate(); |
|
|
|
origInJSXChild = state.inJSXChild; |
|
origInJSXTag = state.inJSXTag; |
|
openingElement = parseJSXOpeningElement(); |
|
|
|
if (!openingElement.selfClosing) { |
|
while (index < length) { |
|
state.inJSXChild = false; // Call lookahead2() with inJSXChild = false because </ should not be considered in the child |
|
if (lookahead.value === '<' && lookahead2().value === '/') { |
|
break; |
|
} |
|
state.inJSXChild = true; |
|
children.push(parseJSXChild()); |
|
} |
|
state.inJSXChild = origInJSXChild; |
|
state.inJSXTag = origInJSXTag; |
|
closingElement = parseJSXClosingElement(); |
|
if (getQualifiedJSXName(closingElement.name) !== getQualifiedJSXName(openingElement.name)) { |
|
throwError({}, Messages.ExpectedJSXClosingTag, getQualifiedJSXName(openingElement.name)); |
|
} |
|
} |
|
|
|
// When (erroneously) writing two adjacent tags like |
|
// |
|
// var x = <div>one</div><div>two</div>; |
|
// |
|
// the default error message is a bit incomprehensible. Since it's |
|
// rarely (never?) useful to write a less-than sign after an JSX |
|
// element, we disallow it here in the parser in order to provide a |
|
// better error message. (In the rare case that the less-than operator |
|
// was intended, the left tag can be wrapped in parentheses.) |
|
if (!origInJSXChild && match('<')) { |
|
throwError(lookahead, Messages.AdjacentJSXElements); |
|
} |
|
|
|
return markerApply(marker, delegate.createJSXElement(openingElement, closingElement, children)); |
|
} |
|
|
|
function parseTypeAlias() { |
|
var id, marker = markerCreate(), typeParameters = null, right; |
|
expectContextualKeyword('type'); |
|
id = parseVariableIdentifier(); |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
expect('='); |
|
right = parseType(); |
|
consumeSemicolon(); |
|
return markerApply(marker, delegate.createTypeAlias(id, typeParameters, right)); |
|
} |
|
|
|
function parseInterfaceExtends() { |
|
var marker = markerCreate(), id, typeParameters = null; |
|
|
|
id = parseVariableIdentifier(); |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterInstantiation(); |
|
} |
|
|
|
return markerApply(marker, delegate.createInterfaceExtends( |
|
id, |
|
typeParameters |
|
)); |
|
} |
|
|
|
function parseInterfaceish(marker, allowStatic) { |
|
var body, bodyMarker, extended = [], id, |
|
typeParameters = null; |
|
|
|
id = parseVariableIdentifier(); |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
|
|
if (matchKeyword('extends')) { |
|
expectKeyword('extends'); |
|
|
|
while (index < length) { |
|
extended.push(parseInterfaceExtends()); |
|
if (!match(',')) { |
|
break; |
|
} |
|
expect(','); |
|
} |
|
} |
|
|
|
bodyMarker = markerCreate(); |
|
body = markerApply(bodyMarker, parseObjectType(allowStatic)); |
|
|
|
return markerApply(marker, delegate.createInterface( |
|
id, |
|
typeParameters, |
|
body, |
|
extended |
|
)); |
|
} |
|
|
|
function parseInterface() { |
|
var marker = markerCreate(); |
|
|
|
if (strict) { |
|
expectKeyword('interface'); |
|
} else { |
|
expectContextualKeyword('interface'); |
|
} |
|
|
|
return parseInterfaceish(marker, /* allowStatic */false); |
|
} |
|
|
|
function parseDeclareClass() { |
|
var marker = markerCreate(), ret; |
|
expectContextualKeyword('declare'); |
|
expectKeyword('class'); |
|
|
|
ret = parseInterfaceish(marker, /* allowStatic */true); |
|
ret.type = Syntax.DeclareClass; |
|
return ret; |
|
} |
|
|
|
function parseDeclareFunction() { |
|
var id, idMarker, |
|
marker = markerCreate(), params, returnType, rest, tmp, |
|
typeParameters = null, value, valueMarker; |
|
|
|
expectContextualKeyword('declare'); |
|
expectKeyword('function'); |
|
idMarker = markerCreate(); |
|
id = parseVariableIdentifier(); |
|
|
|
valueMarker = markerCreate(); |
|
if (match('<')) { |
|
typeParameters = parseTypeParameterDeclaration(); |
|
} |
|
expect('('); |
|
tmp = parseFunctionTypeParams(); |
|
params = tmp.params; |
|
rest = tmp.rest; |
|
expect(')'); |
|
|
|
expect(':'); |
|
returnType = parseType(); |
|
|
|
value = markerApply(valueMarker, delegate.createFunctionTypeAnnotation( |
|
params, |
|
returnType, |
|
rest, |
|
typeParameters |
|
)); |
|
|
|
id.typeAnnotation = markerApply(valueMarker, delegate.createTypeAnnotation( |
|
value |
|
)); |
|
markerApply(idMarker, id); |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createDeclareFunction( |
|
id |
|
)); |
|
} |
|
|
|
function parseDeclareVariable() { |
|
var id, marker = markerCreate(); |
|
expectContextualKeyword('declare'); |
|
expectKeyword('var'); |
|
id = parseTypeAnnotatableIdentifier(); |
|
|
|
consumeSemicolon(); |
|
|
|
return markerApply(marker, delegate.createDeclareVariable( |
|
id |
|
)); |
|
} |
|
|
|
function parseDeclareModule() { |
|
var body = [], bodyMarker, id, idMarker, marker = markerCreate(), token; |
|
expectContextualKeyword('declare'); |
|
expectContextualKeyword('module'); |
|
|
|
if (lookahead.type === Token.StringLiteral) { |
|
if (strict && lookahead.octal) { |
|
throwErrorTolerant(lookahead, Messages.StrictOctalLiteral); |
|
} |
|
idMarker = markerCreate(); |
|
id = markerApply(idMarker, delegate.createLiteral(lex())); |
|
} else { |
|
id = parseVariableIdentifier(); |
|
} |
|
|
|
bodyMarker = markerCreate(); |
|
expect('{'); |
|
while (index < length && !match('}')) { |
|
token = lookahead2(); |
|
switch (token.value) { |
|
case 'class': |
|
body.push(parseDeclareClass()); |
|
break; |
|
case 'function': |
|
body.push(parseDeclareFunction()); |
|
break; |
|
case 'var': |
|
body.push(parseDeclareVariable()); |
|
break; |
|
default: |
|
throwUnexpected(lookahead); |
|
} |
|
} |
|
expect('}'); |
|
|
|
return markerApply(marker, delegate.createDeclareModule( |
|
id, |
|
markerApply(bodyMarker, delegate.createBlockStatement(body)) |
|
)); |
|
} |
|
|
|
function collectToken() { |
|
var loc, token, range, value, entry; |
|
|
|
/* istanbul ignore else */ |
|
if (!state.inJSXChild) { |
|
skipComment(); |
|
} |
|
|
|
loc = { |
|
start: { |
|
line: lineNumber, |
|
column: index - lineStart |
|
} |
|
}; |
|
|
|
token = extra.advance(); |
|
loc.end = { |
|
line: lineNumber, |
|
column: index - lineStart |
|
}; |
|
|
|
if (token.type !== Token.EOF) { |
|
range = [token.range[0], token.range[1]]; |
|
value = source.slice(token.range[0], token.range[1]); |
|
entry = { |
|
type: TokenName[token.type], |
|
value: value, |
|
range: range, |
|
loc: loc |
|
}; |
|
if (token.regex) { |
|
entry.regex = { |
|
pattern: token.regex.pattern, |
|
flags: token.regex.flags |
|
}; |
|
} |
|
extra.tokens.push(entry); |
|
} |
|
|
|
return token; |
|
} |
|
|
|
function collectRegex() { |
|
var pos, loc, regex, token; |
|
|
|
skipComment(); |
|
|
|
pos = index; |
|
loc = { |
|
start: { |
|
line: lineNumber, |
|
column: index - lineStart |
|
} |
|
}; |
|
|
|
regex = extra.scanRegExp(); |
|
loc.end = { |
|
line: lineNumber, |
|
column: index - lineStart |
|
}; |
|
|
|
if (!extra.tokenize) { |
|
/* istanbul ignore next */ |
|
// Pop the previous token, which is likely '/' or '/=' |
|
if (extra.tokens.length > 0) { |
|
token = extra.tokens[extra.tokens.length - 1]; |
|
if (token.range[0] === pos && token.type === 'Punctuator') { |
|
if (token.value === '/' || token.value === '/=') { |
|
extra.tokens.pop(); |
|
} |
|
} |
|
} |
|
|
|
extra.tokens.push({ |
|
type: 'RegularExpression', |
|
value: regex.literal, |
|
regex: regex.regex, |
|
range: [pos, index], |
|
loc: loc |
|
}); |
|
} |
|
|
|
return regex; |
|
} |
|
|
|
function filterTokenLocation() { |
|
var i, entry, token, tokens = []; |
|
|
|
for (i = 0; i < extra.tokens.length; ++i) { |
|
entry = extra.tokens[i]; |
|
token = { |
|
type: entry.type, |
|
value: entry.value |
|
}; |
|
if (entry.regex) { |
|
token.regex = { |
|
pattern: entry.regex.pattern, |
|
flags: entry.regex.flags |
|
}; |
|
} |
|
if (extra.range) { |
|
token.range = entry.range; |
|
} |
|
if (extra.loc) { |
|
token.loc = entry.loc; |
|
} |
|
tokens.push(token); |
|
} |
|
|
|
extra.tokens = tokens; |
|
} |
|
|
|
function patch() { |
|
if (typeof extra.tokens !== 'undefined') { |
|
extra.advance = advance; |
|
extra.scanRegExp = scanRegExp; |
|
|
|
advance = collectToken; |
|
scanRegExp = collectRegex; |
|
} |
|
} |
|
|
|
function unpatch() { |
|
if (typeof extra.scanRegExp === 'function') { |
|
advance = extra.advance; |
|
scanRegExp = extra.scanRegExp; |
|
} |
|
} |
|
|
|
// This is used to modify the delegate. |
|
|
|
function extend(object, properties) { |
|
var entry, result = {}; |
|
|
|
for (entry in object) { |
|
/* istanbul ignore else */ |
|
if (object.hasOwnProperty(entry)) { |
|
result[entry] = object[entry]; |
|
} |
|
} |
|
|
|
for (entry in properties) { |
|
/* istanbul ignore else */ |
|
if (properties.hasOwnProperty(entry)) { |
|
result[entry] = properties[entry]; |
|
} |
|
} |
|
|
|
return result; |
|
} |
|
|
|
function tokenize(code, options) { |
|
var toString, |
|
token, |
|
tokens; |
|
|
|
toString = String; |
|
if (typeof code !== 'string' && !(code instanceof String)) { |
|
code = toString(code); |
|
} |
|
|
|
delegate = SyntaxTreeDelegate; |
|
source = code; |
|
index = 0; |
|
lineNumber = (source.length > 0) ? 1 : 0; |
|
lineStart = 0; |
|
length = source.length; |
|
lookahead = null; |
|
state = { |
|
allowKeyword: true, |
|
allowIn: true, |
|
labelSet: new StringMap(), |
|
inFunctionBody: false, |
|
inIteration: false, |
|
inSwitch: false, |
|
lastCommentStart: -1 |
|
}; |
|
|
|
extra = {}; |
|
|
|
// Options matching. |
|
options = options || {}; |
|
|
|
// Of course we collect tokens here. |
|
options.tokens = true; |
|
extra.tokens = []; |
|
extra.tokenize = true; |
|
// The following two fields are necessary to compute the Regex tokens. |
|
extra.openParenToken = -1; |
|
extra.openCurlyToken = -1; |
|
|
|
extra.range = (typeof options.range === 'boolean') && options.range; |
|
extra.loc = (typeof options.loc === 'boolean') && options.loc; |
|
|
|
if (typeof options.comment === 'boolean' && options.comment) { |
|
extra.comments = []; |
|
} |
|
if (typeof options.tolerant === 'boolean' && options.tolerant) { |
|
extra.errors = []; |
|
} |
|
|
|
patch(); |
|
|
|
try { |
|
peek(); |
|
if (lookahead.type === Token.EOF) { |
|
return extra.tokens; |
|
} |
|
|
|
token = lex(); |
|
while (lookahead.type !== Token.EOF) { |
|
try { |
|
token = lex(); |
|
} catch (lexError) { |
|
token = lookahead; |
|
if (extra.errors) { |
|
extra.errors.push(lexError); |
|
// We have to break on the first error |
|
// to avoid infinite loops. |
|
break; |
|
} else { |
|
throw lexError; |
|
} |
|
} |
|
} |
|
|
|
filterTokenLocation(); |
|
tokens = extra.tokens; |
|
if (typeof extra.comments !== 'undefined') { |
|
tokens.comments = extra.comments; |
|
} |
|
if (typeof extra.errors !== 'undefined') { |
|
tokens.errors = extra.errors; |
|
} |
|
} catch (e) { |
|
throw e; |
|
} finally { |
|
unpatch(); |
|
extra = {}; |
|
} |
|
return tokens; |
|
} |
|
|
|
function parse(code, options) { |
|
var program, toString; |
|
|
|
toString = String; |
|
if (typeof code !== 'string' && !(code instanceof String)) { |
|
code = toString(code); |
|
} |
|
|
|
delegate = SyntaxTreeDelegate; |
|
source = code; |
|
index = 0; |
|
lineNumber = (source.length > 0) ? 1 : 0; |
|
lineStart = 0; |
|
length = source.length; |
|
lookahead = null; |
|
state = { |
|
allowKeyword: false, |
|
allowIn: true, |
|
labelSet: new StringMap(), |
|
parenthesizedCount: 0, |
|
inFunctionBody: false, |
|
inIteration: false, |
|
inSwitch: false, |
|
inJSXChild: false, |
|
inJSXTag: false, |
|
inType: false, |
|
lastCommentStart: -1, |
|
yieldAllowed: false, |
|
awaitAllowed: false |
|
}; |
|
|
|
extra = {}; |
|
if (typeof options !== 'undefined') { |
|
extra.range = (typeof options.range === 'boolean') && options.range; |
|
extra.loc = (typeof options.loc === 'boolean') && options.loc; |
|
extra.attachComment = (typeof options.attachComment === 'boolean') && options.attachComment; |
|
|
|
if (extra.loc && options.source !== null && options.source !== undefined) { |
|
delegate = extend(delegate, { |
|
'postProcess': function (node) { |
|
node.loc.source = toString(options.source); |
|
return node; |
|
} |
|
}); |
|
} |
|
|
|
extra.sourceType = options.sourceType; |
|
if (typeof options.tokens === 'boolean' && options.tokens) { |
|
extra.tokens = []; |
|
} |
|
if (typeof options.comment === 'boolean' && options.comment) { |
|
extra.comments = []; |
|
} |
|
if (typeof options.tolerant === 'boolean' && options.tolerant) { |
|
extra.errors = []; |
|
} |
|
if (extra.attachComment) { |
|
extra.range = true; |
|
extra.comments = []; |
|
extra.bottomRightStack = []; |
|
extra.trailingComments = []; |
|
extra.leadingComments = []; |
|
} |
|
} |
|
|
|
patch(); |
|
try { |
|
program = parseProgram(); |
|
if (typeof extra.comments !== 'undefined') { |
|
program.comments = extra.comments; |
|
} |
|
if (typeof extra.tokens !== 'undefined') { |
|
filterTokenLocation(); |
|
program.tokens = extra.tokens; |
|
} |
|
if (typeof extra.errors !== 'undefined') { |
|
program.errors = extra.errors; |
|
} |
|
} catch (e) { |
|
throw e; |
|
} finally { |
|
unpatch(); |
|
extra = {}; |
|
} |
|
|
|
return program; |
|
} |
|
|
|
// Sync with *.json manifests. |
|
exports.version = '13001.1001.0-dev-harmony-fb'; |
|
|
|
exports.tokenize = tokenize; |
|
|
|
exports.parse = parse; |
|
|
|
// Deep copy. |
|
/* istanbul ignore next */ |
|
exports.Syntax = (function () { |
|
var name, types = {}; |
|
|
|
if (typeof Object.create === 'function') { |
|
types = Object.create(null); |
|
} |
|
|
|
for (name in Syntax) { |
|
if (Syntax.hasOwnProperty(name)) { |
|
types[name] = Syntax[name]; |
|
} |
|
} |
|
|
|
if (typeof Object.freeze === 'function') { |
|
Object.freeze(types); |
|
} |
|
|
|
return types; |
|
}()); |
|
|
|
})); |
|
/* vim: set sw=4 ts=4 et tw=80 : */ |
|
|
|
},{}],10:[function(_dereq_,module,exports){ |
|
var Base62 = (function (my) { |
|
my.chars = ["0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z", "A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z"] |
|
|
|
my.encode = function(i){ |
|
if (i === 0) {return '0'} |
|
var s = '' |
|
while (i > 0) { |
|
s = this.chars[i % 62] + s |
|
i = Math.floor(i/62) |
|
} |
|
return s |
|
}; |
|
my.decode = function(a,b,c,d){ |
|
for ( |
|
b = c = ( |
|
a === (/\W|_|^$/.test(a += "") || a) |
|
) - 1; |
|
d = a.charCodeAt(c++); |
|
) |
|
b = b * 62 + d - [, 48, 29, 87][d >> 5]; |
|
return b |
|
}; |
|
|
|
return my; |
|
}({})); |
|
|
|
module.exports = Base62 |
|
},{}],11:[function(_dereq_,module,exports){ |
|
/* |
|
* Copyright 2009-2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE.txt or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
exports.SourceMapGenerator = _dereq_('./source-map/source-map-generator').SourceMapGenerator; |
|
exports.SourceMapConsumer = _dereq_('./source-map/source-map-consumer').SourceMapConsumer; |
|
exports.SourceNode = _dereq_('./source-map/source-node').SourceNode; |
|
|
|
},{"./source-map/source-map-consumer":16,"./source-map/source-map-generator":17,"./source-map/source-node":18}],12:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
var util = _dereq_('./util'); |
|
|
|
/** |
|
* A data structure which is a combination of an array and a set. Adding a new |
|
* member is O(1), testing for membership is O(1), and finding the index of an |
|
* element is O(1). Removing elements from the set is not supported. Only |
|
* strings are supported for membership. |
|
*/ |
|
function ArraySet() { |
|
this._array = []; |
|
this._set = {}; |
|
} |
|
|
|
/** |
|
* Static method for creating ArraySet instances from an existing array. |
|
*/ |
|
ArraySet.fromArray = function ArraySet_fromArray(aArray, aAllowDuplicates) { |
|
var set = new ArraySet(); |
|
for (var i = 0, len = aArray.length; i < len; i++) { |
|
set.add(aArray[i], aAllowDuplicates); |
|
} |
|
return set; |
|
}; |
|
|
|
/** |
|
* Add the given string to this set. |
|
* |
|
* @param String aStr |
|
*/ |
|
ArraySet.prototype.add = function ArraySet_add(aStr, aAllowDuplicates) { |
|
var isDuplicate = this.has(aStr); |
|
var idx = this._array.length; |
|
if (!isDuplicate || aAllowDuplicates) { |
|
this._array.push(aStr); |
|
} |
|
if (!isDuplicate) { |
|
this._set[util.toSetString(aStr)] = idx; |
|
} |
|
}; |
|
|
|
/** |
|
* Is the given string a member of this set? |
|
* |
|
* @param String aStr |
|
*/ |
|
ArraySet.prototype.has = function ArraySet_has(aStr) { |
|
return Object.prototype.hasOwnProperty.call(this._set, |
|
util.toSetString(aStr)); |
|
}; |
|
|
|
/** |
|
* What is the index of the given string in the array? |
|
* |
|
* @param String aStr |
|
*/ |
|
ArraySet.prototype.indexOf = function ArraySet_indexOf(aStr) { |
|
if (this.has(aStr)) { |
|
return this._set[util.toSetString(aStr)]; |
|
} |
|
throw new Error('"' + aStr + '" is not in the set.'); |
|
}; |
|
|
|
/** |
|
* What is the element at the given index? |
|
* |
|
* @param Number aIdx |
|
*/ |
|
ArraySet.prototype.at = function ArraySet_at(aIdx) { |
|
if (aIdx >= 0 && aIdx < this._array.length) { |
|
return this._array[aIdx]; |
|
} |
|
throw new Error('No element indexed by ' + aIdx); |
|
}; |
|
|
|
/** |
|
* Returns the array representation of this set (which has the proper indices |
|
* indicated by indexOf). Note that this is a copy of the internal array used |
|
* for storing the members so that no one can mess with internal state. |
|
*/ |
|
ArraySet.prototype.toArray = function ArraySet_toArray() { |
|
return this._array.slice(); |
|
}; |
|
|
|
exports.ArraySet = ArraySet; |
|
|
|
}); |
|
|
|
},{"./util":19,"amdefine":20}],13:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
* |
|
* Based on the Base 64 VLQ implementation in Closure Compiler: |
|
* https://code.google.com/p/closure-compiler/source/browse/trunk/src/com/google/debugging/sourcemap/Base64VLQ.java |
|
* |
|
* Copyright 2011 The Closure Compiler Authors. All rights reserved. |
|
* Redistribution and use in source and binary forms, with or without |
|
* modification, are permitted provided that the following conditions are |
|
* met: |
|
* |
|
* * Redistributions of source code must retain the above copyright |
|
* notice, this list of conditions and the following disclaimer. |
|
* * Redistributions in binary form must reproduce the above |
|
* copyright notice, this list of conditions and the following |
|
* disclaimer in the documentation and/or other materials provided |
|
* with the distribution. |
|
* * Neither the name of Google Inc. nor the names of its |
|
* contributors may be used to endorse or promote products derived |
|
* from this software without specific prior written permission. |
|
* |
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
|
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
|
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
|
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
|
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
|
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
|
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
|
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
var base64 = _dereq_('./base64'); |
|
|
|
// A single base 64 digit can contain 6 bits of data. For the base 64 variable |
|
// length quantities we use in the source map spec, the first bit is the sign, |
|
// the next four bits are the actual value, and the 6th bit is the |
|
// continuation bit. The continuation bit tells us whether there are more |
|
// digits in this value following this digit. |
|
// |
|
// Continuation |
|
// | Sign |
|
// | | |
|
// V V |
|
// 101011 |
|
|
|
var VLQ_BASE_SHIFT = 5; |
|
|
|
// binary: 100000 |
|
var VLQ_BASE = 1 << VLQ_BASE_SHIFT; |
|
|
|
// binary: 011111 |
|
var VLQ_BASE_MASK = VLQ_BASE - 1; |
|
|
|
// binary: 100000 |
|
var VLQ_CONTINUATION_BIT = VLQ_BASE; |
|
|
|
/** |
|
* Converts from a two-complement value to a value where the sign bit is |
|
* is placed in the least significant bit. For example, as decimals: |
|
* 1 becomes 2 (10 binary), -1 becomes 3 (11 binary) |
|
* 2 becomes 4 (100 binary), -2 becomes 5 (101 binary) |
|
*/ |
|
function toVLQSigned(aValue) { |
|
return aValue < 0 |
|
? ((-aValue) << 1) + 1 |
|
: (aValue << 1) + 0; |
|
} |
|
|
|
/** |
|
* Converts to a two-complement value from a value where the sign bit is |
|
* is placed in the least significant bit. For example, as decimals: |
|
* 2 (10 binary) becomes 1, 3 (11 binary) becomes -1 |
|
* 4 (100 binary) becomes 2, 5 (101 binary) becomes -2 |
|
*/ |
|
function fromVLQSigned(aValue) { |
|
var isNegative = (aValue & 1) === 1; |
|
var shifted = aValue >> 1; |
|
return isNegative |
|
? -shifted |
|
: shifted; |
|
} |
|
|
|
/** |
|
* Returns the base 64 VLQ encoded value. |
|
*/ |
|
exports.encode = function base64VLQ_encode(aValue) { |
|
var encoded = ""; |
|
var digit; |
|
|
|
var vlq = toVLQSigned(aValue); |
|
|
|
do { |
|
digit = vlq & VLQ_BASE_MASK; |
|
vlq >>>= VLQ_BASE_SHIFT; |
|
if (vlq > 0) { |
|
// There are still more digits in this value, so we must make sure the |
|
// continuation bit is marked. |
|
digit |= VLQ_CONTINUATION_BIT; |
|
} |
|
encoded += base64.encode(digit); |
|
} while (vlq > 0); |
|
|
|
return encoded; |
|
}; |
|
|
|
/** |
|
* Decodes the next base 64 VLQ value from the given string and returns the |
|
* value and the rest of the string. |
|
*/ |
|
exports.decode = function base64VLQ_decode(aStr) { |
|
var i = 0; |
|
var strLen = aStr.length; |
|
var result = 0; |
|
var shift = 0; |
|
var continuation, digit; |
|
|
|
do { |
|
if (i >= strLen) { |
|
throw new Error("Expected more digits in base 64 VLQ value."); |
|
} |
|
digit = base64.decode(aStr.charAt(i++)); |
|
continuation = !!(digit & VLQ_CONTINUATION_BIT); |
|
digit &= VLQ_BASE_MASK; |
|
result = result + (digit << shift); |
|
shift += VLQ_BASE_SHIFT; |
|
} while (continuation); |
|
|
|
return { |
|
value: fromVLQSigned(result), |
|
rest: aStr.slice(i) |
|
}; |
|
}; |
|
|
|
}); |
|
|
|
},{"./base64":14,"amdefine":20}],14:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
var charToIntMap = {}; |
|
var intToCharMap = {}; |
|
|
|
'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/' |
|
.split('') |
|
.forEach(function (ch, index) { |
|
charToIntMap[ch] = index; |
|
intToCharMap[index] = ch; |
|
}); |
|
|
|
/** |
|
* Encode an integer in the range of 0 to 63 to a single base 64 digit. |
|
*/ |
|
exports.encode = function base64_encode(aNumber) { |
|
if (aNumber in intToCharMap) { |
|
return intToCharMap[aNumber]; |
|
} |
|
throw new TypeError("Must be between 0 and 63: " + aNumber); |
|
}; |
|
|
|
/** |
|
* Decode a single base 64 digit to an integer. |
|
*/ |
|
exports.decode = function base64_decode(aChar) { |
|
if (aChar in charToIntMap) { |
|
return charToIntMap[aChar]; |
|
} |
|
throw new TypeError("Not a valid base 64 digit: " + aChar); |
|
}; |
|
|
|
}); |
|
|
|
},{"amdefine":20}],15:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
/** |
|
* Recursive implementation of binary search. |
|
* |
|
* @param aLow Indices here and lower do not contain the needle. |
|
* @param aHigh Indices here and higher do not contain the needle. |
|
* @param aNeedle The element being searched for. |
|
* @param aHaystack The non-empty array being searched. |
|
* @param aCompare Function which takes two elements and returns -1, 0, or 1. |
|
*/ |
|
function recursiveSearch(aLow, aHigh, aNeedle, aHaystack, aCompare) { |
|
// This function terminates when one of the following is true: |
|
// |
|
// 1. We find the exact element we are looking for. |
|
// |
|
// 2. We did not find the exact element, but we can return the next |
|
// closest element that is less than that element. |
|
// |
|
// 3. We did not find the exact element, and there is no next-closest |
|
// element which is less than the one we are searching for, so we |
|
// return null. |
|
var mid = Math.floor((aHigh - aLow) / 2) + aLow; |
|
var cmp = aCompare(aNeedle, aHaystack[mid], true); |
|
if (cmp === 0) { |
|
// Found the element we are looking for. |
|
return aHaystack[mid]; |
|
} |
|
else if (cmp > 0) { |
|
// aHaystack[mid] is greater than our needle. |
|
if (aHigh - mid > 1) { |
|
// The element is in the upper half. |
|
return recursiveSearch(mid, aHigh, aNeedle, aHaystack, aCompare); |
|
} |
|
// We did not find an exact match, return the next closest one |
|
// (termination case 2). |
|
return aHaystack[mid]; |
|
} |
|
else { |
|
// aHaystack[mid] is less than our needle. |
|
if (mid - aLow > 1) { |
|
// The element is in the lower half. |
|
return recursiveSearch(aLow, mid, aNeedle, aHaystack, aCompare); |
|
} |
|
// The exact needle element was not found in this haystack. Determine if |
|
// we are in termination case (2) or (3) and return the appropriate thing. |
|
return aLow < 0 |
|
? null |
|
: aHaystack[aLow]; |
|
} |
|
} |
|
|
|
/** |
|
* This is an implementation of binary search which will always try and return |
|
* the next lowest value checked if there is no exact hit. This is because |
|
* mappings between original and generated line/col pairs are single points, |
|
* and there is an implicit region between each of them, so a miss just means |
|
* that you aren't on the very start of a region. |
|
* |
|
* @param aNeedle The element you are looking for. |
|
* @param aHaystack The array that is being searched. |
|
* @param aCompare A function which takes the needle and an element in the |
|
* array and returns -1, 0, or 1 depending on whether the needle is less |
|
* than, equal to, or greater than the element, respectively. |
|
*/ |
|
exports.search = function search(aNeedle, aHaystack, aCompare) { |
|
return aHaystack.length > 0 |
|
? recursiveSearch(-1, aHaystack.length, aNeedle, aHaystack, aCompare) |
|
: null; |
|
}; |
|
|
|
}); |
|
|
|
},{"amdefine":20}],16:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
var util = _dereq_('./util'); |
|
var binarySearch = _dereq_('./binary-search'); |
|
var ArraySet = _dereq_('./array-set').ArraySet; |
|
var base64VLQ = _dereq_('./base64-vlq'); |
|
|
|
/** |
|
* A SourceMapConsumer instance represents a parsed source map which we can |
|
* query for information about the original file positions by giving it a file |
|
* position in the generated source. |
|
* |
|
* The only parameter is the raw source map (either as a JSON string, or |
|
* already parsed to an object). According to the spec, source maps have the |
|
* following attributes: |
|
* |
|
* - version: Which version of the source map spec this map is following. |
|
* - sources: An array of URLs to the original source files. |
|
* - names: An array of identifiers which can be referrenced by individual mappings. |
|
* - sourceRoot: Optional. The URL root from which all sources are relative. |
|
* - sourcesContent: Optional. An array of contents of the original source files. |
|
* - mappings: A string of base64 VLQs which contain the actual mappings. |
|
* - file: The generated file this source map is associated with. |
|
* |
|
* Here is an example source map, taken from the source map spec[0]: |
|
* |
|
* { |
|
* version : 3, |
|
* file: "out.js", |
|
* sourceRoot : "", |
|
* sources: ["foo.js", "bar.js"], |
|
* names: ["src", "maps", "are", "fun"], |
|
* mappings: "AA,AB;;ABCDE;" |
|
* } |
|
* |
|
* [0]: https://docs.google.com/document/d/1U1RGAehQwRypUTovF1KRlpiOFze0b-_2gc6fAH0KY0k/edit?pli=1# |
|
*/ |
|
function SourceMapConsumer(aSourceMap) { |
|
var sourceMap = aSourceMap; |
|
if (typeof aSourceMap === 'string') { |
|
sourceMap = JSON.parse(aSourceMap.replace(/^\)\]\}'/, '')); |
|
} |
|
|
|
var version = util.getArg(sourceMap, 'version'); |
|
var sources = util.getArg(sourceMap, 'sources'); |
|
// Sass 3.3 leaves out the 'names' array, so we deviate from the spec (which |
|
// requires the array) to play nice here. |
|
var names = util.getArg(sourceMap, 'names', []); |
|
var sourceRoot = util.getArg(sourceMap, 'sourceRoot', null); |
|
var sourcesContent = util.getArg(sourceMap, 'sourcesContent', null); |
|
var mappings = util.getArg(sourceMap, 'mappings'); |
|
var file = util.getArg(sourceMap, 'file', null); |
|
|
|
// Once again, Sass deviates from the spec and supplies the version as a |
|
// string rather than a number, so we use loose equality checking here. |
|
if (version != this._version) { |
|
throw new Error('Unsupported version: ' + version); |
|
} |
|
|
|
// Pass `true` below to allow duplicate names and sources. While source maps |
|
// are intended to be compressed and deduplicated, the TypeScript compiler |
|
// sometimes generates source maps with duplicates in them. See Github issue |
|
// #72 and bugzil.la/889492. |
|
this._names = ArraySet.fromArray(names, true); |
|
this._sources = ArraySet.fromArray(sources, true); |
|
|
|
this.sourceRoot = sourceRoot; |
|
this.sourcesContent = sourcesContent; |
|
this._mappings = mappings; |
|
this.file = file; |
|
} |
|
|
|
/** |
|
* Create a SourceMapConsumer from a SourceMapGenerator. |
|
* |
|
* @param SourceMapGenerator aSourceMap |
|
* The source map that will be consumed. |
|
* @returns SourceMapConsumer |
|
*/ |
|
SourceMapConsumer.fromSourceMap = |
|
function SourceMapConsumer_fromSourceMap(aSourceMap) { |
|
var smc = Object.create(SourceMapConsumer.prototype); |
|
|
|
smc._names = ArraySet.fromArray(aSourceMap._names.toArray(), true); |
|
smc._sources = ArraySet.fromArray(aSourceMap._sources.toArray(), true); |
|
smc.sourceRoot = aSourceMap._sourceRoot; |
|
smc.sourcesContent = aSourceMap._generateSourcesContent(smc._sources.toArray(), |
|
smc.sourceRoot); |
|
smc.file = aSourceMap._file; |
|
|
|
smc.__generatedMappings = aSourceMap._mappings.slice() |
|
.sort(util.compareByGeneratedPositions); |
|
smc.__originalMappings = aSourceMap._mappings.slice() |
|
.sort(util.compareByOriginalPositions); |
|
|
|
return smc; |
|
}; |
|
|
|
/** |
|
* The version of the source mapping spec that we are consuming. |
|
*/ |
|
SourceMapConsumer.prototype._version = 3; |
|
|
|
/** |
|
* The list of original sources. |
|
*/ |
|
Object.defineProperty(SourceMapConsumer.prototype, 'sources', { |
|
get: function () { |
|
return this._sources.toArray().map(function (s) { |
|
return this.sourceRoot ? util.join(this.sourceRoot, s) : s; |
|
}, this); |
|
} |
|
}); |
|
|
|
// `__generatedMappings` and `__originalMappings` are arrays that hold the |
|
// parsed mapping coordinates from the source map's "mappings" attribute. They |
|
// are lazily instantiated, accessed via the `_generatedMappings` and |
|
// `_originalMappings` getters respectively, and we only parse the mappings |
|
// and create these arrays once queried for a source location. We jump through |
|
// these hoops because there can be many thousands of mappings, and parsing |
|
// them is expensive, so we only want to do it if we must. |
|
// |
|
// Each object in the arrays is of the form: |
|
// |
|
// { |
|
// generatedLine: The line number in the generated code, |
|
// generatedColumn: The column number in the generated code, |
|
// source: The path to the original source file that generated this |
|
// chunk of code, |
|
// originalLine: The line number in the original source that |
|
// corresponds to this chunk of generated code, |
|
// originalColumn: The column number in the original source that |
|
// corresponds to this chunk of generated code, |
|
// name: The name of the original symbol which generated this chunk of |
|
// code. |
|
// } |
|
// |
|
// All properties except for `generatedLine` and `generatedColumn` can be |
|
// `null`. |
|
// |
|
// `_generatedMappings` is ordered by the generated positions. |
|
// |
|
// `_originalMappings` is ordered by the original positions. |
|
|
|
SourceMapConsumer.prototype.__generatedMappings = null; |
|
Object.defineProperty(SourceMapConsumer.prototype, '_generatedMappings', { |
|
get: function () { |
|
if (!this.__generatedMappings) { |
|
this.__generatedMappings = []; |
|
this.__originalMappings = []; |
|
this._parseMappings(this._mappings, this.sourceRoot); |
|
} |
|
|
|
return this.__generatedMappings; |
|
} |
|
}); |
|
|
|
SourceMapConsumer.prototype.__originalMappings = null; |
|
Object.defineProperty(SourceMapConsumer.prototype, '_originalMappings', { |
|
get: function () { |
|
if (!this.__originalMappings) { |
|
this.__generatedMappings = []; |
|
this.__originalMappings = []; |
|
this._parseMappings(this._mappings, this.sourceRoot); |
|
} |
|
|
|
return this.__originalMappings; |
|
} |
|
}); |
|
|
|
/** |
|
* Parse the mappings in a string in to a data structure which we can easily |
|
* query (the ordered arrays in the `this.__generatedMappings` and |
|
* `this.__originalMappings` properties). |
|
*/ |
|
SourceMapConsumer.prototype._parseMappings = |
|
function SourceMapConsumer_parseMappings(aStr, aSourceRoot) { |
|
var generatedLine = 1; |
|
var previousGeneratedColumn = 0; |
|
var previousOriginalLine = 0; |
|
var previousOriginalColumn = 0; |
|
var previousSource = 0; |
|
var previousName = 0; |
|
var mappingSeparator = /^[,;]/; |
|
var str = aStr; |
|
var mapping; |
|
var temp; |
|
|
|
while (str.length > 0) { |
|
if (str.charAt(0) === ';') { |
|
generatedLine++; |
|
str = str.slice(1); |
|
previousGeneratedColumn = 0; |
|
} |
|
else if (str.charAt(0) === ',') { |
|
str = str.slice(1); |
|
} |
|
else { |
|
mapping = {}; |
|
mapping.generatedLine = generatedLine; |
|
|
|
// Generated column. |
|
temp = base64VLQ.decode(str); |
|
mapping.generatedColumn = previousGeneratedColumn + temp.value; |
|
previousGeneratedColumn = mapping.generatedColumn; |
|
str = temp.rest; |
|
|
|
if (str.length > 0 && !mappingSeparator.test(str.charAt(0))) { |
|
// Original source. |
|
temp = base64VLQ.decode(str); |
|
mapping.source = this._sources.at(previousSource + temp.value); |
|
previousSource += temp.value; |
|
str = temp.rest; |
|
if (str.length === 0 || mappingSeparator.test(str.charAt(0))) { |
|
throw new Error('Found a source, but no line and column'); |
|
} |
|
|
|
// Original line. |
|
temp = base64VLQ.decode(str); |
|
mapping.originalLine = previousOriginalLine + temp.value; |
|
previousOriginalLine = mapping.originalLine; |
|
// Lines are stored 0-based |
|
mapping.originalLine += 1; |
|
str = temp.rest; |
|
if (str.length === 0 || mappingSeparator.test(str.charAt(0))) { |
|
throw new Error('Found a source and line, but no column'); |
|
} |
|
|
|
// Original column. |
|
temp = base64VLQ.decode(str); |
|
mapping.originalColumn = previousOriginalColumn + temp.value; |
|
previousOriginalColumn = mapping.originalColumn; |
|
str = temp.rest; |
|
|
|
if (str.length > 0 && !mappingSeparator.test(str.charAt(0))) { |
|
// Original name. |
|
temp = base64VLQ.decode(str); |
|
mapping.name = this._names.at(previousName + temp.value); |
|
previousName += temp.value; |
|
str = temp.rest; |
|
} |
|
} |
|
|
|
this.__generatedMappings.push(mapping); |
|
if (typeof mapping.originalLine === 'number') { |
|
this.__originalMappings.push(mapping); |
|
} |
|
} |
|
} |
|
|
|
this.__originalMappings.sort(util.compareByOriginalPositions); |
|
}; |
|
|
|
/** |
|
* Find the mapping that best matches the hypothetical "needle" mapping that |
|
* we are searching for in the given "haystack" of mappings. |
|
*/ |
|
SourceMapConsumer.prototype._findMapping = |
|
function SourceMapConsumer_findMapping(aNeedle, aMappings, aLineName, |
|
aColumnName, aComparator) { |
|
// To return the position we are searching for, we must first find the |
|
// mapping for the given position and then return the opposite position it |
|
// points to. Because the mappings are sorted, we can use binary search to |
|
// find the best mapping. |
|
|
|
if (aNeedle[aLineName] <= 0) { |
|
throw new TypeError('Line must be greater than or equal to 1, got ' |
|
+ aNeedle[aLineName]); |
|
} |
|
if (aNeedle[aColumnName] < 0) { |
|
throw new TypeError('Column must be greater than or equal to 0, got ' |
|
+ aNeedle[aColumnName]); |
|
} |
|
|
|
return binarySearch.search(aNeedle, aMappings, aComparator); |
|
}; |
|
|
|
/** |
|
* Returns the original source, line, and column information for the generated |
|
* source's line and column positions provided. The only argument is an object |
|
* with the following properties: |
|
* |
|
* - line: The line number in the generated source. |
|
* - column: The column number in the generated source. |
|
* |
|
* and an object is returned with the following properties: |
|
* |
|
* - source: The original source file, or null. |
|
* - line: The line number in the original source, or null. |
|
* - column: The column number in the original source, or null. |
|
* - name: The original identifier, or null. |
|
*/ |
|
SourceMapConsumer.prototype.originalPositionFor = |
|
function SourceMapConsumer_originalPositionFor(aArgs) { |
|
var needle = { |
|
generatedLine: util.getArg(aArgs, 'line'), |
|
generatedColumn: util.getArg(aArgs, 'column') |
|
}; |
|
|
|
var mapping = this._findMapping(needle, |
|
this._generatedMappings, |
|
"generatedLine", |
|
"generatedColumn", |
|
util.compareByGeneratedPositions); |
|
|
|
if (mapping) { |
|
var source = util.getArg(mapping, 'source', null); |
|
if (source && this.sourceRoot) { |
|
source = util.join(this.sourceRoot, source); |
|
} |
|
return { |
|
source: source, |
|
line: util.getArg(mapping, 'originalLine', null), |
|
column: util.getArg(mapping, 'originalColumn', null), |
|
name: util.getArg(mapping, 'name', null) |
|
}; |
|
} |
|
|
|
return { |
|
source: null, |
|
line: null, |
|
column: null, |
|
name: null |
|
}; |
|
}; |
|
|
|
/** |
|
* Returns the original source content. The only argument is the url of the |
|
* original source file. Returns null if no original source content is |
|
* availible. |
|
*/ |
|
SourceMapConsumer.prototype.sourceContentFor = |
|
function SourceMapConsumer_sourceContentFor(aSource) { |
|
if (!this.sourcesContent) { |
|
return null; |
|
} |
|
|
|
if (this.sourceRoot) { |
|
aSource = util.relative(this.sourceRoot, aSource); |
|
} |
|
|
|
if (this._sources.has(aSource)) { |
|
return this.sourcesContent[this._sources.indexOf(aSource)]; |
|
} |
|
|
|
var url; |
|
if (this.sourceRoot |
|
&& (url = util.urlParse(this.sourceRoot))) { |
|
// XXX: file:// URIs and absolute paths lead to unexpected behavior for |
|
// many users. We can help them out when they expect file:// URIs to |
|
// behave like it would if they were running a local HTTP server. See |
|
// https://bugzilla.mozilla.org/show_bug.cgi?id=885597. |
|
var fileUriAbsPath = aSource.replace(/^file:\/\//, ""); |
|
if (url.scheme == "file" |
|
&& this._sources.has(fileUriAbsPath)) { |
|
return this.sourcesContent[this._sources.indexOf(fileUriAbsPath)] |
|
} |
|
|
|
if ((!url.path || url.path == "/") |
|
&& this._sources.has("/" + aSource)) { |
|
return this.sourcesContent[this._sources.indexOf("/" + aSource)]; |
|
} |
|
} |
|
|
|
throw new Error('"' + aSource + '" is not in the SourceMap.'); |
|
}; |
|
|
|
/** |
|
* Returns the generated line and column information for the original source, |
|
* line, and column positions provided. The only argument is an object with |
|
* the following properties: |
|
* |
|
* - source: The filename of the original source. |
|
* - line: The line number in the original source. |
|
* - column: The column number in the original source. |
|
* |
|
* and an object is returned with the following properties: |
|
* |
|
* - line: The line number in the generated source, or null. |
|
* - column: The column number in the generated source, or null. |
|
*/ |
|
SourceMapConsumer.prototype.generatedPositionFor = |
|
function SourceMapConsumer_generatedPositionFor(aArgs) { |
|
var needle = { |
|
source: util.getArg(aArgs, 'source'), |
|
originalLine: util.getArg(aArgs, 'line'), |
|
originalColumn: util.getArg(aArgs, 'column') |
|
}; |
|
|
|
if (this.sourceRoot) { |
|
needle.source = util.relative(this.sourceRoot, needle.source); |
|
} |
|
|
|
var mapping = this._findMapping(needle, |
|
this._originalMappings, |
|
"originalLine", |
|
"originalColumn", |
|
util.compareByOriginalPositions); |
|
|
|
if (mapping) { |
|
return { |
|
line: util.getArg(mapping, 'generatedLine', null), |
|
column: util.getArg(mapping, 'generatedColumn', null) |
|
}; |
|
} |
|
|
|
return { |
|
line: null, |
|
column: null |
|
}; |
|
}; |
|
|
|
SourceMapConsumer.GENERATED_ORDER = 1; |
|
SourceMapConsumer.ORIGINAL_ORDER = 2; |
|
|
|
/** |
|
* Iterate over each mapping between an original source/line/column and a |
|
* generated line/column in this source map. |
|
* |
|
* @param Function aCallback |
|
* The function that is called with each mapping. |
|
* @param Object aContext |
|
* Optional. If specified, this object will be the value of `this` every |
|
* time that `aCallback` is called. |
|
* @param aOrder |
|
* Either `SourceMapConsumer.GENERATED_ORDER` or |
|
* `SourceMapConsumer.ORIGINAL_ORDER`. Specifies whether you want to |
|
* iterate over the mappings sorted by the generated file's line/column |
|
* order or the original's source/line/column order, respectively. Defaults to |
|
* `SourceMapConsumer.GENERATED_ORDER`. |
|
*/ |
|
SourceMapConsumer.prototype.eachMapping = |
|
function SourceMapConsumer_eachMapping(aCallback, aContext, aOrder) { |
|
var context = aContext || null; |
|
var order = aOrder || SourceMapConsumer.GENERATED_ORDER; |
|
|
|
var mappings; |
|
switch (order) { |
|
case SourceMapConsumer.GENERATED_ORDER: |
|
mappings = this._generatedMappings; |
|
break; |
|
case SourceMapConsumer.ORIGINAL_ORDER: |
|
mappings = this._originalMappings; |
|
break; |
|
default: |
|
throw new Error("Unknown order of iteration."); |
|
} |
|
|
|
var sourceRoot = this.sourceRoot; |
|
mappings.map(function (mapping) { |
|
var source = mapping.source; |
|
if (source && sourceRoot) { |
|
source = util.join(sourceRoot, source); |
|
} |
|
return { |
|
source: source, |
|
generatedLine: mapping.generatedLine, |
|
generatedColumn: mapping.generatedColumn, |
|
originalLine: mapping.originalLine, |
|
originalColumn: mapping.originalColumn, |
|
name: mapping.name |
|
}; |
|
}).forEach(aCallback, context); |
|
}; |
|
|
|
exports.SourceMapConsumer = SourceMapConsumer; |
|
|
|
}); |
|
|
|
},{"./array-set":12,"./base64-vlq":13,"./binary-search":15,"./util":19,"amdefine":20}],17:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
var base64VLQ = _dereq_('./base64-vlq'); |
|
var util = _dereq_('./util'); |
|
var ArraySet = _dereq_('./array-set').ArraySet; |
|
|
|
/** |
|
* An instance of the SourceMapGenerator represents a source map which is |
|
* being built incrementally. To create a new one, you must pass an object |
|
* with the following properties: |
|
* |
|
* - file: The filename of the generated source. |
|
* - sourceRoot: An optional root for all URLs in this source map. |
|
*/ |
|
function SourceMapGenerator(aArgs) { |
|
this._file = util.getArg(aArgs, 'file'); |
|
this._sourceRoot = util.getArg(aArgs, 'sourceRoot', null); |
|
this._sources = new ArraySet(); |
|
this._names = new ArraySet(); |
|
this._mappings = []; |
|
this._sourcesContents = null; |
|
} |
|
|
|
SourceMapGenerator.prototype._version = 3; |
|
|
|
/** |
|
* Creates a new SourceMapGenerator based on a SourceMapConsumer |
|
* |
|
* @param aSourceMapConsumer The SourceMap. |
|
*/ |
|
SourceMapGenerator.fromSourceMap = |
|
function SourceMapGenerator_fromSourceMap(aSourceMapConsumer) { |
|
var sourceRoot = aSourceMapConsumer.sourceRoot; |
|
var generator = new SourceMapGenerator({ |
|
file: aSourceMapConsumer.file, |
|
sourceRoot: sourceRoot |
|
}); |
|
aSourceMapConsumer.eachMapping(function (mapping) { |
|
var newMapping = { |
|
generated: { |
|
line: mapping.generatedLine, |
|
column: mapping.generatedColumn |
|
} |
|
}; |
|
|
|
if (mapping.source) { |
|
newMapping.source = mapping.source; |
|
if (sourceRoot) { |
|
newMapping.source = util.relative(sourceRoot, newMapping.source); |
|
} |
|
|
|
newMapping.original = { |
|
line: mapping.originalLine, |
|
column: mapping.originalColumn |
|
}; |
|
|
|
if (mapping.name) { |
|
newMapping.name = mapping.name; |
|
} |
|
} |
|
|
|
generator.addMapping(newMapping); |
|
}); |
|
aSourceMapConsumer.sources.forEach(function (sourceFile) { |
|
var content = aSourceMapConsumer.sourceContentFor(sourceFile); |
|
if (content) { |
|
generator.setSourceContent(sourceFile, content); |
|
} |
|
}); |
|
return generator; |
|
}; |
|
|
|
/** |
|
* Add a single mapping from original source line and column to the generated |
|
* source's line and column for this source map being created. The mapping |
|
* object should have the following properties: |
|
* |
|
* - generated: An object with the generated line and column positions. |
|
* - original: An object with the original line and column positions. |
|
* - source: The original source file (relative to the sourceRoot). |
|
* - name: An optional original token name for this mapping. |
|
*/ |
|
SourceMapGenerator.prototype.addMapping = |
|
function SourceMapGenerator_addMapping(aArgs) { |
|
var generated = util.getArg(aArgs, 'generated'); |
|
var original = util.getArg(aArgs, 'original', null); |
|
var source = util.getArg(aArgs, 'source', null); |
|
var name = util.getArg(aArgs, 'name', null); |
|
|
|
this._validateMapping(generated, original, source, name); |
|
|
|
if (source && !this._sources.has(source)) { |
|
this._sources.add(source); |
|
} |
|
|
|
if (name && !this._names.has(name)) { |
|
this._names.add(name); |
|
} |
|
|
|
this._mappings.push({ |
|
generatedLine: generated.line, |
|
generatedColumn: generated.column, |
|
originalLine: original != null && original.line, |
|
originalColumn: original != null && original.column, |
|
source: source, |
|
name: name |
|
}); |
|
}; |
|
|
|
/** |
|
* Set the source content for a source file. |
|
*/ |
|
SourceMapGenerator.prototype.setSourceContent = |
|
function SourceMapGenerator_setSourceContent(aSourceFile, aSourceContent) { |
|
var source = aSourceFile; |
|
if (this._sourceRoot) { |
|
source = util.relative(this._sourceRoot, source); |
|
} |
|
|
|
if (aSourceContent !== null) { |
|
// Add the source content to the _sourcesContents map. |
|
// Create a new _sourcesContents map if the property is null. |
|
if (!this._sourcesContents) { |
|
this._sourcesContents = {}; |
|
} |
|
this._sourcesContents[util.toSetString(source)] = aSourceContent; |
|
} else { |
|
// Remove the source file from the _sourcesContents map. |
|
// If the _sourcesContents map is empty, set the property to null. |
|
delete this._sourcesContents[util.toSetString(source)]; |
|
if (Object.keys(this._sourcesContents).length === 0) { |
|
this._sourcesContents = null; |
|
} |
|
} |
|
}; |
|
|
|
/** |
|
* Applies the mappings of a sub-source-map for a specific source file to the |
|
* source map being generated. Each mapping to the supplied source file is |
|
* rewritten using the supplied source map. Note: The resolution for the |
|
* resulting mappings is the minimium of this map and the supplied map. |
|
* |
|
* @param aSourceMapConsumer The source map to be applied. |
|
* @param aSourceFile Optional. The filename of the source file. |
|
* If omitted, SourceMapConsumer's file property will be used. |
|
*/ |
|
SourceMapGenerator.prototype.applySourceMap = |
|
function SourceMapGenerator_applySourceMap(aSourceMapConsumer, aSourceFile) { |
|
// If aSourceFile is omitted, we will use the file property of the SourceMap |
|
if (!aSourceFile) { |
|
aSourceFile = aSourceMapConsumer.file; |
|
} |
|
var sourceRoot = this._sourceRoot; |
|
// Make "aSourceFile" relative if an absolute Url is passed. |
|
if (sourceRoot) { |
|
aSourceFile = util.relative(sourceRoot, aSourceFile); |
|
} |
|
// Applying the SourceMap can add and remove items from the sources and |
|
// the names array. |
|
var newSources = new ArraySet(); |
|
var newNames = new ArraySet(); |
|
|
|
// Find mappings for the "aSourceFile" |
|
this._mappings.forEach(function (mapping) { |
|
if (mapping.source === aSourceFile && mapping.originalLine) { |
|
// Check if it can be mapped by the source map, then update the mapping. |
|
var original = aSourceMapConsumer.originalPositionFor({ |
|
line: mapping.originalLine, |
|
column: mapping.originalColumn |
|
}); |
|
if (original.source !== null) { |
|
// Copy mapping |
|
if (sourceRoot) { |
|
mapping.source = util.relative(sourceRoot, original.source); |
|
} else { |
|
mapping.source = original.source; |
|
} |
|
mapping.originalLine = original.line; |
|
mapping.originalColumn = original.column; |
|
if (original.name !== null && mapping.name !== null) { |
|
// Only use the identifier name if it's an identifier |
|
// in both SourceMaps |
|
mapping.name = original.name; |
|
} |
|
} |
|
} |
|
|
|
var source = mapping.source; |
|
if (source && !newSources.has(source)) { |
|
newSources.add(source); |
|
} |
|
|
|
var name = mapping.name; |
|
if (name && !newNames.has(name)) { |
|
newNames.add(name); |
|
} |
|
|
|
}, this); |
|
this._sources = newSources; |
|
this._names = newNames; |
|
|
|
// Copy sourcesContents of applied map. |
|
aSourceMapConsumer.sources.forEach(function (sourceFile) { |
|
var content = aSourceMapConsumer.sourceContentFor(sourceFile); |
|
if (content) { |
|
if (sourceRoot) { |
|
sourceFile = util.relative(sourceRoot, sourceFile); |
|
} |
|
this.setSourceContent(sourceFile, content); |
|
} |
|
}, this); |
|
}; |
|
|
|
/** |
|
* A mapping can have one of the three levels of data: |
|
* |
|
* 1. Just the generated position. |
|
* 2. The Generated position, original position, and original source. |
|
* 3. Generated and original position, original source, as well as a name |
|
* token. |
|
* |
|
* To maintain consistency, we validate that any new mapping being added falls |
|
* in to one of these categories. |
|
*/ |
|
SourceMapGenerator.prototype._validateMapping = |
|
function SourceMapGenerator_validateMapping(aGenerated, aOriginal, aSource, |
|
aName) { |
|
if (aGenerated && 'line' in aGenerated && 'column' in aGenerated |
|
&& aGenerated.line > 0 && aGenerated.column >= 0 |
|
&& !aOriginal && !aSource && !aName) { |
|
// Case 1. |
|
return; |
|
} |
|
else if (aGenerated && 'line' in aGenerated && 'column' in aGenerated |
|
&& aOriginal && 'line' in aOriginal && 'column' in aOriginal |
|
&& aGenerated.line > 0 && aGenerated.column >= 0 |
|
&& aOriginal.line > 0 && aOriginal.column >= 0 |
|
&& aSource) { |
|
// Cases 2 and 3. |
|
return; |
|
} |
|
else { |
|
throw new Error('Invalid mapping: ' + JSON.stringify({ |
|
generated: aGenerated, |
|
source: aSource, |
|
orginal: aOriginal, |
|
name: aName |
|
})); |
|
} |
|
}; |
|
|
|
/** |
|
* Serialize the accumulated mappings in to the stream of base 64 VLQs |
|
* specified by the source map format. |
|
*/ |
|
SourceMapGenerator.prototype._serializeMappings = |
|
function SourceMapGenerator_serializeMappings() { |
|
var previousGeneratedColumn = 0; |
|
var previousGeneratedLine = 1; |
|
var previousOriginalColumn = 0; |
|
var previousOriginalLine = 0; |
|
var previousName = 0; |
|
var previousSource = 0; |
|
var result = ''; |
|
var mapping; |
|
|
|
// The mappings must be guaranteed to be in sorted order before we start |
|
// serializing them or else the generated line numbers (which are defined |
|
// via the ';' separators) will be all messed up. Note: it might be more |
|
// performant to maintain the sorting as we insert them, rather than as we |
|
// serialize them, but the big O is the same either way. |
|
this._mappings.sort(util.compareByGeneratedPositions); |
|
|
|
for (var i = 0, len = this._mappings.length; i < len; i++) { |
|
mapping = this._mappings[i]; |
|
|
|
if (mapping.generatedLine !== previousGeneratedLine) { |
|
previousGeneratedColumn = 0; |
|
while (mapping.generatedLine !== previousGeneratedLine) { |
|
result += ';'; |
|
previousGeneratedLine++; |
|
} |
|
} |
|
else { |
|
if (i > 0) { |
|
if (!util.compareByGeneratedPositions(mapping, this._mappings[i - 1])) { |
|
continue; |
|
} |
|
result += ','; |
|
} |
|
} |
|
|
|
result += base64VLQ.encode(mapping.generatedColumn |
|
- previousGeneratedColumn); |
|
previousGeneratedColumn = mapping.generatedColumn; |
|
|
|
if (mapping.source) { |
|
result += base64VLQ.encode(this._sources.indexOf(mapping.source) |
|
- previousSource); |
|
previousSource = this._sources.indexOf(mapping.source); |
|
|
|
// lines are stored 0-based in SourceMap spec version 3 |
|
result += base64VLQ.encode(mapping.originalLine - 1 |
|
- previousOriginalLine); |
|
previousOriginalLine = mapping.originalLine - 1; |
|
|
|
result += base64VLQ.encode(mapping.originalColumn |
|
- previousOriginalColumn); |
|
previousOriginalColumn = mapping.originalColumn; |
|
|
|
if (mapping.name) { |
|
result += base64VLQ.encode(this._names.indexOf(mapping.name) |
|
- previousName); |
|
previousName = this._names.indexOf(mapping.name); |
|
} |
|
} |
|
} |
|
|
|
return result; |
|
}; |
|
|
|
SourceMapGenerator.prototype._generateSourcesContent = |
|
function SourceMapGenerator_generateSourcesContent(aSources, aSourceRoot) { |
|
return aSources.map(function (source) { |
|
if (!this._sourcesContents) { |
|
return null; |
|
} |
|
if (aSourceRoot) { |
|
source = util.relative(aSourceRoot, source); |
|
} |
|
var key = util.toSetString(source); |
|
return Object.prototype.hasOwnProperty.call(this._sourcesContents, |
|
key) |
|
? this._sourcesContents[key] |
|
: null; |
|
}, this); |
|
}; |
|
|
|
/** |
|
* Externalize the source map. |
|
*/ |
|
SourceMapGenerator.prototype.toJSON = |
|
function SourceMapGenerator_toJSON() { |
|
var map = { |
|
version: this._version, |
|
file: this._file, |
|
sources: this._sources.toArray(), |
|
names: this._names.toArray(), |
|
mappings: this._serializeMappings() |
|
}; |
|
if (this._sourceRoot) { |
|
map.sourceRoot = this._sourceRoot; |
|
} |
|
if (this._sourcesContents) { |
|
map.sourcesContent = this._generateSourcesContent(map.sources, map.sourceRoot); |
|
} |
|
|
|
return map; |
|
}; |
|
|
|
/** |
|
* Render the source map being generated to a string. |
|
*/ |
|
SourceMapGenerator.prototype.toString = |
|
function SourceMapGenerator_toString() { |
|
return JSON.stringify(this); |
|
}; |
|
|
|
exports.SourceMapGenerator = SourceMapGenerator; |
|
|
|
}); |
|
|
|
},{"./array-set":12,"./base64-vlq":13,"./util":19,"amdefine":20}],18:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
var SourceMapGenerator = _dereq_('./source-map-generator').SourceMapGenerator; |
|
var util = _dereq_('./util'); |
|
|
|
/** |
|
* SourceNodes provide a way to abstract over interpolating/concatenating |
|
* snippets of generated JavaScript source code while maintaining the line and |
|
* column information associated with the original source code. |
|
* |
|
* @param aLine The original line number. |
|
* @param aColumn The original column number. |
|
* @param aSource The original source's filename. |
|
* @param aChunks Optional. An array of strings which are snippets of |
|
* generated JS, or other SourceNodes. |
|
* @param aName The original identifier. |
|
*/ |
|
function SourceNode(aLine, aColumn, aSource, aChunks, aName) { |
|
this.children = []; |
|
this.sourceContents = {}; |
|
this.line = aLine === undefined ? null : aLine; |
|
this.column = aColumn === undefined ? null : aColumn; |
|
this.source = aSource === undefined ? null : aSource; |
|
this.name = aName === undefined ? null : aName; |
|
if (aChunks != null) this.add(aChunks); |
|
} |
|
|
|
/** |
|
* Creates a SourceNode from generated code and a SourceMapConsumer. |
|
* |
|
* @param aGeneratedCode The generated code |
|
* @param aSourceMapConsumer The SourceMap for the generated code |
|
*/ |
|
SourceNode.fromStringWithSourceMap = |
|
function SourceNode_fromStringWithSourceMap(aGeneratedCode, aSourceMapConsumer) { |
|
// The SourceNode we want to fill with the generated code |
|
// and the SourceMap |
|
var node = new SourceNode(); |
|
|
|
// The generated code |
|
// Processed fragments are removed from this array. |
|
var remainingLines = aGeneratedCode.split('\n'); |
|
|
|
// We need to remember the position of "remainingLines" |
|
var lastGeneratedLine = 1, lastGeneratedColumn = 0; |
|
|
|
// The generate SourceNodes we need a code range. |
|
// To extract it current and last mapping is used. |
|
// Here we store the last mapping. |
|
var lastMapping = null; |
|
|
|
aSourceMapConsumer.eachMapping(function (mapping) { |
|
if (lastMapping === null) { |
|
// We add the generated code until the first mapping |
|
// to the SourceNode without any mapping. |
|
// Each line is added as separate string. |
|
while (lastGeneratedLine < mapping.generatedLine) { |
|
node.add(remainingLines.shift() + "\n"); |
|
lastGeneratedLine++; |
|
} |
|
if (lastGeneratedColumn < mapping.generatedColumn) { |
|
var nextLine = remainingLines[0]; |
|
node.add(nextLine.substr(0, mapping.generatedColumn)); |
|
remainingLines[0] = nextLine.substr(mapping.generatedColumn); |
|
lastGeneratedColumn = mapping.generatedColumn; |
|
} |
|
} else { |
|
// We add the code from "lastMapping" to "mapping": |
|
// First check if there is a new line in between. |
|
if (lastGeneratedLine < mapping.generatedLine) { |
|
var code = ""; |
|
// Associate full lines with "lastMapping" |
|
do { |
|
code += remainingLines.shift() + "\n"; |
|
lastGeneratedLine++; |
|
lastGeneratedColumn = 0; |
|
} while (lastGeneratedLine < mapping.generatedLine); |
|
// When we reached the correct line, we add code until we |
|
// reach the correct column too. |
|
if (lastGeneratedColumn < mapping.generatedColumn) { |
|
var nextLine = remainingLines[0]; |
|
code += nextLine.substr(0, mapping.generatedColumn); |
|
remainingLines[0] = nextLine.substr(mapping.generatedColumn); |
|
lastGeneratedColumn = mapping.generatedColumn; |
|
} |
|
// Create the SourceNode. |
|
addMappingWithCode(lastMapping, code); |
|
} else { |
|
// There is no new line in between. |
|
// Associate the code between "lastGeneratedColumn" and |
|
// "mapping.generatedColumn" with "lastMapping" |
|
var nextLine = remainingLines[0]; |
|
var code = nextLine.substr(0, mapping.generatedColumn - |
|
lastGeneratedColumn); |
|
remainingLines[0] = nextLine.substr(mapping.generatedColumn - |
|
lastGeneratedColumn); |
|
lastGeneratedColumn = mapping.generatedColumn; |
|
addMappingWithCode(lastMapping, code); |
|
} |
|
} |
|
lastMapping = mapping; |
|
}, this); |
|
// We have processed all mappings. |
|
// Associate the remaining code in the current line with "lastMapping" |
|
// and add the remaining lines without any mapping |
|
addMappingWithCode(lastMapping, remainingLines.join("\n")); |
|
|
|
// Copy sourcesContent into SourceNode |
|
aSourceMapConsumer.sources.forEach(function (sourceFile) { |
|
var content = aSourceMapConsumer.sourceContentFor(sourceFile); |
|
if (content) { |
|
node.setSourceContent(sourceFile, content); |
|
} |
|
}); |
|
|
|
return node; |
|
|
|
function addMappingWithCode(mapping, code) { |
|
if (mapping === null || mapping.source === undefined) { |
|
node.add(code); |
|
} else { |
|
node.add(new SourceNode(mapping.originalLine, |
|
mapping.originalColumn, |
|
mapping.source, |
|
code, |
|
mapping.name)); |
|
} |
|
} |
|
}; |
|
|
|
/** |
|
* Add a chunk of generated JS to this source node. |
|
* |
|
* @param aChunk A string snippet of generated JS code, another instance of |
|
* SourceNode, or an array where each member is one of those things. |
|
*/ |
|
SourceNode.prototype.add = function SourceNode_add(aChunk) { |
|
if (Array.isArray(aChunk)) { |
|
aChunk.forEach(function (chunk) { |
|
this.add(chunk); |
|
}, this); |
|
} |
|
else if (aChunk instanceof SourceNode || typeof aChunk === "string") { |
|
if (aChunk) { |
|
this.children.push(aChunk); |
|
} |
|
} |
|
else { |
|
throw new TypeError( |
|
"Expected a SourceNode, string, or an array of SourceNodes and strings. Got " + aChunk |
|
); |
|
} |
|
return this; |
|
}; |
|
|
|
/** |
|
* Add a chunk of generated JS to the beginning of this source node. |
|
* |
|
* @param aChunk A string snippet of generated JS code, another instance of |
|
* SourceNode, or an array where each member is one of those things. |
|
*/ |
|
SourceNode.prototype.prepend = function SourceNode_prepend(aChunk) { |
|
if (Array.isArray(aChunk)) { |
|
for (var i = aChunk.length-1; i >= 0; i--) { |
|
this.prepend(aChunk[i]); |
|
} |
|
} |
|
else if (aChunk instanceof SourceNode || typeof aChunk === "string") { |
|
this.children.unshift(aChunk); |
|
} |
|
else { |
|
throw new TypeError( |
|
"Expected a SourceNode, string, or an array of SourceNodes and strings. Got " + aChunk |
|
); |
|
} |
|
return this; |
|
}; |
|
|
|
/** |
|
* Walk over the tree of JS snippets in this node and its children. The |
|
* walking function is called once for each snippet of JS and is passed that |
|
* snippet and the its original associated source's line/column location. |
|
* |
|
* @param aFn The traversal function. |
|
*/ |
|
SourceNode.prototype.walk = function SourceNode_walk(aFn) { |
|
var chunk; |
|
for (var i = 0, len = this.children.length; i < len; i++) { |
|
chunk = this.children[i]; |
|
if (chunk instanceof SourceNode) { |
|
chunk.walk(aFn); |
|
} |
|
else { |
|
if (chunk !== '') { |
|
aFn(chunk, { source: this.source, |
|
line: this.line, |
|
column: this.column, |
|
name: this.name }); |
|
} |
|
} |
|
} |
|
}; |
|
|
|
/** |
|
* Like `String.prototype.join` except for SourceNodes. Inserts `aStr` between |
|
* each of `this.children`. |
|
* |
|
* @param aSep The separator. |
|
*/ |
|
SourceNode.prototype.join = function SourceNode_join(aSep) { |
|
var newChildren; |
|
var i; |
|
var len = this.children.length; |
|
if (len > 0) { |
|
newChildren = []; |
|
for (i = 0; i < len-1; i++) { |
|
newChildren.push(this.children[i]); |
|
newChildren.push(aSep); |
|
} |
|
newChildren.push(this.children[i]); |
|
this.children = newChildren; |
|
} |
|
return this; |
|
}; |
|
|
|
/** |
|
* Call String.prototype.replace on the very right-most source snippet. Useful |
|
* for trimming whitespace from the end of a source node, etc. |
|
* |
|
* @param aPattern The pattern to replace. |
|
* @param aReplacement The thing to replace the pattern with. |
|
*/ |
|
SourceNode.prototype.replaceRight = function SourceNode_replaceRight(aPattern, aReplacement) { |
|
var lastChild = this.children[this.children.length - 1]; |
|
if (lastChild instanceof SourceNode) { |
|
lastChild.replaceRight(aPattern, aReplacement); |
|
} |
|
else if (typeof lastChild === 'string') { |
|
this.children[this.children.length - 1] = lastChild.replace(aPattern, aReplacement); |
|
} |
|
else { |
|
this.children.push(''.replace(aPattern, aReplacement)); |
|
} |
|
return this; |
|
}; |
|
|
|
/** |
|
* Set the source content for a source file. This will be added to the SourceMapGenerator |
|
* in the sourcesContent field. |
|
* |
|
* @param aSourceFile The filename of the source file |
|
* @param aSourceContent The content of the source file |
|
*/ |
|
SourceNode.prototype.setSourceContent = |
|
function SourceNode_setSourceContent(aSourceFile, aSourceContent) { |
|
this.sourceContents[util.toSetString(aSourceFile)] = aSourceContent; |
|
}; |
|
|
|
/** |
|
* Walk over the tree of SourceNodes. The walking function is called for each |
|
* source file content and is passed the filename and source content. |
|
* |
|
* @param aFn The traversal function. |
|
*/ |
|
SourceNode.prototype.walkSourceContents = |
|
function SourceNode_walkSourceContents(aFn) { |
|
for (var i = 0, len = this.children.length; i < len; i++) { |
|
if (this.children[i] instanceof SourceNode) { |
|
this.children[i].walkSourceContents(aFn); |
|
} |
|
} |
|
|
|
var sources = Object.keys(this.sourceContents); |
|
for (var i = 0, len = sources.length; i < len; i++) { |
|
aFn(util.fromSetString(sources[i]), this.sourceContents[sources[i]]); |
|
} |
|
}; |
|
|
|
/** |
|
* Return the string representation of this source node. Walks over the tree |
|
* and concatenates all the various snippets together to one string. |
|
*/ |
|
SourceNode.prototype.toString = function SourceNode_toString() { |
|
var str = ""; |
|
this.walk(function (chunk) { |
|
str += chunk; |
|
}); |
|
return str; |
|
}; |
|
|
|
/** |
|
* Returns the string representation of this source node along with a source |
|
* map. |
|
*/ |
|
SourceNode.prototype.toStringWithSourceMap = function SourceNode_toStringWithSourceMap(aArgs) { |
|
var generated = { |
|
code: "", |
|
line: 1, |
|
column: 0 |
|
}; |
|
var map = new SourceMapGenerator(aArgs); |
|
var sourceMappingActive = false; |
|
var lastOriginalSource = null; |
|
var lastOriginalLine = null; |
|
var lastOriginalColumn = null; |
|
var lastOriginalName = null; |
|
this.walk(function (chunk, original) { |
|
generated.code += chunk; |
|
if (original.source !== null |
|
&& original.line !== null |
|
&& original.column !== null) { |
|
if(lastOriginalSource !== original.source |
|
|| lastOriginalLine !== original.line |
|
|| lastOriginalColumn !== original.column |
|
|| lastOriginalName !== original.name) { |
|
map.addMapping({ |
|
source: original.source, |
|
original: { |
|
line: original.line, |
|
column: original.column |
|
}, |
|
generated: { |
|
line: generated.line, |
|
column: generated.column |
|
}, |
|
name: original.name |
|
}); |
|
} |
|
lastOriginalSource = original.source; |
|
lastOriginalLine = original.line; |
|
lastOriginalColumn = original.column; |
|
lastOriginalName = original.name; |
|
sourceMappingActive = true; |
|
} else if (sourceMappingActive) { |
|
map.addMapping({ |
|
generated: { |
|
line: generated.line, |
|
column: generated.column |
|
} |
|
}); |
|
lastOriginalSource = null; |
|
sourceMappingActive = false; |
|
} |
|
chunk.split('').forEach(function (ch) { |
|
if (ch === '\n') { |
|
generated.line++; |
|
generated.column = 0; |
|
} else { |
|
generated.column++; |
|
} |
|
}); |
|
}); |
|
this.walkSourceContents(function (sourceFile, sourceContent) { |
|
map.setSourceContent(sourceFile, sourceContent); |
|
}); |
|
|
|
return { code: generated.code, map: map }; |
|
}; |
|
|
|
exports.SourceNode = SourceNode; |
|
|
|
}); |
|
|
|
},{"./source-map-generator":17,"./util":19,"amdefine":20}],19:[function(_dereq_,module,exports){ |
|
/* -*- Mode: js; js-indent-level: 2; -*- */ |
|
/* |
|
* Copyright 2011 Mozilla Foundation and contributors |
|
* Licensed under the New BSD license. See LICENSE or: |
|
* http://opensource.org/licenses/BSD-3-Clause |
|
*/ |
|
if (typeof define !== 'function') { |
|
var define = _dereq_('amdefine')(module, _dereq_); |
|
} |
|
define(function (_dereq_, exports, module) { |
|
|
|
/** |
|
* This is a helper function for getting values from parameter/options |
|
* objects. |
|
* |
|
* @param args The object we are extracting values from |
|
* @param name The name of the property we are getting. |
|
* @param defaultValue An optional value to return if the property is missing |
|
* from the object. If this is not specified and the property is missing, an |
|
* error will be thrown. |
|
*/ |
|
function getArg(aArgs, aName, aDefaultValue) { |
|
if (aName in aArgs) { |
|
return aArgs[aName]; |
|
} else if (arguments.length === 3) { |
|
return aDefaultValue; |
|
} else { |
|
throw new Error('"' + aName + '" is a required argument.'); |
|
} |
|
} |
|
exports.getArg = getArg; |
|
|
|
var urlRegexp = /([\w+\-.]+):\/\/((\w+:\w+)@)?([\w.]+)?(:(\d+))?(\S+)?/; |
|
var dataUrlRegexp = /^data:.+\,.+/; |
|
|
|
function urlParse(aUrl) { |
|
var match = aUrl.match(urlRegexp); |
|
if (!match) { |
|
return null; |
|
} |
|
return { |
|
scheme: match[1], |
|
auth: match[3], |
|
host: match[4], |
|
port: match[6], |
|
path: match[7] |
|
}; |
|
} |
|
exports.urlParse = urlParse; |
|
|
|
function urlGenerate(aParsedUrl) { |
|
var url = aParsedUrl.scheme + "://"; |
|
if (aParsedUrl.auth) { |
|
url += aParsedUrl.auth + "@" |
|
} |
|
if (aParsedUrl.host) { |
|
url += aParsedUrl.host; |
|
} |
|
if (aParsedUrl.port) { |
|
url += ":" + aParsedUrl.port |
|
} |
|
if (aParsedUrl.path) { |
|
url += aParsedUrl.path; |
|
} |
|
return url; |
|
} |
|
exports.urlGenerate = urlGenerate; |
|
|
|
function join(aRoot, aPath) { |
|
var url; |
|
|
|
if (aPath.match(urlRegexp) || aPath.match(dataUrlRegexp)) { |
|
return aPath; |
|
} |
|
|
|
if (aPath.charAt(0) === '/' && (url = urlParse(aRoot))) { |
|
url.path = aPath; |
|
return urlGenerate(url); |
|
} |
|
|
|
return aRoot.replace(/\/$/, '') + '/' + aPath; |
|
} |
|
exports.join = join; |
|
|
|
/** |
|
* Because behavior goes wacky when you set `__proto__` on objects, we |
|
* have to prefix all the strings in our set with an arbitrary character. |
|
* |
|
* See https://github.com/mozilla/source-map/pull/31 and |
|
* https://github.com/mozilla/source-map/issues/30 |
|
* |
|
* @param String aStr |
|
*/ |
|
function toSetString(aStr) { |
|
return '$' + aStr; |
|
} |
|
exports.toSetString = toSetString; |
|
|
|
function fromSetString(aStr) { |
|
return aStr.substr(1); |
|
} |
|
exports.fromSetString = fromSetString; |
|
|
|
function relative(aRoot, aPath) { |
|
aRoot = aRoot.replace(/\/$/, ''); |
|
|
|
var url = urlParse(aRoot); |
|
if (aPath.charAt(0) == "/" && url && url.path == "/") { |
|
return aPath.slice(1); |
|
} |
|
|
|
return aPath.indexOf(aRoot + '/') === 0 |
|
? aPath.substr(aRoot.length + 1) |
|
: aPath; |
|
} |
|
exports.relative = relative; |
|
|
|
function strcmp(aStr1, aStr2) { |
|
var s1 = aStr1 || ""; |
|
var s2 = aStr2 || ""; |
|
return (s1 > s2) - (s1 < s2); |
|
} |
|
|
|
/** |
|
* Comparator between two mappings where the original positions are compared. |
|
* |
|
* Optionally pass in `true` as `onlyCompareGenerated` to consider two |
|
* mappings with the same original source/line/column, but different generated |
|
* line and column the same. Useful when searching for a mapping with a |
|
* stubbed out mapping. |
|
*/ |
|
function compareByOriginalPositions(mappingA, mappingB, onlyCompareOriginal) { |
|
var cmp; |
|
|
|
cmp = strcmp(mappingA.source, mappingB.source); |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
cmp = mappingA.originalLine - mappingB.originalLine; |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
cmp = mappingA.originalColumn - mappingB.originalColumn; |
|
if (cmp || onlyCompareOriginal) { |
|
return cmp; |
|
} |
|
|
|
cmp = strcmp(mappingA.name, mappingB.name); |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
cmp = mappingA.generatedLine - mappingB.generatedLine; |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
return mappingA.generatedColumn - mappingB.generatedColumn; |
|
}; |
|
exports.compareByOriginalPositions = compareByOriginalPositions; |
|
|
|
/** |
|
* Comparator between two mappings where the generated positions are |
|
* compared. |
|
* |
|
* Optionally pass in `true` as `onlyCompareGenerated` to consider two |
|
* mappings with the same generated line and column, but different |
|
* source/name/original line and column the same. Useful when searching for a |
|
* mapping with a stubbed out mapping. |
|
*/ |
|
function compareByGeneratedPositions(mappingA, mappingB, onlyCompareGenerated) { |
|
var cmp; |
|
|
|
cmp = mappingA.generatedLine - mappingB.generatedLine; |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
cmp = mappingA.generatedColumn - mappingB.generatedColumn; |
|
if (cmp || onlyCompareGenerated) { |
|
return cmp; |
|
} |
|
|
|
cmp = strcmp(mappingA.source, mappingB.source); |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
cmp = mappingA.originalLine - mappingB.originalLine; |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
cmp = mappingA.originalColumn - mappingB.originalColumn; |
|
if (cmp) { |
|
return cmp; |
|
} |
|
|
|
return strcmp(mappingA.name, mappingB.name); |
|
}; |
|
exports.compareByGeneratedPositions = compareByGeneratedPositions; |
|
|
|
}); |
|
|
|
},{"amdefine":20}],20:[function(_dereq_,module,exports){ |
|
(function (process,__filename){ |
|
/** vim: et:ts=4:sw=4:sts=4 |
|
* @license amdefine 0.1.0 Copyright (c) 2011, The Dojo Foundation All Rights Reserved. |
|
* Available via the MIT or new BSD license. |
|
* see: http://github.com/jrburke/amdefine for details |
|
*/ |
|
|
|
/*jslint node: true */ |
|
/*global module, process */ |
|
'use strict'; |
|
|
|
/** |
|
* Creates a define for node. |
|
* @param {Object} module the "module" object that is defined by Node for the |
|
* current module. |
|
* @param {Function} [requireFn]. Node's require function for the current module. |
|
* It only needs to be passed in Node versions before 0.5, when module.require |
|
* did not exist. |
|
* @returns {Function} a define function that is usable for the current node |
|
* module. |
|
*/ |
|
function amdefine(module, requireFn) { |
|
'use strict'; |
|
var defineCache = {}, |
|
loaderCache = {}, |
|
alreadyCalled = false, |
|
path = _dereq_('path'), |
|
makeRequire, stringRequire; |
|
|
|
/** |
|
* Trims the . and .. from an array of path segments. |
|
* It will keep a leading path segment if a .. will become |
|
* the first path segment, to help with module name lookups, |
|
* which act like paths, but can be remapped. But the end result, |
|
* all paths that use this function should look normalized. |
|
* NOTE: this method MODIFIES the input array. |
|
* @param {Array} ary the array of path segments. |
|
*/ |
|
function trimDots(ary) { |
|
var i, part; |
|
for (i = 0; ary[i]; i+= 1) { |
|
part = ary[i]; |
|
if (part === '.') { |
|
ary.splice(i, 1); |
|
i -= 1; |
|
} else if (part === '..') { |
|
if (i === 1 && (ary[2] === '..' || ary[0] === '..')) { |
|
//End of the line. Keep at least one non-dot |
|
//path segment at the front so it can be mapped |
|
//correctly to disk. Otherwise, there is likely |
|
//no path mapping for a path starting with '..'. |
|
//This can still fail, but catches the most reasonable |
|
//uses of .. |
|
break; |
|
} else if (i > 0) { |
|
ary.splice(i - 1, 2); |
|
i -= 2; |
|
} |
|
} |
|
} |
|
} |
|
|
|
function normalize(name, baseName) { |
|
var baseParts; |
|
|
|
//Adjust any relative paths. |
|
if (name && name.charAt(0) === '.') { |
|
//If have a base name, try to normalize against it, |
|
//otherwise, assume it is a top-level require that will |
|
//be relative to baseUrl in the end. |
|
if (baseName) { |
|
baseParts = baseName.split('/'); |
|
baseParts = baseParts.slice(0, baseParts.length - 1); |
|
baseParts = baseParts.concat(name.split('/')); |
|
trimDots(baseParts); |
|
name = baseParts.join('/'); |
|
} |
|
} |
|
|
|
return name; |
|
} |
|
|
|
/** |
|
* Create the normalize() function passed to a loader plugin's |
|
* normalize method. |
|
*/ |
|
function makeNormalize(relName) { |
|
return function (name) { |
|
return normalize(name, relName); |
|
}; |
|
} |
|
|
|
function makeLoad(id) { |
|
function load(value) { |
|
loaderCache[id] = value; |
|
} |
|
|
|
load.fromText = function (id, text) { |
|
//This one is difficult because the text can/probably uses |
|
//define, and any relative paths and requires should be relative |
|
//to that id was it would be found on disk. But this would require |
|
//bootstrapping a module/require fairly deeply from node core. |
|
//Not sure how best to go about that yet. |
|
throw new Error('amdefine does not implement load.fromText'); |
|
}; |
|
|
|
return load; |
|
} |
|
|
|
makeRequire = function (systemRequire, exports, module, relId) { |
|
function amdRequire(deps, callback) { |
|
if (typeof deps === 'string') { |
|
//Synchronous, single module require('') |
|
return stringRequire(systemRequire, exports, module, deps, relId); |
|
} else { |
|
//Array of dependencies with a callback. |
|
|
|
//Convert the dependencies to modules. |
|
deps = deps.map(function (depName) { |
|
return stringRequire(systemRequire, exports, module, depName, relId); |
|
}); |
|
|
|
//Wait for next tick to call back the require call. |
|
process.nextTick(function () { |
|
callback.apply(null, deps); |
|
}); |
|
} |
|
} |
|
|
|
amdRequire.toUrl = function (filePath) { |
|
if (filePath.indexOf('.') === 0) { |
|
return normalize(filePath, path.dirname(module.filename)); |
|
} else { |
|
return filePath; |
|
} |
|
}; |
|
|
|
return amdRequire; |
|
}; |
|
|
|
//Favor explicit value, passed in if the module wants to support Node 0.4. |
|
requireFn = requireFn || function req() { |
|
return module.require.apply(module, arguments); |
|
}; |
|
|
|
function runFactory(id, deps, factory) { |
|
var r, e, m, result; |
|
|
|
if (id) { |
|
e = loaderCache[id] = {}; |
|
m = { |
|
id: id, |
|
uri: __filename, |
|
exports: e |
|
}; |
|
r = makeRequire(requireFn, e, m, id); |
|
} else { |
|
//Only support one define call per file |
|
if (alreadyCalled) { |
|
throw new Error('amdefine with no module ID cannot be called more than once per file.'); |
|
} |
|
alreadyCalled = true; |
|
|
|
//Use the real variables from node |
|
//Use module.exports for exports, since |
|
//the exports in here is amdefine exports. |
|
e = module.exports; |
|
m = module; |
|
r = makeRequire(requireFn, e, m, module.id); |
|
} |
|
|
|
//If there are dependencies, they are strings, so need |
|
//to convert them to dependency values. |
|
if (deps) { |
|
deps = deps.map(function (depName) { |
|
return r(depName); |
|
}); |
|
} |
|
|
|
//Call the factory with the right dependencies. |
|
if (typeof factory === 'function') { |
|
result = factory.apply(m.exports, deps); |
|
} else { |
|
result = factory; |
|
} |
|
|
|
if (result !== undefined) { |
|
m.exports = result; |
|
if (id) { |
|
loaderCache[id] = m.exports; |
|
} |
|
} |
|
} |
|
|
|
stringRequire = function (systemRequire, exports, module, id, relId) { |
|
//Split the ID by a ! so that |
|
var index = id.indexOf('!'), |
|
originalId = id, |
|
prefix, plugin; |
|
|
|
if (index === -1) { |
|
id = normalize(id, relId); |
|
|
|
//Straight module lookup. If it is one of the special dependencies, |
|
//deal with it, otherwise, delegate to node. |
|
if (id === 'require') { |
|
return makeRequire(systemRequire, exports, module, relId); |
|
} else if (id === 'exports') { |
|
return exports; |
|
} else if (id === 'module') { |
|
return module; |
|
} else if (loaderCache.hasOwnProperty(id)) { |
|
return loaderCache[id]; |
|
} else if (defineCache[id]) { |
|
runFactory.apply(null, defineCache[id]); |
|
return loaderCache[id]; |
|
} else { |
|
if(systemRequire) { |
|
return systemRequire(originalId); |
|
} else { |
|
throw new Error('No module with ID: ' + id); |
|
} |
|
} |
|
} else { |
|
//There is a plugin in play. |
|
prefix = id.substring(0, index); |
|
id = id.substring(index + 1, id.length); |
|
|
|
plugin = stringRequire(systemRequire, exports, module, prefix, relId); |
|
|
|
if (plugin.normalize) { |
|
id = plugin.normalize(id, makeNormalize(relId)); |
|
} else { |
|
//Normalize the ID normally. |
|
id = normalize(id, relId); |
|
} |
|
|
|
if (loaderCache[id]) { |
|
return loaderCache[id]; |
|
} else { |
|
plugin.load(id, makeRequire(systemRequire, exports, module, relId), makeLoad(id), {}); |
|
|
|
return loaderCache[id]; |
|
} |
|
} |
|
}; |
|
|
|
//Create a define function specific to the module asking for amdefine. |
|
function define(id, deps, factory) { |
|
if (Array.isArray(id)) { |
|
factory = deps; |
|
deps = id; |
|
id = undefined; |
|
} else if (typeof id !== 'string') { |
|
factory = id; |
|
id = deps = undefined; |
|
} |
|
|
|
if (deps && !Array.isArray(deps)) { |
|
factory = deps; |
|
deps = undefined; |
|
} |
|
|
|
if (!deps) { |
|
deps = ['require', 'exports', 'module']; |
|
} |
|
|
|
//Set up properties for this module. If an ID, then use |
|
//internal cache. If no ID, then use the external variables |
|
//for this node module. |
|
if (id) { |
|
//Put the module in deep freeze until there is a |
|
//require call for it. |
|
defineCache[id] = [id, deps, factory]; |
|
} else { |
|
runFactory(id, deps, factory); |
|
} |
|
} |
|
|
|
//define.require, which has access to all the values in the |
|
//cache. Useful for AMD modules that all have IDs in the file, |
|
//but need to finally export a value to node based on one of those |
|
//IDs. |
|
define.require = function (id) { |
|
if (loaderCache[id]) { |
|
return loaderCache[id]; |
|
} |
|
|
|
if (defineCache[id]) { |
|
runFactory.apply(null, defineCache[id]); |
|
return loaderCache[id]; |
|
} |
|
}; |
|
|
|
define.amd = {}; |
|
|
|
return define; |
|
} |
|
|
|
module.exports = amdefine; |
|
|
|
}).call(this,_dereq_('_process'),"/node_modules/jstransform/node_modules/source-map/node_modules/amdefine/amdefine.js") |
|
},{"_process":8,"path":7}],21:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
var docblockRe = /^\s*(\/\*\*(.|\r?\n)*?\*\/)/; |
|
var ltrimRe = /^\s*/; |
|
/** |
|
* @param {String} contents |
|
* @return {String} |
|
*/ |
|
function extract(contents) { |
|
var match = contents.match(docblockRe); |
|
if (match) { |
|
return match[0].replace(ltrimRe, '') || ''; |
|
} |
|
return ''; |
|
} |
|
|
|
|
|
var commentStartRe = /^\/\*\*?/; |
|
var commentEndRe = /\*+\/$/; |
|
var wsRe = /[\t ]+/g; |
|
var stringStartRe = /(\r?\n|^) *\*/g; |
|
var multilineRe = /(?:^|\r?\n) *(@[^\r\n]*?) *\r?\n *([^@\r\n\s][^@\r\n]+?) *\r?\n/g; |
|
var propertyRe = /(?:^|\r?\n) *@(\S+) *([^\r\n]*)/g; |
|
|
|
/** |
|
* @param {String} contents |
|
* @return {Array} |
|
*/ |
|
function parse(docblock) { |
|
docblock = docblock |
|
.replace(commentStartRe, '') |
|
.replace(commentEndRe, '') |
|
.replace(wsRe, ' ') |
|
.replace(stringStartRe, '$1'); |
|
|
|
// Normalize multi-line directives |
|
var prev = ''; |
|
while (prev != docblock) { |
|
prev = docblock; |
|
docblock = docblock.replace(multilineRe, "\n$1 $2\n"); |
|
} |
|
docblock = docblock.trim(); |
|
|
|
var result = []; |
|
var match; |
|
while (match = propertyRe.exec(docblock)) { |
|
result.push([match[1], match[2]]); |
|
} |
|
|
|
return result; |
|
} |
|
|
|
/** |
|
* Same as parse but returns an object of prop: value instead of array of paris |
|
* If a property appers more than once the last one will be returned |
|
* |
|
* @param {String} contents |
|
* @return {Object} |
|
*/ |
|
function parseAsObject(docblock) { |
|
var pairs = parse(docblock); |
|
var result = {}; |
|
for (var i = 0; i < pairs.length; i++) { |
|
result[pairs[i][0]] = pairs[i][1]; |
|
} |
|
return result; |
|
} |
|
|
|
|
|
exports.extract = extract; |
|
exports.parse = parse; |
|
exports.parseAsObject = parseAsObject; |
|
|
|
},{}],22:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
|
|
/*jslint node: true*/ |
|
"use strict"; |
|
|
|
var esprima = _dereq_('esprima-fb'); |
|
var utils = _dereq_('./utils'); |
|
|
|
var getBoundaryNode = utils.getBoundaryNode; |
|
var declareIdentInScope = utils.declareIdentInLocalScope; |
|
var initScopeMetadata = utils.initScopeMetadata; |
|
var Syntax = esprima.Syntax; |
|
|
|
/** |
|
* @param {object} node |
|
* @param {object} parentNode |
|
* @return {boolean} |
|
*/ |
|
function _nodeIsClosureScopeBoundary(node, parentNode) { |
|
if (node.type === Syntax.Program) { |
|
return true; |
|
} |
|
|
|
var parentIsFunction = |
|
parentNode.type === Syntax.FunctionDeclaration |
|
|| parentNode.type === Syntax.FunctionExpression |
|
|| parentNode.type === Syntax.ArrowFunctionExpression; |
|
|
|
var parentIsCurlylessArrowFunc = |
|
parentNode.type === Syntax.ArrowFunctionExpression |
|
&& node === parentNode.body; |
|
|
|
return parentIsFunction |
|
&& (node.type === Syntax.BlockStatement || parentIsCurlylessArrowFunc); |
|
} |
|
|
|
function _nodeIsBlockScopeBoundary(node, parentNode) { |
|
if (node.type === Syntax.Program) { |
|
return false; |
|
} |
|
|
|
return node.type === Syntax.BlockStatement |
|
&& parentNode.type === Syntax.CatchClause; |
|
} |
|
|
|
/** |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function traverse(node, path, state) { |
|
/*jshint -W004*/ |
|
// Create a scope stack entry if this is the first node we've encountered in |
|
// its local scope |
|
var startIndex = null; |
|
var parentNode = path[0]; |
|
if (!Array.isArray(node) && state.localScope.parentNode !== parentNode) { |
|
if (_nodeIsClosureScopeBoundary(node, parentNode)) { |
|
var scopeIsStrict = state.scopeIsStrict; |
|
if (!scopeIsStrict |
|
&& (node.type === Syntax.BlockStatement |
|
|| node.type === Syntax.Program)) { |
|
scopeIsStrict = |
|
node.body.length > 0 |
|
&& node.body[0].type === Syntax.ExpressionStatement |
|
&& node.body[0].expression.type === Syntax.Literal |
|
&& node.body[0].expression.value === 'use strict'; |
|
} |
|
|
|
if (node.type === Syntax.Program) { |
|
startIndex = state.g.buffer.length; |
|
state = utils.updateState(state, { |
|
scopeIsStrict: scopeIsStrict |
|
}); |
|
} else { |
|
startIndex = state.g.buffer.length + 1; |
|
state = utils.updateState(state, { |
|
localScope: { |
|
parentNode: parentNode, |
|
parentScope: state.localScope, |
|
identifiers: {}, |
|
tempVarIndex: 0, |
|
tempVars: [] |
|
}, |
|
scopeIsStrict: scopeIsStrict |
|
}); |
|
|
|
// All functions have an implicit 'arguments' object in scope |
|
declareIdentInScope('arguments', initScopeMetadata(node), state); |
|
|
|
// Include function arg identifiers in the scope boundaries of the |
|
// function |
|
if (parentNode.params.length > 0) { |
|
var param; |
|
var metadata = initScopeMetadata(parentNode, path.slice(1), path[0]); |
|
for (var i = 0; i < parentNode.params.length; i++) { |
|
param = parentNode.params[i]; |
|
if (param.type === Syntax.Identifier) { |
|
declareIdentInScope(param.name, metadata, state); |
|
} |
|
} |
|
} |
|
|
|
// Include rest arg identifiers in the scope boundaries of their |
|
// functions |
|
if (parentNode.rest) { |
|
var metadata = initScopeMetadata( |
|
parentNode, |
|
path.slice(1), |
|
path[0] |
|
); |
|
declareIdentInScope(parentNode.rest.name, metadata, state); |
|
} |
|
|
|
// Named FunctionExpressions scope their name within the body block of |
|
// themselves only |
|
if (parentNode.type === Syntax.FunctionExpression && parentNode.id) { |
|
var metaData = |
|
initScopeMetadata(parentNode, path.parentNodeslice, parentNode); |
|
declareIdentInScope(parentNode.id.name, metaData, state); |
|
} |
|
} |
|
|
|
// Traverse and find all local identifiers in this closure first to |
|
// account for function/variable declaration hoisting |
|
collectClosureIdentsAndTraverse(node, path, state); |
|
} |
|
|
|
if (_nodeIsBlockScopeBoundary(node, parentNode)) { |
|
startIndex = state.g.buffer.length; |
|
state = utils.updateState(state, { |
|
localScope: { |
|
parentNode: parentNode, |
|
parentScope: state.localScope, |
|
identifiers: {}, |
|
tempVarIndex: 0, |
|
tempVars: [] |
|
} |
|
}); |
|
|
|
if (parentNode.type === Syntax.CatchClause) { |
|
var metadata = initScopeMetadata( |
|
parentNode, |
|
path.slice(1), |
|
parentNode |
|
); |
|
declareIdentInScope(parentNode.param.name, metadata, state); |
|
} |
|
collectBlockIdentsAndTraverse(node, path, state); |
|
} |
|
} |
|
|
|
// Only catchup() before and after traversing a child node |
|
function traverser(node, path, state) { |
|
node.range && utils.catchup(node.range[0], state); |
|
traverse(node, path, state); |
|
node.range && utils.catchup(node.range[1], state); |
|
} |
|
|
|
utils.analyzeAndTraverse(walker, traverser, node, path, state); |
|
|
|
// Inject temp variables into the scope. |
|
if (startIndex !== null) { |
|
utils.injectTempVarDeclarations(state, startIndex); |
|
} |
|
} |
|
|
|
function collectClosureIdentsAndTraverse(node, path, state) { |
|
utils.analyzeAndTraverse( |
|
visitLocalClosureIdentifiers, |
|
collectClosureIdentsAndTraverse, |
|
node, |
|
path, |
|
state |
|
); |
|
} |
|
|
|
function collectBlockIdentsAndTraverse(node, path, state) { |
|
utils.analyzeAndTraverse( |
|
visitLocalBlockIdentifiers, |
|
collectBlockIdentsAndTraverse, |
|
node, |
|
path, |
|
state |
|
); |
|
} |
|
|
|
function visitLocalClosureIdentifiers(node, path, state) { |
|
var metaData; |
|
switch (node.type) { |
|
case Syntax.ArrowFunctionExpression: |
|
case Syntax.FunctionExpression: |
|
// Function expressions don't get their names (if there is one) added to |
|
// the closure scope they're defined in |
|
return false; |
|
case Syntax.ClassDeclaration: |
|
case Syntax.ClassExpression: |
|
case Syntax.FunctionDeclaration: |
|
if (node.id) { |
|
metaData = initScopeMetadata(getBoundaryNode(path), path.slice(), node); |
|
declareIdentInScope(node.id.name, metaData, state); |
|
} |
|
return false; |
|
case Syntax.VariableDeclarator: |
|
// Variables have function-local scope |
|
if (path[0].kind === 'var') { |
|
metaData = initScopeMetadata(getBoundaryNode(path), path.slice(), node); |
|
declareIdentInScope(node.id.name, metaData, state); |
|
} |
|
break; |
|
} |
|
} |
|
|
|
function visitLocalBlockIdentifiers(node, path, state) { |
|
// TODO: Support 'let' here...maybe...one day...or something... |
|
if (node.type === Syntax.CatchClause) { |
|
return false; |
|
} |
|
} |
|
|
|
function walker(node, path, state) { |
|
var visitors = state.g.visitors; |
|
for (var i = 0; i < visitors.length; i++) { |
|
if (visitors[i].test(node, path, state)) { |
|
return visitors[i](traverse, node, path, state); |
|
} |
|
} |
|
} |
|
|
|
var _astCache = {}; |
|
|
|
function getAstForSource(source, options) { |
|
if (_astCache[source] && !options.disableAstCache) { |
|
return _astCache[source]; |
|
} |
|
var ast = esprima.parse(source, { |
|
comment: true, |
|
loc: true, |
|
range: true, |
|
sourceType: options.sourceType |
|
}); |
|
if (!options.disableAstCache) { |
|
_astCache[source] = ast; |
|
} |
|
return ast; |
|
} |
|
|
|
/** |
|
* Applies all available transformations to the source |
|
* @param {array} visitors |
|
* @param {string} source |
|
* @param {?object} options |
|
* @return {object} |
|
*/ |
|
function transform(visitors, source, options) { |
|
options = options || {}; |
|
var ast; |
|
try { |
|
ast = getAstForSource(source, options); |
|
} catch (e) { |
|
e.message = 'Parse Error: ' + e.message; |
|
throw e; |
|
} |
|
var state = utils.createState(source, ast, options); |
|
state.g.visitors = visitors; |
|
|
|
if (options.sourceMap) { |
|
var SourceMapGenerator = _dereq_('source-map').SourceMapGenerator; |
|
state.g.sourceMap = new SourceMapGenerator({file: options.filename || 'transformed.js'}); |
|
} |
|
|
|
traverse(ast, [], state); |
|
utils.catchup(source.length, state); |
|
|
|
var ret = {code: state.g.buffer, extra: state.g.extra}; |
|
if (options.sourceMap) { |
|
ret.sourceMap = state.g.sourceMap; |
|
ret.sourceMapFilename = options.filename || 'source.js'; |
|
} |
|
return ret; |
|
} |
|
|
|
exports.transform = transform; |
|
exports.Syntax = Syntax; |
|
|
|
},{"./utils":23,"esprima-fb":9,"source-map":11}],23:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
|
|
/*jslint node: true*/ |
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var leadingIndentRegexp = /(^|\n)( {2}|\t)/g; |
|
var nonWhiteRegexp = /(\S)/g; |
|
|
|
/** |
|
* A `state` object represents the state of the parser. It has "local" and |
|
* "global" parts. Global contains parser position, source, etc. Local contains |
|
* scope based properties like current class name. State should contain all the |
|
* info required for transformation. It's the only mandatory object that is |
|
* being passed to every function in transform chain. |
|
* |
|
* @param {string} source |
|
* @param {object} transformOptions |
|
* @return {object} |
|
*/ |
|
function createState(source, rootNode, transformOptions) { |
|
return { |
|
/** |
|
* A tree representing the current local scope (and its lexical scope chain) |
|
* Useful for tracking identifiers from parent scopes, etc. |
|
* @type {Object} |
|
*/ |
|
localScope: { |
|
parentNode: rootNode, |
|
parentScope: null, |
|
identifiers: {}, |
|
tempVarIndex: 0, |
|
tempVars: [] |
|
}, |
|
/** |
|
* The name (and, if applicable, expression) of the super class |
|
* @type {Object} |
|
*/ |
|
superClass: null, |
|
/** |
|
* The namespace to use when munging identifiers |
|
* @type {String} |
|
*/ |
|
mungeNamespace: '', |
|
/** |
|
* Ref to the node for the current MethodDefinition |
|
* @type {Object} |
|
*/ |
|
methodNode: null, |
|
/** |
|
* Ref to the node for the FunctionExpression of the enclosing |
|
* MethodDefinition |
|
* @type {Object} |
|
*/ |
|
methodFuncNode: null, |
|
/** |
|
* Name of the enclosing class |
|
* @type {String} |
|
*/ |
|
className: null, |
|
/** |
|
* Whether we're currently within a `strict` scope |
|
* @type {Bool} |
|
*/ |
|
scopeIsStrict: null, |
|
/** |
|
* Indentation offset |
|
* @type {Number} |
|
*/ |
|
indentBy: 0, |
|
/** |
|
* Global state (not affected by updateState) |
|
* @type {Object} |
|
*/ |
|
g: { |
|
/** |
|
* A set of general options that transformations can consider while doing |
|
* a transformation: |
|
* |
|
* - minify |
|
* Specifies that transformation steps should do their best to minify |
|
* the output source when possible. This is useful for places where |
|
* minification optimizations are possible with higher-level context |
|
* info than what jsxmin can provide. |
|
* |
|
* For example, the ES6 class transform will minify munged private |
|
* variables if this flag is set. |
|
*/ |
|
opts: transformOptions, |
|
/** |
|
* Current position in the source code |
|
* @type {Number} |
|
*/ |
|
position: 0, |
|
/** |
|
* Auxiliary data to be returned by transforms |
|
* @type {Object} |
|
*/ |
|
extra: {}, |
|
/** |
|
* Buffer containing the result |
|
* @type {String} |
|
*/ |
|
buffer: '', |
|
/** |
|
* Source that is being transformed |
|
* @type {String} |
|
*/ |
|
source: source, |
|
|
|
/** |
|
* Cached parsed docblock (see getDocblock) |
|
* @type {object} |
|
*/ |
|
docblock: null, |
|
|
|
/** |
|
* Whether the thing was used |
|
* @type {Boolean} |
|
*/ |
|
tagNamespaceUsed: false, |
|
|
|
/** |
|
* If using bolt xjs transformation |
|
* @type {Boolean} |
|
*/ |
|
isBolt: undefined, |
|
|
|
/** |
|
* Whether to record source map (expensive) or not |
|
* @type {SourceMapGenerator|null} |
|
*/ |
|
sourceMap: null, |
|
|
|
/** |
|
* Filename of the file being processed. Will be returned as a source |
|
* attribute in the source map |
|
*/ |
|
sourceMapFilename: 'source.js', |
|
|
|
/** |
|
* Only when source map is used: last line in the source for which |
|
* source map was generated |
|
* @type {Number} |
|
*/ |
|
sourceLine: 1, |
|
|
|
/** |
|
* Only when source map is used: last line in the buffer for which |
|
* source map was generated |
|
* @type {Number} |
|
*/ |
|
bufferLine: 1, |
|
|
|
/** |
|
* The top-level Program AST for the original file. |
|
*/ |
|
originalProgramAST: null, |
|
|
|
sourceColumn: 0, |
|
bufferColumn: 0 |
|
} |
|
}; |
|
} |
|
|
|
/** |
|
* Updates a copy of a given state with "update" and returns an updated state. |
|
* |
|
* @param {object} state |
|
* @param {object} update |
|
* @return {object} |
|
*/ |
|
function updateState(state, update) { |
|
var ret = Object.create(state); |
|
Object.keys(update).forEach(function(updatedKey) { |
|
ret[updatedKey] = update[updatedKey]; |
|
}); |
|
return ret; |
|
} |
|
|
|
/** |
|
* Given a state fill the resulting buffer from the original source up to |
|
* the end |
|
* |
|
* @param {number} end |
|
* @param {object} state |
|
* @param {?function} contentTransformer Optional callback to transform newly |
|
* added content. |
|
*/ |
|
function catchup(end, state, contentTransformer) { |
|
if (end < state.g.position) { |
|
// cannot move backwards |
|
return; |
|
} |
|
var source = state.g.source.substring(state.g.position, end); |
|
var transformed = updateIndent(source, state); |
|
if (state.g.sourceMap && transformed) { |
|
// record where we are |
|
state.g.sourceMap.addMapping({ |
|
generated: { line: state.g.bufferLine, column: state.g.bufferColumn }, |
|
original: { line: state.g.sourceLine, column: state.g.sourceColumn }, |
|
source: state.g.sourceMapFilename |
|
}); |
|
|
|
// record line breaks in transformed source |
|
var sourceLines = source.split('\n'); |
|
var transformedLines = transformed.split('\n'); |
|
// Add line break mappings between last known mapping and the end of the |
|
// added piece. So for the code piece |
|
// (foo, bar); |
|
// > var x = 2; |
|
// > var b = 3; |
|
// var c = |
|
// only add lines marked with ">": 2, 3. |
|
for (var i = 1; i < sourceLines.length - 1; i++) { |
|
state.g.sourceMap.addMapping({ |
|
generated: { line: state.g.bufferLine, column: 0 }, |
|
original: { line: state.g.sourceLine, column: 0 }, |
|
source: state.g.sourceMapFilename |
|
}); |
|
state.g.sourceLine++; |
|
state.g.bufferLine++; |
|
} |
|
// offset for the last piece |
|
if (sourceLines.length > 1) { |
|
state.g.sourceLine++; |
|
state.g.bufferLine++; |
|
state.g.sourceColumn = 0; |
|
state.g.bufferColumn = 0; |
|
} |
|
state.g.sourceColumn += sourceLines[sourceLines.length - 1].length; |
|
state.g.bufferColumn += |
|
transformedLines[transformedLines.length - 1].length; |
|
} |
|
state.g.buffer += |
|
contentTransformer ? contentTransformer(transformed) : transformed; |
|
state.g.position = end; |
|
} |
|
|
|
/** |
|
* Returns original source for an AST node. |
|
* @param {object} node |
|
* @param {object} state |
|
* @return {string} |
|
*/ |
|
function getNodeSourceText(node, state) { |
|
return state.g.source.substring(node.range[0], node.range[1]); |
|
} |
|
|
|
function _replaceNonWhite(value) { |
|
return value.replace(nonWhiteRegexp, ' '); |
|
} |
|
|
|
/** |
|
* Removes all non-whitespace characters |
|
*/ |
|
function _stripNonWhite(value) { |
|
return value.replace(nonWhiteRegexp, ''); |
|
} |
|
|
|
/** |
|
* Finds the position of the next instance of the specified syntactic char in |
|
* the pending source. |
|
* |
|
* NOTE: This will skip instances of the specified char if they sit inside a |
|
* comment body. |
|
* |
|
* NOTE: This function also assumes that the buffer's current position is not |
|
* already within a comment or a string. This is rarely the case since all |
|
* of the buffer-advancement utility methods tend to be used on syntactic |
|
* nodes' range values -- but it's a small gotcha that's worth mentioning. |
|
*/ |
|
function getNextSyntacticCharOffset(char, state) { |
|
var pendingSource = state.g.source.substring(state.g.position); |
|
var pendingSourceLines = pendingSource.split('\n'); |
|
|
|
var charOffset = 0; |
|
var line; |
|
var withinBlockComment = false; |
|
var withinString = false; |
|
lineLoop: while ((line = pendingSourceLines.shift()) !== undefined) { |
|
var lineEndPos = charOffset + line.length; |
|
charLoop: for (; charOffset < lineEndPos; charOffset++) { |
|
var currChar = pendingSource[charOffset]; |
|
if (currChar === '"' || currChar === '\'') { |
|
withinString = !withinString; |
|
continue charLoop; |
|
} else if (withinString) { |
|
continue charLoop; |
|
} else if (charOffset + 1 < lineEndPos) { |
|
var nextTwoChars = currChar + line[charOffset + 1]; |
|
if (nextTwoChars === '//') { |
|
charOffset = lineEndPos + 1; |
|
continue lineLoop; |
|
} else if (nextTwoChars === '/*') { |
|
withinBlockComment = true; |
|
charOffset += 1; |
|
continue charLoop; |
|
} else if (nextTwoChars === '*/') { |
|
withinBlockComment = false; |
|
charOffset += 1; |
|
continue charLoop; |
|
} |
|
} |
|
|
|
if (!withinBlockComment && currChar === char) { |
|
return charOffset + state.g.position; |
|
} |
|
} |
|
|
|
// Account for '\n' |
|
charOffset++; |
|
withinString = false; |
|
} |
|
|
|
throw new Error('`' + char + '` not found!'); |
|
} |
|
|
|
/** |
|
* Catches up as `catchup` but replaces non-whitespace chars with spaces. |
|
*/ |
|
function catchupWhiteOut(end, state) { |
|
catchup(end, state, _replaceNonWhite); |
|
} |
|
|
|
/** |
|
* Catches up as `catchup` but removes all non-whitespace characters. |
|
*/ |
|
function catchupWhiteSpace(end, state) { |
|
catchup(end, state, _stripNonWhite); |
|
} |
|
|
|
/** |
|
* Removes all non-newline characters |
|
*/ |
|
var reNonNewline = /[^\n]/g; |
|
function stripNonNewline(value) { |
|
return value.replace(reNonNewline, function() { |
|
return ''; |
|
}); |
|
} |
|
|
|
/** |
|
* Catches up as `catchup` but removes all non-newline characters. |
|
* |
|
* Equivalent to appending as many newlines as there are in the original source |
|
* between the current position and `end`. |
|
*/ |
|
function catchupNewlines(end, state) { |
|
catchup(end, state, stripNonNewline); |
|
} |
|
|
|
|
|
/** |
|
* Same as catchup but does not touch the buffer |
|
* |
|
* @param {number} end |
|
* @param {object} state |
|
*/ |
|
function move(end, state) { |
|
// move the internal cursors |
|
if (state.g.sourceMap) { |
|
if (end < state.g.position) { |
|
state.g.position = 0; |
|
state.g.sourceLine = 1; |
|
state.g.sourceColumn = 0; |
|
} |
|
|
|
var source = state.g.source.substring(state.g.position, end); |
|
var sourceLines = source.split('\n'); |
|
if (sourceLines.length > 1) { |
|
state.g.sourceLine += sourceLines.length - 1; |
|
state.g.sourceColumn = 0; |
|
} |
|
state.g.sourceColumn += sourceLines[sourceLines.length - 1].length; |
|
} |
|
state.g.position = end; |
|
} |
|
|
|
/** |
|
* Appends a string of text to the buffer |
|
* |
|
* @param {string} str |
|
* @param {object} state |
|
*/ |
|
function append(str, state) { |
|
if (state.g.sourceMap && str) { |
|
state.g.sourceMap.addMapping({ |
|
generated: { line: state.g.bufferLine, column: state.g.bufferColumn }, |
|
original: { line: state.g.sourceLine, column: state.g.sourceColumn }, |
|
source: state.g.sourceMapFilename |
|
}); |
|
var transformedLines = str.split('\n'); |
|
if (transformedLines.length > 1) { |
|
state.g.bufferLine += transformedLines.length - 1; |
|
state.g.bufferColumn = 0; |
|
} |
|
state.g.bufferColumn += |
|
transformedLines[transformedLines.length - 1].length; |
|
} |
|
state.g.buffer += str; |
|
} |
|
|
|
/** |
|
* Update indent using state.indentBy property. Indent is measured in |
|
* double spaces. Updates a single line only. |
|
* |
|
* @param {string} str |
|
* @param {object} state |
|
* @return {string} |
|
*/ |
|
function updateIndent(str, state) { |
|
/*jshint -W004*/ |
|
var indentBy = state.indentBy; |
|
if (indentBy < 0) { |
|
for (var i = 0; i < -indentBy; i++) { |
|
str = str.replace(leadingIndentRegexp, '$1'); |
|
} |
|
} else { |
|
for (var i = 0; i < indentBy; i++) { |
|
str = str.replace(leadingIndentRegexp, '$1$2$2'); |
|
} |
|
} |
|
return str; |
|
} |
|
|
|
/** |
|
* Calculates indent from the beginning of the line until "start" or the first |
|
* character before start. |
|
* @example |
|
* " foo.bar()" |
|
* ^ |
|
* start |
|
* indent will be " " |
|
* |
|
* @param {number} start |
|
* @param {object} state |
|
* @return {string} |
|
*/ |
|
function indentBefore(start, state) { |
|
var end = start; |
|
start = start - 1; |
|
|
|
while (start > 0 && state.g.source[start] != '\n') { |
|
if (!state.g.source[start].match(/[ \t]/)) { |
|
end = start; |
|
} |
|
start--; |
|
} |
|
return state.g.source.substring(start + 1, end); |
|
} |
|
|
|
function getDocblock(state) { |
|
if (!state.g.docblock) { |
|
var docblock = _dereq_('./docblock'); |
|
state.g.docblock = |
|
docblock.parseAsObject(docblock.extract(state.g.source)); |
|
} |
|
return state.g.docblock; |
|
} |
|
|
|
function identWithinLexicalScope(identName, state, stopBeforeNode) { |
|
var currScope = state.localScope; |
|
while (currScope) { |
|
if (currScope.identifiers[identName] !== undefined) { |
|
return true; |
|
} |
|
|
|
if (stopBeforeNode && currScope.parentNode === stopBeforeNode) { |
|
break; |
|
} |
|
|
|
currScope = currScope.parentScope; |
|
} |
|
return false; |
|
} |
|
|
|
function identInLocalScope(identName, state) { |
|
return state.localScope.identifiers[identName] !== undefined; |
|
} |
|
|
|
/** |
|
* @param {object} boundaryNode |
|
* @param {?array} path |
|
* @return {?object} node |
|
*/ |
|
function initScopeMetadata(boundaryNode, path, node) { |
|
return { |
|
boundaryNode: boundaryNode, |
|
bindingPath: path, |
|
bindingNode: node |
|
}; |
|
} |
|
|
|
function declareIdentInLocalScope(identName, metaData, state) { |
|
state.localScope.identifiers[identName] = { |
|
boundaryNode: metaData.boundaryNode, |
|
path: metaData.bindingPath, |
|
node: metaData.bindingNode, |
|
state: Object.create(state) |
|
}; |
|
} |
|
|
|
function getLexicalBindingMetadata(identName, state) { |
|
var currScope = state.localScope; |
|
while (currScope) { |
|
if (currScope.identifiers[identName] !== undefined) { |
|
return currScope.identifiers[identName]; |
|
} |
|
|
|
currScope = currScope.parentScope; |
|
} |
|
} |
|
|
|
function getLocalBindingMetadata(identName, state) { |
|
return state.localScope.identifiers[identName]; |
|
} |
|
|
|
/** |
|
* Apply the given analyzer function to the current node. If the analyzer |
|
* doesn't return false, traverse each child of the current node using the given |
|
* traverser function. |
|
* |
|
* @param {function} analyzer |
|
* @param {function} traverser |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function analyzeAndTraverse(analyzer, traverser, node, path, state) { |
|
if (node.type) { |
|
if (analyzer(node, path, state) === false) { |
|
return; |
|
} |
|
path.unshift(node); |
|
} |
|
|
|
getOrderedChildren(node).forEach(function(child) { |
|
traverser(child, path, state); |
|
}); |
|
|
|
node.type && path.shift(); |
|
} |
|
|
|
/** |
|
* It is crucial that we traverse in order, or else catchup() on a later |
|
* node that is processed out of order can move the buffer past a node |
|
* that we haven't handled yet, preventing us from modifying that node. |
|
* |
|
* This can happen when a node has multiple properties containing children. |
|
* For example, XJSElement nodes have `openingElement`, `closingElement` and |
|
* `children`. If we traverse `openingElement`, then `closingElement`, then |
|
* when we get to `children`, the buffer has already caught up to the end of |
|
* the closing element, after the children. |
|
* |
|
* This is basically a Schwartzian transform. Collects an array of children, |
|
* each one represented as [child, startIndex]; sorts the array by start |
|
* index; then traverses the children in that order. |
|
*/ |
|
function getOrderedChildren(node) { |
|
var queue = []; |
|
for (var key in node) { |
|
if (node.hasOwnProperty(key)) { |
|
enqueueNodeWithStartIndex(queue, node[key]); |
|
} |
|
} |
|
queue.sort(function(a, b) { return a[1] - b[1]; }); |
|
return queue.map(function(pair) { return pair[0]; }); |
|
} |
|
|
|
/** |
|
* Helper function for analyzeAndTraverse which queues up all of the children |
|
* of the given node. |
|
* |
|
* Children can also be found in arrays, so we basically want to merge all of |
|
* those arrays together so we can sort them and then traverse the children |
|
* in order. |
|
* |
|
* One example is the Program node. It contains `body` and `comments`, both |
|
* arrays. Lexographically, comments are interspersed throughout the body |
|
* nodes, but esprima's AST groups them together. |
|
*/ |
|
function enqueueNodeWithStartIndex(queue, node) { |
|
if (typeof node !== 'object' || node === null) { |
|
return; |
|
} |
|
if (node.range) { |
|
queue.push([node, node.range[0]]); |
|
} else if (Array.isArray(node)) { |
|
for (var ii = 0; ii < node.length; ii++) { |
|
enqueueNodeWithStartIndex(queue, node[ii]); |
|
} |
|
} |
|
} |
|
|
|
/** |
|
* Checks whether a node or any of its sub-nodes contains |
|
* a syntactic construct of the passed type. |
|
* @param {object} node - AST node to test. |
|
* @param {string} type - node type to lookup. |
|
*/ |
|
function containsChildOfType(node, type) { |
|
return containsChildMatching(node, function(node) { |
|
return node.type === type; |
|
}); |
|
} |
|
|
|
function containsChildMatching(node, matcher) { |
|
var foundMatchingChild = false; |
|
function nodeTypeAnalyzer(node) { |
|
if (matcher(node) === true) { |
|
foundMatchingChild = true; |
|
return false; |
|
} |
|
} |
|
function nodeTypeTraverser(child, path, state) { |
|
if (!foundMatchingChild) { |
|
foundMatchingChild = containsChildMatching(child, matcher); |
|
} |
|
} |
|
analyzeAndTraverse( |
|
nodeTypeAnalyzer, |
|
nodeTypeTraverser, |
|
node, |
|
[] |
|
); |
|
return foundMatchingChild; |
|
} |
|
|
|
var scopeTypes = {}; |
|
scopeTypes[Syntax.ArrowFunctionExpression] = true; |
|
scopeTypes[Syntax.FunctionExpression] = true; |
|
scopeTypes[Syntax.FunctionDeclaration] = true; |
|
scopeTypes[Syntax.Program] = true; |
|
|
|
function getBoundaryNode(path) { |
|
for (var ii = 0; ii < path.length; ++ii) { |
|
if (scopeTypes[path[ii].type]) { |
|
return path[ii]; |
|
} |
|
} |
|
throw new Error( |
|
'Expected to find a node with one of the following types in path:\n' + |
|
JSON.stringify(Object.keys(scopeTypes)) |
|
); |
|
} |
|
|
|
function getTempVar(tempVarIndex) { |
|
return '$__' + tempVarIndex; |
|
} |
|
|
|
function injectTempVar(state) { |
|
var tempVar = '$__' + (state.localScope.tempVarIndex++); |
|
state.localScope.tempVars.push(tempVar); |
|
return tempVar; |
|
} |
|
|
|
function injectTempVarDeclarations(state, index) { |
|
if (state.localScope.tempVars.length) { |
|
state.g.buffer = |
|
state.g.buffer.slice(0, index) + |
|
'var ' + state.localScope.tempVars.join(', ') + ';' + |
|
state.g.buffer.slice(index); |
|
state.localScope.tempVars = []; |
|
} |
|
} |
|
|
|
exports.analyzeAndTraverse = analyzeAndTraverse; |
|
exports.append = append; |
|
exports.catchup = catchup; |
|
exports.catchupNewlines = catchupNewlines; |
|
exports.catchupWhiteOut = catchupWhiteOut; |
|
exports.catchupWhiteSpace = catchupWhiteSpace; |
|
exports.containsChildMatching = containsChildMatching; |
|
exports.containsChildOfType = containsChildOfType; |
|
exports.createState = createState; |
|
exports.declareIdentInLocalScope = declareIdentInLocalScope; |
|
exports.getBoundaryNode = getBoundaryNode; |
|
exports.getDocblock = getDocblock; |
|
exports.getLexicalBindingMetadata = getLexicalBindingMetadata; |
|
exports.getLocalBindingMetadata = getLocalBindingMetadata; |
|
exports.getNextSyntacticCharOffset = getNextSyntacticCharOffset; |
|
exports.getNodeSourceText = getNodeSourceText; |
|
exports.getOrderedChildren = getOrderedChildren; |
|
exports.getTempVar = getTempVar; |
|
exports.identInLocalScope = identInLocalScope; |
|
exports.identWithinLexicalScope = identWithinLexicalScope; |
|
exports.indentBefore = indentBefore; |
|
exports.initScopeMetadata = initScopeMetadata; |
|
exports.injectTempVar = injectTempVar; |
|
exports.injectTempVarDeclarations = injectTempVarDeclarations; |
|
exports.move = move; |
|
exports.scopeTypes = scopeTypes; |
|
exports.updateIndent = updateIndent; |
|
exports.updateState = updateState; |
|
|
|
},{"./docblock":21,"esprima-fb":9}],24:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*global exports:true*/ |
|
|
|
/** |
|
* Desugars ES6 Arrow functions to ES3 function expressions. |
|
* If the function contains `this` expression -- automatically |
|
* binds the function to current value of `this`. |
|
* |
|
* Single parameter, simple expression: |
|
* |
|
* [1, 2, 3].map(x => x * x); |
|
* |
|
* [1, 2, 3].map(function(x) { return x * x; }); |
|
* |
|
* Several parameters, complex block: |
|
* |
|
* this.users.forEach((user, idx) => { |
|
* return this.isActive(idx) && this.send(user); |
|
* }); |
|
* |
|
* this.users.forEach(function(user, idx) { |
|
* return this.isActive(idx) && this.send(user); |
|
* }.bind(this)); |
|
* |
|
*/ |
|
var restParamVisitors = _dereq_('./es6-rest-param-visitors'); |
|
var destructuringVisitors = _dereq_('./es6-destructuring-visitors'); |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
/** |
|
* @public |
|
*/ |
|
function visitArrowFunction(traverse, node, path, state) { |
|
var notInExpression = (path[0].type === Syntax.ExpressionStatement); |
|
|
|
// Wrap a function into a grouping operator, if it's not |
|
// in the expression position. |
|
if (notInExpression) { |
|
utils.append('(', state); |
|
} |
|
|
|
utils.append('function', state); |
|
renderParams(traverse, node, path, state); |
|
|
|
// Skip arrow. |
|
utils.catchupWhiteSpace(node.body.range[0], state); |
|
|
|
var renderBody = node.body.type == Syntax.BlockStatement |
|
? renderStatementBody |
|
: renderExpressionBody; |
|
|
|
path.unshift(node); |
|
renderBody(traverse, node, path, state); |
|
path.shift(); |
|
|
|
// Bind the function only if `this` value is used |
|
// inside it or inside any sub-expression. |
|
var containsBindingSyntax = |
|
utils.containsChildMatching(node.body, function(node) { |
|
return node.type === Syntax.ThisExpression |
|
|| (node.type === Syntax.Identifier |
|
&& node.name === "super"); |
|
}); |
|
|
|
if (containsBindingSyntax) { |
|
utils.append('.bind(this)', state); |
|
} |
|
|
|
utils.catchupWhiteSpace(node.range[1], state); |
|
|
|
// Close wrapper if not in the expression. |
|
if (notInExpression) { |
|
utils.append(')', state); |
|
} |
|
|
|
return false; |
|
} |
|
|
|
function renderParams(traverse, node, path, state) { |
|
// To preserve inline typechecking directives, we |
|
// distinguish between parens-free and paranthesized single param. |
|
if (isParensFreeSingleParam(node, state) || !node.params.length) { |
|
utils.append('(', state); |
|
} |
|
if (node.params.length !== 0) { |
|
path.unshift(node); |
|
traverse(node.params, path, state); |
|
path.unshift(); |
|
} |
|
utils.append(')', state); |
|
} |
|
|
|
function isParensFreeSingleParam(node, state) { |
|
return node.params.length === 1 && |
|
state.g.source[state.g.position] !== '('; |
|
} |
|
|
|
function renderExpressionBody(traverse, node, path, state) { |
|
// Wrap simple expression bodies into a block |
|
// with explicit return statement. |
|
utils.append('{', state); |
|
|
|
// Special handling of rest param. |
|
if (node.rest) { |
|
utils.append( |
|
restParamVisitors.renderRestParamSetup(node, state), |
|
state |
|
); |
|
} |
|
|
|
// Special handling of destructured params. |
|
destructuringVisitors.renderDestructuredComponents( |
|
node, |
|
utils.updateState(state, { |
|
localScope: { |
|
parentNode: state.parentNode, |
|
parentScope: state.parentScope, |
|
identifiers: state.identifiers, |
|
tempVarIndex: 0 |
|
} |
|
}) |
|
); |
|
|
|
utils.append('return ', state); |
|
renderStatementBody(traverse, node, path, state); |
|
utils.append(';}', state); |
|
} |
|
|
|
function renderStatementBody(traverse, node, path, state) { |
|
traverse(node.body, path, state); |
|
utils.catchup(node.body.range[1], state); |
|
} |
|
|
|
visitArrowFunction.test = function(node, path, state) { |
|
return node.type === Syntax.ArrowFunctionExpression; |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitArrowFunction |
|
]; |
|
|
|
|
|
},{"../src/utils":23,"./es6-destructuring-visitors":27,"./es6-rest-param-visitors":30,"esprima-fb":9}],25:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2004-present Facebook. All Rights Reserved. |
|
*/ |
|
/*global exports:true*/ |
|
|
|
/** |
|
* Implements ES6 call spread. |
|
* |
|
* instance.method(a, b, c, ...d) |
|
* |
|
* instance.method.apply(instance, [a, b, c].concat(d)) |
|
* |
|
*/ |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
function process(traverse, node, path, state) { |
|
utils.move(node.range[0], state); |
|
traverse(node, path, state); |
|
utils.catchup(node.range[1], state); |
|
} |
|
|
|
function visitCallSpread(traverse, node, path, state) { |
|
utils.catchup(node.range[0], state); |
|
|
|
if (node.type === Syntax.NewExpression) { |
|
// Input = new Set(1, 2, ...list) |
|
// Output = new (Function.prototype.bind.apply(Set, [null, 1, 2].concat(list))) |
|
utils.append('new (Function.prototype.bind.apply(', state); |
|
process(traverse, node.callee, path, state); |
|
} else if (node.callee.type === Syntax.MemberExpression) { |
|
// Input = get().fn(1, 2, ...more) |
|
// Output = (_ = get()).fn.apply(_, [1, 2].apply(more)) |
|
var tempVar = utils.injectTempVar(state); |
|
utils.append('(' + tempVar + ' = ', state); |
|
process(traverse, node.callee.object, path, state); |
|
utils.append(')', state); |
|
if (node.callee.property.type === Syntax.Identifier) { |
|
utils.append('.', state); |
|
process(traverse, node.callee.property, path, state); |
|
} else { |
|
utils.append('[', state); |
|
process(traverse, node.callee.property, path, state); |
|
utils.append(']', state); |
|
} |
|
utils.append('.apply(' + tempVar, state); |
|
} else { |
|
// Input = max(1, 2, ...list) |
|
// Output = max.apply(null, [1, 2].concat(list)) |
|
var needsToBeWrappedInParenthesis = |
|
node.callee.type === Syntax.FunctionDeclaration || |
|
node.callee.type === Syntax.FunctionExpression; |
|
if (needsToBeWrappedInParenthesis) { |
|
utils.append('(', state); |
|
} |
|
process(traverse, node.callee, path, state); |
|
if (needsToBeWrappedInParenthesis) { |
|
utils.append(')', state); |
|
} |
|
utils.append('.apply(null', state); |
|
} |
|
utils.append(', ', state); |
|
|
|
var args = node.arguments.slice(); |
|
var spread = args.pop(); |
|
if (args.length || node.type === Syntax.NewExpression) { |
|
utils.append('[', state); |
|
if (node.type === Syntax.NewExpression) { |
|
utils.append('null' + (args.length ? ', ' : ''), state); |
|
} |
|
while (args.length) { |
|
var arg = args.shift(); |
|
utils.move(arg.range[0], state); |
|
traverse(arg, path, state); |
|
if (args.length) { |
|
utils.catchup(args[0].range[0], state); |
|
} else { |
|
utils.catchup(arg.range[1], state); |
|
} |
|
} |
|
utils.append('].concat(', state); |
|
process(traverse, spread.argument, path, state); |
|
utils.append(')', state); |
|
} else { |
|
process(traverse, spread.argument, path, state); |
|
} |
|
utils.append(node.type === Syntax.NewExpression ? '))' : ')', state); |
|
|
|
utils.move(node.range[1], state); |
|
return false; |
|
} |
|
|
|
visitCallSpread.test = function(node, path, state) { |
|
return ( |
|
( |
|
node.type === Syntax.CallExpression || |
|
node.type === Syntax.NewExpression |
|
) && |
|
node.arguments.length > 0 && |
|
node.arguments[node.arguments.length - 1].type === Syntax.SpreadElement |
|
); |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitCallSpread |
|
]; |
|
|
|
},{"../src/utils":23,"esprima-fb":9}],26:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*jslint node:true*/ |
|
|
|
/** |
|
* @typechecks |
|
*/ |
|
'use strict'; |
|
|
|
var base62 = _dereq_('base62'); |
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
var reservedWordsHelper = _dereq_('./reserved-words-helper'); |
|
|
|
var declareIdentInLocalScope = utils.declareIdentInLocalScope; |
|
var initScopeMetadata = utils.initScopeMetadata; |
|
|
|
var SUPER_PROTO_IDENT_PREFIX = '____SuperProtoOf'; |
|
|
|
var _anonClassUUIDCounter = 0; |
|
var _mungedSymbolMaps = {}; |
|
|
|
function resetSymbols() { |
|
_anonClassUUIDCounter = 0; |
|
_mungedSymbolMaps = {}; |
|
} |
|
|
|
/** |
|
* Used to generate a unique class for use with code-gens for anonymous class |
|
* expressions. |
|
* |
|
* @param {object} state |
|
* @return {string} |
|
*/ |
|
function _generateAnonymousClassName(state) { |
|
var mungeNamespace = state.mungeNamespace || ''; |
|
return '____Class' + mungeNamespace + base62.encode(_anonClassUUIDCounter++); |
|
} |
|
|
|
/** |
|
* Given an identifier name, munge it using the current state's mungeNamespace. |
|
* |
|
* @param {string} identName |
|
* @param {object} state |
|
* @return {string} |
|
*/ |
|
function _getMungedName(identName, state) { |
|
var mungeNamespace = state.mungeNamespace; |
|
var shouldMinify = state.g.opts.minify; |
|
|
|
if (shouldMinify) { |
|
if (!_mungedSymbolMaps[mungeNamespace]) { |
|
_mungedSymbolMaps[mungeNamespace] = { |
|
symbolMap: {}, |
|
identUUIDCounter: 0 |
|
}; |
|
} |
|
|
|
var symbolMap = _mungedSymbolMaps[mungeNamespace].symbolMap; |
|
if (!symbolMap[identName]) { |
|
symbolMap[identName] = |
|
base62.encode(_mungedSymbolMaps[mungeNamespace].identUUIDCounter++); |
|
} |
|
identName = symbolMap[identName]; |
|
} |
|
return '$' + mungeNamespace + identName; |
|
} |
|
|
|
/** |
|
* Extracts super class information from a class node. |
|
* |
|
* Information includes name of the super class and/or the expression string |
|
* (if extending from an expression) |
|
* |
|
* @param {object} node |
|
* @param {object} state |
|
* @return {object} |
|
*/ |
|
function _getSuperClassInfo(node, state) { |
|
var ret = { |
|
name: null, |
|
expression: null |
|
}; |
|
if (node.superClass) { |
|
if (node.superClass.type === Syntax.Identifier) { |
|
ret.name = node.superClass.name; |
|
} else { |
|
// Extension from an expression |
|
ret.name = _generateAnonymousClassName(state); |
|
ret.expression = state.g.source.substring( |
|
node.superClass.range[0], |
|
node.superClass.range[1] |
|
); |
|
} |
|
} |
|
return ret; |
|
} |
|
|
|
/** |
|
* Used with .filter() to find the constructor method in a list of |
|
* MethodDefinition nodes. |
|
* |
|
* @param {object} classElement |
|
* @return {boolean} |
|
*/ |
|
function _isConstructorMethod(classElement) { |
|
return classElement.type === Syntax.MethodDefinition && |
|
classElement.key.type === Syntax.Identifier && |
|
classElement.key.name === 'constructor'; |
|
} |
|
|
|
/** |
|
* @param {object} node |
|
* @param {object} state |
|
* @return {boolean} |
|
*/ |
|
function _shouldMungeIdentifier(node, state) { |
|
return ( |
|
!!state.methodFuncNode && |
|
!utils.getDocblock(state).hasOwnProperty('preventMunge') && |
|
/^_(?!_)/.test(node.name) |
|
); |
|
} |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitClassMethod(traverse, node, path, state) { |
|
if (!state.g.opts.es5 && (node.kind === 'get' || node.kind === 'set')) { |
|
throw new Error( |
|
'This transform does not support ' + node.kind + 'ter methods for ES6 ' + |
|
'classes. (line: ' + node.loc.start.line + ', col: ' + |
|
node.loc.start.column + ')' |
|
); |
|
} |
|
state = utils.updateState(state, { |
|
methodNode: node |
|
}); |
|
utils.catchup(node.range[0], state); |
|
path.unshift(node); |
|
traverse(node.value, path, state); |
|
path.shift(); |
|
return false; |
|
} |
|
visitClassMethod.test = function(node, path, state) { |
|
return node.type === Syntax.MethodDefinition; |
|
}; |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitClassFunctionExpression(traverse, node, path, state) { |
|
var methodNode = path[0]; |
|
var isGetter = methodNode.kind === 'get'; |
|
var isSetter = methodNode.kind === 'set'; |
|
|
|
state = utils.updateState(state, { |
|
methodFuncNode: node |
|
}); |
|
|
|
if (methodNode.key.name === 'constructor') { |
|
utils.append('function ' + state.className, state); |
|
} else { |
|
var methodAccessorComputed = false; |
|
var methodAccessor; |
|
var prototypeOrStatic = methodNode["static"] ? '' : '.prototype'; |
|
var objectAccessor = state.className + prototypeOrStatic; |
|
|
|
if (methodNode.key.type === Syntax.Identifier) { |
|
// foo() {} |
|
methodAccessor = methodNode.key.name; |
|
if (_shouldMungeIdentifier(methodNode.key, state)) { |
|
methodAccessor = _getMungedName(methodAccessor, state); |
|
} |
|
if (isGetter || isSetter) { |
|
methodAccessor = JSON.stringify(methodAccessor); |
|
} else if (reservedWordsHelper.isReservedWord(methodAccessor)) { |
|
methodAccessorComputed = true; |
|
methodAccessor = JSON.stringify(methodAccessor); |
|
} |
|
} else if (methodNode.key.type === Syntax.Literal) { |
|
// 'foo bar'() {} | get 'foo bar'() {} | set 'foo bar'() {} |
|
methodAccessor = JSON.stringify(methodNode.key.value); |
|
methodAccessorComputed = true; |
|
} |
|
|
|
if (isSetter || isGetter) { |
|
utils.append( |
|
'Object.defineProperty(' + |
|
objectAccessor + ',' + |
|
methodAccessor + ',' + |
|
'{configurable:true,' + |
|
methodNode.kind + ':function', |
|
state |
|
); |
|
} else { |
|
if (state.g.opts.es3) { |
|
if (methodAccessorComputed) { |
|
methodAccessor = '[' + methodAccessor + ']'; |
|
} else { |
|
methodAccessor = '.' + methodAccessor; |
|
} |
|
utils.append( |
|
objectAccessor + |
|
methodAccessor + '=function' + (node.generator ? '*' : ''), |
|
state |
|
); |
|
} else { |
|
if (!methodAccessorComputed) { |
|
methodAccessor = JSON.stringify(methodAccessor); |
|
} |
|
utils.append( |
|
'Object.defineProperty(' + |
|
objectAccessor + ',' + |
|
methodAccessor + ',' + |
|
'{writable:true,configurable:true,' + |
|
'value:function' + (node.generator ? '*' : ''), |
|
state |
|
); |
|
} |
|
} |
|
} |
|
utils.move(methodNode.key.range[1], state); |
|
utils.append('(', state); |
|
|
|
var params = node.params; |
|
if (params.length > 0) { |
|
utils.catchupNewlines(params[0].range[0], state); |
|
for (var i = 0; i < params.length; i++) { |
|
utils.catchup(node.params[i].range[0], state); |
|
path.unshift(node); |
|
traverse(params[i], path, state); |
|
path.shift(); |
|
} |
|
} |
|
|
|
var closingParenPosition = utils.getNextSyntacticCharOffset(')', state); |
|
utils.catchupWhiteSpace(closingParenPosition, state); |
|
|
|
var openingBracketPosition = utils.getNextSyntacticCharOffset('{', state); |
|
utils.catchup(openingBracketPosition + 1, state); |
|
|
|
if (!state.scopeIsStrict) { |
|
utils.append('"use strict";', state); |
|
state = utils.updateState(state, { |
|
scopeIsStrict: true |
|
}); |
|
} |
|
utils.move(node.body.range[0] + '{'.length, state); |
|
|
|
path.unshift(node); |
|
traverse(node.body, path, state); |
|
path.shift(); |
|
utils.catchup(node.body.range[1], state); |
|
|
|
if (methodNode.key.name !== 'constructor') { |
|
if (isGetter || isSetter || !state.g.opts.es3) { |
|
utils.append('})', state); |
|
} |
|
utils.append(';', state); |
|
} |
|
return false; |
|
} |
|
visitClassFunctionExpression.test = function(node, path, state) { |
|
return node.type === Syntax.FunctionExpression |
|
&& path[0].type === Syntax.MethodDefinition; |
|
}; |
|
|
|
function visitClassMethodParam(traverse, node, path, state) { |
|
var paramName = node.name; |
|
if (_shouldMungeIdentifier(node, state)) { |
|
paramName = _getMungedName(node.name, state); |
|
} |
|
utils.append(paramName, state); |
|
utils.move(node.range[1], state); |
|
} |
|
visitClassMethodParam.test = function(node, path, state) { |
|
if (!path[0] || !path[1]) { |
|
return; |
|
} |
|
|
|
var parentFuncExpr = path[0]; |
|
var parentClassMethod = path[1]; |
|
|
|
return parentFuncExpr.type === Syntax.FunctionExpression |
|
&& parentClassMethod.type === Syntax.MethodDefinition |
|
&& node.type === Syntax.Identifier; |
|
}; |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function _renderClassBody(traverse, node, path, state) { |
|
var className = state.className; |
|
var superClass = state.superClass; |
|
|
|
// Set up prototype of constructor on same line as `extends` for line-number |
|
// preservation. This relies on function-hoisting if a constructor function is |
|
// defined in the class body. |
|
if (superClass.name) { |
|
// If the super class is an expression, we need to memoize the output of the |
|
// expression into the generated class name variable and use that to refer |
|
// to the super class going forward. Example: |
|
// |
|
// class Foo extends mixin(Bar, Baz) {} |
|
// --transforms to-- |
|
// function Foo() {} var ____Class0Blah = mixin(Bar, Baz); |
|
if (superClass.expression !== null) { |
|
utils.append( |
|
'var ' + superClass.name + '=' + superClass.expression + ';', |
|
state |
|
); |
|
} |
|
|
|
var keyName = superClass.name + '____Key'; |
|
var keyNameDeclarator = ''; |
|
if (!utils.identWithinLexicalScope(keyName, state)) { |
|
keyNameDeclarator = 'var '; |
|
declareIdentInLocalScope(keyName, initScopeMetadata(node), state); |
|
} |
|
utils.append( |
|
'for(' + keyNameDeclarator + keyName + ' in ' + superClass.name + '){' + |
|
'if(' + superClass.name + '.hasOwnProperty(' + keyName + ')){' + |
|
className + '[' + keyName + ']=' + |
|
superClass.name + '[' + keyName + '];' + |
|
'}' + |
|
'}', |
|
state |
|
); |
|
|
|
var superProtoIdentStr = SUPER_PROTO_IDENT_PREFIX + superClass.name; |
|
if (!utils.identWithinLexicalScope(superProtoIdentStr, state)) { |
|
utils.append( |
|
'var ' + superProtoIdentStr + '=' + superClass.name + '===null?' + |
|
'null:' + superClass.name + '.prototype;', |
|
state |
|
); |
|
declareIdentInLocalScope(superProtoIdentStr, initScopeMetadata(node), state); |
|
} |
|
|
|
utils.append( |
|
className + '.prototype=Object.create(' + superProtoIdentStr + ');', |
|
state |
|
); |
|
utils.append( |
|
className + '.prototype.constructor=' + className + ';', |
|
state |
|
); |
|
utils.append( |
|
className + '.__superConstructor__=' + superClass.name + ';', |
|
state |
|
); |
|
} |
|
|
|
// If there's no constructor method specified in the class body, create an |
|
// empty constructor function at the top (same line as the class keyword) |
|
if (!node.body.body.filter(_isConstructorMethod).pop()) { |
|
utils.append('function ' + className + '(){', state); |
|
if (!state.scopeIsStrict) { |
|
utils.append('"use strict";', state); |
|
} |
|
if (superClass.name) { |
|
utils.append( |
|
'if(' + superClass.name + '!==null){' + |
|
superClass.name + '.apply(this,arguments);}', |
|
state |
|
); |
|
} |
|
utils.append('}', state); |
|
} |
|
|
|
utils.move(node.body.range[0] + '{'.length, state); |
|
traverse(node.body, path, state); |
|
utils.catchupWhiteSpace(node.range[1], state); |
|
} |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitClassDeclaration(traverse, node, path, state) { |
|
var className = node.id.name; |
|
var superClass = _getSuperClassInfo(node, state); |
|
|
|
state = utils.updateState(state, { |
|
mungeNamespace: className, |
|
className: className, |
|
superClass: superClass |
|
}); |
|
|
|
_renderClassBody(traverse, node, path, state); |
|
|
|
return false; |
|
} |
|
visitClassDeclaration.test = function(node, path, state) { |
|
return node.type === Syntax.ClassDeclaration; |
|
}; |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitClassExpression(traverse, node, path, state) { |
|
var className = node.id && node.id.name || _generateAnonymousClassName(state); |
|
var superClass = _getSuperClassInfo(node, state); |
|
|
|
utils.append('(function(){', state); |
|
|
|
state = utils.updateState(state, { |
|
mungeNamespace: className, |
|
className: className, |
|
superClass: superClass |
|
}); |
|
|
|
_renderClassBody(traverse, node, path, state); |
|
|
|
utils.append('return ' + className + ';})()', state); |
|
return false; |
|
} |
|
visitClassExpression.test = function(node, path, state) { |
|
return node.type === Syntax.ClassExpression; |
|
}; |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitPrivateIdentifier(traverse, node, path, state) { |
|
utils.append(_getMungedName(node.name, state), state); |
|
utils.move(node.range[1], state); |
|
} |
|
visitPrivateIdentifier.test = function(node, path, state) { |
|
if (node.type === Syntax.Identifier && _shouldMungeIdentifier(node, state)) { |
|
// Always munge non-computed properties of MemberExpressions |
|
// (a la preventing access of properties of unowned objects) |
|
if (path[0].type === Syntax.MemberExpression && path[0].object !== node |
|
&& path[0].computed === false) { |
|
return true; |
|
} |
|
|
|
// Always munge identifiers that were declared within the method function |
|
// scope |
|
if (utils.identWithinLexicalScope(node.name, state, state.methodFuncNode)) { |
|
return true; |
|
} |
|
|
|
// Always munge private keys on object literals defined within a method's |
|
// scope. |
|
if (path[0].type === Syntax.Property |
|
&& path[1].type === Syntax.ObjectExpression) { |
|
return true; |
|
} |
|
|
|
// Always munge function parameters |
|
if (path[0].type === Syntax.FunctionExpression |
|
|| path[0].type === Syntax.FunctionDeclaration |
|
|| path[0].type === Syntax.ArrowFunctionExpression) { |
|
for (var i = 0; i < path[0].params.length; i++) { |
|
if (path[0].params[i] === node) { |
|
return true; |
|
} |
|
} |
|
} |
|
} |
|
return false; |
|
}; |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitSuperCallExpression(traverse, node, path, state) { |
|
var superClassName = state.superClass.name; |
|
|
|
if (node.callee.type === Syntax.Identifier) { |
|
if (_isConstructorMethod(state.methodNode)) { |
|
utils.append(superClassName + '.call(', state); |
|
} else { |
|
var protoProp = SUPER_PROTO_IDENT_PREFIX + superClassName; |
|
if (state.methodNode.key.type === Syntax.Identifier) { |
|
protoProp += '.' + state.methodNode.key.name; |
|
} else if (state.methodNode.key.type === Syntax.Literal) { |
|
protoProp += '[' + JSON.stringify(state.methodNode.key.value) + ']'; |
|
} |
|
utils.append(protoProp + ".call(", state); |
|
} |
|
utils.move(node.callee.range[1], state); |
|
} else if (node.callee.type === Syntax.MemberExpression) { |
|
utils.append(SUPER_PROTO_IDENT_PREFIX + superClassName, state); |
|
utils.move(node.callee.object.range[1], state); |
|
|
|
if (node.callee.computed) { |
|
// ["a" + "b"] |
|
utils.catchup(node.callee.property.range[1] + ']'.length, state); |
|
} else { |
|
// .ab |
|
utils.append('.' + node.callee.property.name, state); |
|
} |
|
|
|
utils.append('.call(', state); |
|
utils.move(node.callee.range[1], state); |
|
} |
|
|
|
utils.append('this', state); |
|
if (node.arguments.length > 0) { |
|
utils.append(',', state); |
|
utils.catchupWhiteSpace(node.arguments[0].range[0], state); |
|
traverse(node.arguments, path, state); |
|
} |
|
|
|
utils.catchupWhiteSpace(node.range[1], state); |
|
utils.append(')', state); |
|
return false; |
|
} |
|
visitSuperCallExpression.test = function(node, path, state) { |
|
if (state.superClass && node.type === Syntax.CallExpression) { |
|
var callee = node.callee; |
|
if (callee.type === Syntax.Identifier && callee.name === 'super' |
|
|| callee.type == Syntax.MemberExpression |
|
&& callee.object.name === 'super') { |
|
return true; |
|
} |
|
} |
|
return false; |
|
}; |
|
|
|
/** |
|
* @param {function} traverse |
|
* @param {object} node |
|
* @param {array} path |
|
* @param {object} state |
|
*/ |
|
function visitSuperMemberExpression(traverse, node, path, state) { |
|
var superClassName = state.superClass.name; |
|
|
|
utils.append(SUPER_PROTO_IDENT_PREFIX + superClassName, state); |
|
utils.move(node.object.range[1], state); |
|
} |
|
visitSuperMemberExpression.test = function(node, path, state) { |
|
return state.superClass |
|
&& node.type === Syntax.MemberExpression |
|
&& node.object.type === Syntax.Identifier |
|
&& node.object.name === 'super'; |
|
}; |
|
|
|
exports.resetSymbols = resetSymbols; |
|
|
|
exports.visitorList = [ |
|
visitClassDeclaration, |
|
visitClassExpression, |
|
visitClassFunctionExpression, |
|
visitClassMethod, |
|
visitClassMethodParam, |
|
visitPrivateIdentifier, |
|
visitSuperCallExpression, |
|
visitSuperMemberExpression |
|
]; |
|
|
|
},{"../src/utils":23,"./reserved-words-helper":34,"base62":10,"esprima-fb":9}],27:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2014 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
/*global exports:true*/ |
|
|
|
/** |
|
* Implements ES6 destructuring assignment and pattern matchng. |
|
* |
|
* function init({port, ip, coords: [x, y]}) { |
|
* return (x && y) ? {id, port} : {ip}; |
|
* }; |
|
* |
|
* function init($__0) { |
|
* var |
|
* port = $__0.port, |
|
* ip = $__0.ip, |
|
* $__1 = $__0.coords, |
|
* x = $__1[0], |
|
* y = $__1[1]; |
|
* return (x && y) ? {id, port} : {ip}; |
|
* } |
|
* |
|
* var x, {ip, port} = init({ip, port}); |
|
* |
|
* var x, $__0 = init({ip, port}), ip = $__0.ip, port = $__0.port; |
|
* |
|
*/ |
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
var reservedWordsHelper = _dereq_('./reserved-words-helper'); |
|
var restParamVisitors = _dereq_('./es6-rest-param-visitors'); |
|
var restPropertyHelpers = _dereq_('./es7-rest-property-helpers'); |
|
|
|
// ------------------------------------------------------- |
|
// 1. Structured variable declarations. |
|
// |
|
// var [a, b] = [b, a]; |
|
// var {x, y} = {y, x}; |
|
// ------------------------------------------------------- |
|
|
|
function visitStructuredVariable(traverse, node, path, state) { |
|
// Allocate new temp for the pattern. |
|
utils.append(utils.getTempVar(state.localScope.tempVarIndex) + '=', state); |
|
// Skip the pattern and assign the init to the temp. |
|
utils.catchupWhiteSpace(node.init.range[0], state); |
|
traverse(node.init, path, state); |
|
utils.catchup(node.init.range[1], state); |
|
// Render the destructured data. |
|
utils.append(',' + getDestructuredComponents(node.id, state), state); |
|
state.localScope.tempVarIndex++; |
|
return false; |
|
} |
|
|
|
visitStructuredVariable.test = function(node, path, state) { |
|
return node.type === Syntax.VariableDeclarator && |
|
isStructuredPattern(node.id); |
|
}; |
|
|
|
function isStructuredPattern(node) { |
|
return node.type === Syntax.ObjectPattern || |
|
node.type === Syntax.ArrayPattern; |
|
} |
|
|
|
// Main function which does actual recursive destructuring |
|
// of nested complex structures. |
|
function getDestructuredComponents(node, state) { |
|
var tmpIndex = state.localScope.tempVarIndex; |
|
var components = []; |
|
var patternItems = getPatternItems(node); |
|
|
|
for (var idx = 0; idx < patternItems.length; idx++) { |
|
var item = patternItems[idx]; |
|
if (!item) { |
|
continue; |
|
} |
|
|
|
if (item.type === Syntax.SpreadElement) { |
|
// Spread/rest of an array. |
|
// TODO(dmitrys): support spread in the middle of a pattern |
|
// and also for function param patterns: [x, ...xs, y] |
|
components.push(item.argument.name + |
|
'=Array.prototype.slice.call(' + |
|
utils.getTempVar(tmpIndex) + ',' + idx + ')' |
|
); |
|
continue; |
|
} |
|
|
|
if (item.type === Syntax.SpreadProperty) { |
|
var restExpression = restPropertyHelpers.renderRestExpression( |
|
utils.getTempVar(tmpIndex), |
|
patternItems |
|
); |
|
components.push(item.argument.name + '=' + restExpression); |
|
continue; |
|
} |
|
|
|
// Depending on pattern type (Array or Object), we get |
|
// corresponding pattern item parts. |
|
var accessor = getPatternItemAccessor(node, item, tmpIndex, idx); |
|
var value = getPatternItemValue(node, item); |
|
|
|
// TODO(dmitrys): implement default values: {x, y=5} |
|
if (value.type === Syntax.Identifier) { |
|
// Simple pattern item. |
|
components.push(value.name + '=' + accessor); |
|
} else { |
|
// Complex sub-structure. |
|
components.push( |
|
utils.getTempVar(++state.localScope.tempVarIndex) + '=' + accessor + |
|
',' + getDestructuredComponents(value, state) |
|
); |
|
} |
|
} |
|
|
|
return components.join(','); |
|
} |
|
|
|
function getPatternItems(node) { |
|
return node.properties || node.elements; |
|
} |
|
|
|
function getPatternItemAccessor(node, patternItem, tmpIndex, idx) { |
|
var tmpName = utils.getTempVar(tmpIndex); |
|
if (node.type === Syntax.ObjectPattern) { |
|
if (reservedWordsHelper.isReservedWord(patternItem.key.name)) { |
|
return tmpName + '["' + patternItem.key.name + '"]'; |
|
} else if (patternItem.key.type === Syntax.Literal) { |
|
return tmpName + '[' + JSON.stringify(patternItem.key.value) + ']'; |
|
} else if (patternItem.key.type === Syntax.Identifier) { |
|
return tmpName + '.' + patternItem.key.name; |
|
} |
|
} else if (node.type === Syntax.ArrayPattern) { |
|
return tmpName + '[' + idx + ']'; |
|
} |
|
} |
|
|
|
function getPatternItemValue(node, patternItem) { |
|
return node.type === Syntax.ObjectPattern |
|
? patternItem.value |
|
: patternItem; |
|
} |
|
|
|
// ------------------------------------------------------- |
|
// 2. Assignment expression. |
|
// |
|
// [a, b] = [b, a]; |
|
// ({x, y} = {y, x}); |
|
// ------------------------------------------------------- |
|
|
|
function visitStructuredAssignment(traverse, node, path, state) { |
|
var exprNode = node.expression; |
|
utils.append('var ' + utils.getTempVar(state.localScope.tempVarIndex) + '=', state); |
|
|
|
utils.catchupWhiteSpace(exprNode.right.range[0], state); |
|
traverse(exprNode.right, path, state); |
|
utils.catchup(exprNode.right.range[1], state); |
|
|
|
utils.append( |
|
';' + getDestructuredComponents(exprNode.left, state) + ';', |
|
state |
|
); |
|
|
|
utils.catchupWhiteSpace(node.range[1], state); |
|
state.localScope.tempVarIndex++; |
|
return false; |
|
} |
|
|
|
visitStructuredAssignment.test = function(node, path, state) { |
|
// We consider the expression statement rather than just assignment |
|
// expression to cover case with object patters which should be |
|
// wrapped in grouping operator: ({x, y} = {y, x}); |
|
return node.type === Syntax.ExpressionStatement && |
|
node.expression.type === Syntax.AssignmentExpression && |
|
isStructuredPattern(node.expression.left); |
|
}; |
|
|
|
// ------------------------------------------------------- |
|
// 3. Structured parameter. |
|
// |
|
// function foo({x, y}) { ... } |
|
// ------------------------------------------------------- |
|
|
|
function visitStructuredParameter(traverse, node, path, state) { |
|
utils.append(utils.getTempVar(getParamIndex(node, path)), state); |
|
utils.catchupWhiteSpace(node.range[1], state); |
|
return true; |
|
} |
|
|
|
function getParamIndex(paramNode, path) { |
|
var funcNode = path[0]; |
|
var tmpIndex = 0; |
|
for (var k = 0; k < funcNode.params.length; k++) { |
|
var param = funcNode.params[k]; |
|
if (param === paramNode) { |
|
break; |
|
} |
|
if (isStructuredPattern(param)) { |
|
tmpIndex++; |
|
} |
|
} |
|
return tmpIndex; |
|
} |
|
|
|
visitStructuredParameter.test = function(node, path, state) { |
|
return isStructuredPattern(node) && isFunctionNode(path[0]); |
|
}; |
|
|
|
function isFunctionNode(node) { |
|
return (node.type == Syntax.FunctionDeclaration || |
|
node.type == Syntax.FunctionExpression || |
|
node.type == Syntax.MethodDefinition || |
|
node.type == Syntax.ArrowFunctionExpression); |
|
} |
|
|
|
// ------------------------------------------------------- |
|
// 4. Function body for structured parameters. |
|
// |
|
// function foo({x, y}) { x; y; } |
|
// ------------------------------------------------------- |
|
|
|
function visitFunctionBodyForStructuredParameter(traverse, node, path, state) { |
|
var funcNode = path[0]; |
|
|
|
utils.catchup(funcNode.body.range[0] + 1, state); |
|
renderDestructuredComponents(funcNode, state); |
|
|
|
if (funcNode.rest) { |
|
utils.append( |
|
restParamVisitors.renderRestParamSetup(funcNode, state), |
|
state |
|
); |
|
} |
|
|
|
return true; |
|
} |
|
|
|
function renderDestructuredComponents(funcNode, state) { |
|
var destructuredComponents = []; |
|
|
|
for (var k = 0; k < funcNode.params.length; k++) { |
|
var param = funcNode.params[k]; |
|
if (isStructuredPattern(param)) { |
|
destructuredComponents.push( |
|
getDestructuredComponents(param, state) |
|
); |
|
state.localScope.tempVarIndex++; |
|
} |
|
} |
|
|
|
if (destructuredComponents.length) { |
|
utils.append('var ' + destructuredComponents.join(',') + ';', state); |
|
} |
|
} |
|
|
|
visitFunctionBodyForStructuredParameter.test = function(node, path, state) { |
|
return node.type === Syntax.BlockStatement && isFunctionNode(path[0]); |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitStructuredVariable, |
|
visitStructuredAssignment, |
|
visitStructuredParameter, |
|
visitFunctionBodyForStructuredParameter |
|
]; |
|
|
|
exports.renderDestructuredComponents = renderDestructuredComponents; |
|
|
|
|
|
},{"../src/utils":23,"./es6-rest-param-visitors":30,"./es7-rest-property-helpers":32,"./reserved-words-helper":34,"esprima-fb":9}],28:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*jslint node:true*/ |
|
|
|
/** |
|
* Desugars concise methods of objects to function expressions. |
|
* |
|
* var foo = { |
|
* method(x, y) { ... } |
|
* }; |
|
* |
|
* var foo = { |
|
* method: function(x, y) { ... } |
|
* }; |
|
* |
|
*/ |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
var reservedWordsHelper = _dereq_('./reserved-words-helper'); |
|
|
|
function visitObjectConciseMethod(traverse, node, path, state) { |
|
var isGenerator = node.value.generator; |
|
if (isGenerator) { |
|
utils.catchupWhiteSpace(node.range[0] + 1, state); |
|
} |
|
if (node.computed) { // [<expr>]() { ...} |
|
utils.catchup(node.key.range[1] + 1, state); |
|
} else if (reservedWordsHelper.isReservedWord(node.key.name)) { |
|
utils.catchup(node.key.range[0], state); |
|
utils.append('"', state); |
|
utils.catchup(node.key.range[1], state); |
|
utils.append('"', state); |
|
} |
|
|
|
utils.catchup(node.key.range[1], state); |
|
utils.append( |
|
':function' + (isGenerator ? '*' : ''), |
|
state |
|
); |
|
path.unshift(node); |
|
traverse(node.value, path, state); |
|
path.shift(); |
|
return false; |
|
} |
|
|
|
visitObjectConciseMethod.test = function(node, path, state) { |
|
return node.type === Syntax.Property && |
|
node.value.type === Syntax.FunctionExpression && |
|
node.method === true; |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitObjectConciseMethod |
|
]; |
|
|
|
},{"../src/utils":23,"./reserved-words-helper":34,"esprima-fb":9}],29:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*jslint node: true*/ |
|
|
|
/** |
|
* Desugars ES6 Object Literal short notations into ES3 full notation. |
|
* |
|
* // Easier return values. |
|
* function foo(x, y) { |
|
* return {x, y}; // {x: x, y: y} |
|
* }; |
|
* |
|
* // Destructuring. |
|
* function init({port, ip, coords: {x, y}}) { ... } |
|
* |
|
*/ |
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
/** |
|
* @public |
|
*/ |
|
function visitObjectLiteralShortNotation(traverse, node, path, state) { |
|
utils.catchup(node.key.range[1], state); |
|
utils.append(':' + node.key.name, state); |
|
return false; |
|
} |
|
|
|
visitObjectLiteralShortNotation.test = function(node, path, state) { |
|
return node.type === Syntax.Property && |
|
node.kind === 'init' && |
|
node.shorthand === true && |
|
path[0].type !== Syntax.ObjectPattern; |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitObjectLiteralShortNotation |
|
]; |
|
|
|
|
|
},{"../src/utils":23,"esprima-fb":9}],30:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*jslint node:true*/ |
|
|
|
/** |
|
* Desugars ES6 rest parameters into an ES3 arguments array. |
|
* |
|
* function printf(template, ...args) { |
|
* args.forEach(...); |
|
* } |
|
* |
|
* We could use `Array.prototype.slice.call`, but that usage of arguments causes |
|
* functions to be deoptimized in V8, so instead we use a for-loop. |
|
* |
|
* function printf(template) { |
|
* for (var args = [], $__0 = 1, $__1 = arguments.length; $__0 < $__1; $__0++) |
|
* args.push(arguments[$__0]); |
|
* args.forEach(...); |
|
* } |
|
* |
|
*/ |
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
|
|
|
|
function _nodeIsFunctionWithRestParam(node) { |
|
return (node.type === Syntax.FunctionDeclaration |
|
|| node.type === Syntax.FunctionExpression |
|
|| node.type === Syntax.ArrowFunctionExpression) |
|
&& node.rest; |
|
} |
|
|
|
function visitFunctionParamsWithRestParam(traverse, node, path, state) { |
|
if (node.parametricType) { |
|
utils.catchup(node.parametricType.range[0], state); |
|
path.unshift(node); |
|
traverse(node.parametricType, path, state); |
|
path.shift(); |
|
} |
|
|
|
// Render params. |
|
if (node.params.length) { |
|
path.unshift(node); |
|
traverse(node.params, path, state); |
|
path.shift(); |
|
} else { |
|
// -3 is for ... of the rest. |
|
utils.catchup(node.rest.range[0] - 3, state); |
|
} |
|
utils.catchupWhiteSpace(node.rest.range[1], state); |
|
|
|
path.unshift(node); |
|
traverse(node.body, path, state); |
|
path.shift(); |
|
|
|
return false; |
|
} |
|
|
|
visitFunctionParamsWithRestParam.test = function(node, path, state) { |
|
return _nodeIsFunctionWithRestParam(node); |
|
}; |
|
|
|
function renderRestParamSetup(functionNode, state) { |
|
var idx = state.localScope.tempVarIndex++; |
|
var len = state.localScope.tempVarIndex++; |
|
|
|
return 'for (var ' + functionNode.rest.name + '=[],' + |
|
utils.getTempVar(idx) + '=' + functionNode.params.length + ',' + |
|
utils.getTempVar(len) + '=arguments.length;' + |
|
utils.getTempVar(idx) + '<' + utils.getTempVar(len) + ';' + |
|
utils.getTempVar(idx) + '++) ' + |
|
functionNode.rest.name + '.push(arguments[' + utils.getTempVar(idx) + ']);'; |
|
} |
|
|
|
function visitFunctionBodyWithRestParam(traverse, node, path, state) { |
|
utils.catchup(node.range[0] + 1, state); |
|
var parentNode = path[0]; |
|
utils.append(renderRestParamSetup(parentNode, state), state); |
|
return true; |
|
} |
|
|
|
visitFunctionBodyWithRestParam.test = function(node, path, state) { |
|
return node.type === Syntax.BlockStatement |
|
&& _nodeIsFunctionWithRestParam(path[0]); |
|
}; |
|
|
|
exports.renderRestParamSetup = renderRestParamSetup; |
|
exports.visitorList = [ |
|
visitFunctionParamsWithRestParam, |
|
visitFunctionBodyWithRestParam |
|
]; |
|
|
|
},{"../src/utils":23,"esprima-fb":9}],31:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*jslint node:true*/ |
|
|
|
/** |
|
* @typechecks |
|
*/ |
|
'use strict'; |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
/** |
|
* http://people.mozilla.org/~jorendorff/es6-draft.html#sec-12.1.9 |
|
*/ |
|
function visitTemplateLiteral(traverse, node, path, state) { |
|
var templateElements = node.quasis; |
|
|
|
utils.append('(', state); |
|
for (var ii = 0; ii < templateElements.length; ii++) { |
|
var templateElement = templateElements[ii]; |
|
if (templateElement.value.raw !== '') { |
|
utils.append(getCookedValue(templateElement), state); |
|
if (!templateElement.tail) { |
|
// + between element and substitution |
|
utils.append(' + ', state); |
|
} |
|
// maintain line numbers |
|
utils.move(templateElement.range[0], state); |
|
utils.catchupNewlines(templateElement.range[1], state); |
|
} else { // templateElement.value.raw === '' |
|
// Concatenat adjacent substitutions, e.g. `${x}${y}`. Empty templates |
|
// appear before the first and after the last element - nothing to add in |
|
// those cases. |
|
if (ii > 0 && !templateElement.tail) { |
|
// + between substitution and substitution |
|
utils.append(' + ', state); |
|
} |
|
} |
|
|
|
utils.move(templateElement.range[1], state); |
|
if (!templateElement.tail) { |
|
var substitution = node.expressions[ii]; |
|
if (substitution.type === Syntax.Identifier || |
|
substitution.type === Syntax.MemberExpression || |
|
substitution.type === Syntax.CallExpression) { |
|
utils.catchup(substitution.range[1], state); |
|
} else { |
|
utils.append('(', state); |
|
traverse(substitution, path, state); |
|
utils.catchup(substitution.range[1], state); |
|
utils.append(')', state); |
|
} |
|
// if next templateElement isn't empty... |
|
if (templateElements[ii + 1].value.cooked !== '') { |
|
utils.append(' + ', state); |
|
} |
|
} |
|
} |
|
utils.move(node.range[1], state); |
|
utils.append(')', state); |
|
return false; |
|
} |
|
|
|
visitTemplateLiteral.test = function(node, path, state) { |
|
return node.type === Syntax.TemplateLiteral; |
|
}; |
|
|
|
/** |
|
* http://people.mozilla.org/~jorendorff/es6-draft.html#sec-12.2.6 |
|
*/ |
|
function visitTaggedTemplateExpression(traverse, node, path, state) { |
|
var template = node.quasi; |
|
var numQuasis = template.quasis.length; |
|
|
|
// print the tag |
|
utils.move(node.tag.range[0], state); |
|
traverse(node.tag, path, state); |
|
utils.catchup(node.tag.range[1], state); |
|
|
|
// print array of template elements |
|
utils.append('(function() { var siteObj = [', state); |
|
for (var ii = 0; ii < numQuasis; ii++) { |
|
utils.append(getCookedValue(template.quasis[ii]), state); |
|
if (ii !== numQuasis - 1) { |
|
utils.append(', ', state); |
|
} |
|
} |
|
utils.append(']; siteObj.raw = [', state); |
|
for (ii = 0; ii < numQuasis; ii++) { |
|
utils.append(getRawValue(template.quasis[ii]), state); |
|
if (ii !== numQuasis - 1) { |
|
utils.append(', ', state); |
|
} |
|
} |
|
utils.append( |
|
']; Object.freeze(siteObj.raw); Object.freeze(siteObj); return siteObj; }()', |
|
state |
|
); |
|
|
|
// print substitutions |
|
if (numQuasis > 1) { |
|
for (ii = 0; ii < template.expressions.length; ii++) { |
|
var expression = template.expressions[ii]; |
|
utils.append(', ', state); |
|
|
|
// maintain line numbers by calling catchupWhiteSpace over the whole |
|
// previous TemplateElement |
|
utils.move(template.quasis[ii].range[0], state); |
|
utils.catchupNewlines(template.quasis[ii].range[1], state); |
|
|
|
utils.move(expression.range[0], state); |
|
traverse(expression, path, state); |
|
utils.catchup(expression.range[1], state); |
|
} |
|
} |
|
|
|
// print blank lines to push the closing ) down to account for the final |
|
// TemplateElement. |
|
utils.catchupNewlines(node.range[1], state); |
|
|
|
utils.append(')', state); |
|
|
|
return false; |
|
} |
|
|
|
visitTaggedTemplateExpression.test = function(node, path, state) { |
|
return node.type === Syntax.TaggedTemplateExpression; |
|
}; |
|
|
|
function getCookedValue(templateElement) { |
|
return JSON.stringify(templateElement.value.cooked); |
|
} |
|
|
|
function getRawValue(templateElement) { |
|
return JSON.stringify(templateElement.value.raw); |
|
} |
|
|
|
exports.visitorList = [ |
|
visitTemplateLiteral, |
|
visitTaggedTemplateExpression |
|
]; |
|
|
|
},{"../src/utils":23,"esprima-fb":9}],32:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
/*jslint node:true*/ |
|
|
|
/** |
|
* Desugars ES7 rest properties into ES5 object iteration. |
|
*/ |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
|
|
// TODO: This is a pretty massive helper, it should only be defined once, in the |
|
// transform's runtime environment. We don't currently have a runtime though. |
|
var restFunction = |
|
'(function(source, exclusion) {' + |
|
'var rest = {};' + |
|
'var hasOwn = Object.prototype.hasOwnProperty;' + |
|
'if (source == null) {' + |
|
'throw new TypeError();' + |
|
'}' + |
|
'for (var key in source) {' + |
|
'if (hasOwn.call(source, key) && !hasOwn.call(exclusion, key)) {' + |
|
'rest[key] = source[key];' + |
|
'}' + |
|
'}' + |
|
'return rest;' + |
|
'})'; |
|
|
|
function getPropertyNames(properties) { |
|
var names = []; |
|
for (var i = 0; i < properties.length; i++) { |
|
var property = properties[i]; |
|
if (property.type === Syntax.SpreadProperty) { |
|
continue; |
|
} |
|
if (property.type === Syntax.Identifier) { |
|
names.push(property.name); |
|
} else { |
|
names.push(property.key.name); |
|
} |
|
} |
|
return names; |
|
} |
|
|
|
function getRestFunctionCall(source, exclusion) { |
|
return restFunction + '(' + source + ',' + exclusion + ')'; |
|
} |
|
|
|
function getSimpleShallowCopy(accessorExpression) { |
|
// This could be faster with 'Object.assign({}, ' + accessorExpression + ')' |
|
// but to unify code paths and avoid a ES6 dependency we use the same |
|
// helper as for the exclusion case. |
|
return getRestFunctionCall(accessorExpression, '{}'); |
|
} |
|
|
|
function renderRestExpression(accessorExpression, excludedProperties) { |
|
var excludedNames = getPropertyNames(excludedProperties); |
|
if (!excludedNames.length) { |
|
return getSimpleShallowCopy(accessorExpression); |
|
} |
|
return getRestFunctionCall( |
|
accessorExpression, |
|
'{' + excludedNames.join(':1,') + ':1}' |
|
); |
|
} |
|
|
|
exports.renderRestExpression = renderRestExpression; |
|
|
|
},{"esprima-fb":9}],33:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2004-present Facebook. All Rights Reserved. |
|
*/ |
|
/*global exports:true*/ |
|
|
|
/** |
|
* Implements ES7 object spread property. |
|
* https://gist.github.com/sebmarkbage/aa849c7973cb4452c547 |
|
* |
|
* { ...a, x: 1 } |
|
* |
|
* Object.assign({}, a, {x: 1 }) |
|
* |
|
*/ |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
|
|
function visitObjectLiteralSpread(traverse, node, path, state) { |
|
utils.catchup(node.range[0], state); |
|
|
|
utils.append('Object.assign({', state); |
|
|
|
// Skip the original { |
|
utils.move(node.range[0] + 1, state); |
|
|
|
var previousWasSpread = false; |
|
|
|
for (var i = 0; i < node.properties.length; i++) { |
|
var property = node.properties[i]; |
|
if (property.type === Syntax.SpreadProperty) { |
|
|
|
// Close the previous object or initial object |
|
if (!previousWasSpread) { |
|
utils.append('}', state); |
|
} |
|
|
|
if (i === 0) { |
|
// Normally there will be a comma when we catch up, but not before |
|
// the first property. |
|
utils.append(',', state); |
|
} |
|
|
|
utils.catchup(property.range[0], state); |
|
|
|
// skip ... |
|
utils.move(property.range[0] + 3, state); |
|
|
|
traverse(property.argument, path, state); |
|
|
|
utils.catchup(property.range[1], state); |
|
|
|
previousWasSpread = true; |
|
|
|
} else { |
|
|
|
utils.catchup(property.range[0], state); |
|
|
|
if (previousWasSpread) { |
|
utils.append('{', state); |
|
} |
|
|
|
traverse(property, path, state); |
|
|
|
utils.catchup(property.range[1], state); |
|
|
|
previousWasSpread = false; |
|
|
|
} |
|
} |
|
|
|
// Strip any non-whitespace between the last item and the end. |
|
// We only catch up on whitespace so that we ignore any trailing commas which |
|
// are stripped out for IE8 support. Unfortunately, this also strips out any |
|
// trailing comments. |
|
utils.catchupWhiteSpace(node.range[1] - 1, state); |
|
|
|
// Skip the trailing } |
|
utils.move(node.range[1], state); |
|
|
|
if (!previousWasSpread) { |
|
utils.append('}', state); |
|
} |
|
|
|
utils.append(')', state); |
|
return false; |
|
} |
|
|
|
visitObjectLiteralSpread.test = function(node, path, state) { |
|
if (node.type !== Syntax.ObjectExpression) { |
|
return false; |
|
} |
|
// Tight loop optimization |
|
var hasAtLeastOneSpreadProperty = false; |
|
for (var i = 0; i < node.properties.length; i++) { |
|
var property = node.properties[i]; |
|
if (property.type === Syntax.SpreadProperty) { |
|
hasAtLeastOneSpreadProperty = true; |
|
} else if (property.kind !== 'init') { |
|
return false; |
|
} |
|
} |
|
return hasAtLeastOneSpreadProperty; |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitObjectLiteralSpread |
|
]; |
|
|
|
},{"../src/utils":23,"esprima-fb":9}],34:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2014 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
*/ |
|
|
|
var KEYWORDS = [ |
|
'break', 'do', 'in', 'typeof', 'case', 'else', 'instanceof', 'var', 'catch', |
|
'export', 'new', 'void', 'class', 'extends', 'return', 'while', 'const', |
|
'finally', 'super', 'with', 'continue', 'for', 'switch', 'yield', 'debugger', |
|
'function', 'this', 'default', 'if', 'throw', 'delete', 'import', 'try' |
|
]; |
|
|
|
var FUTURE_RESERVED_WORDS = [ |
|
'enum', 'await', 'implements', 'package', 'protected', 'static', 'interface', |
|
'private', 'public' |
|
]; |
|
|
|
var LITERALS = [ |
|
'null', |
|
'true', |
|
'false' |
|
]; |
|
|
|
// https://people.mozilla.org/~jorendorff/es6-draft.html#sec-reserved-words |
|
var RESERVED_WORDS = [].concat( |
|
KEYWORDS, |
|
FUTURE_RESERVED_WORDS, |
|
LITERALS |
|
); |
|
|
|
var reservedWordsMap = Object.create(null); |
|
RESERVED_WORDS.forEach(function(k) { |
|
reservedWordsMap[k] = true; |
|
}); |
|
|
|
/** |
|
* This list should not grow as new reserved words are introdued. This list is |
|
* of words that need to be quoted because ES3-ish browsers do not allow their |
|
* use as identifier names. |
|
*/ |
|
var ES3_FUTURE_RESERVED_WORDS = [ |
|
'enum', 'implements', 'package', 'protected', 'static', 'interface', |
|
'private', 'public' |
|
]; |
|
|
|
var ES3_RESERVED_WORDS = [].concat( |
|
KEYWORDS, |
|
ES3_FUTURE_RESERVED_WORDS, |
|
LITERALS |
|
); |
|
|
|
var es3ReservedWordsMap = Object.create(null); |
|
ES3_RESERVED_WORDS.forEach(function(k) { |
|
es3ReservedWordsMap[k] = true; |
|
}); |
|
|
|
exports.isReservedWord = function(word) { |
|
return !!reservedWordsMap[word]; |
|
}; |
|
|
|
exports.isES3ReservedWord = function(word) { |
|
return !!es3ReservedWordsMap[word]; |
|
}; |
|
|
|
},{}],35:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2014 Facebook, Inc. |
|
* |
|
* Licensed under the Apache License, Version 2.0 (the "License"); |
|
* you may not use this file except in compliance with the License. |
|
* You may obtain a copy of the License at |
|
* |
|
* http://www.apache.org/licenses/LICENSE-2.0 |
|
* |
|
* Unless required by applicable law or agreed to in writing, software |
|
* distributed under the License is distributed on an "AS IS" BASIS, |
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
* See the License for the specific language governing permissions and |
|
* limitations under the License. |
|
* |
|
*/ |
|
/*global exports:true*/ |
|
|
|
var Syntax = _dereq_('esprima-fb').Syntax; |
|
var utils = _dereq_('../src/utils'); |
|
var reserverdWordsHelper = _dereq_('./reserved-words-helper'); |
|
|
|
/** |
|
* Code adapted from https://github.com/spicyj/es3ify |
|
* The MIT License (MIT) |
|
* Copyright (c) 2014 Ben Alpert |
|
*/ |
|
|
|
function visitProperty(traverse, node, path, state) { |
|
utils.catchup(node.key.range[0], state); |
|
utils.append('"', state); |
|
utils.catchup(node.key.range[1], state); |
|
utils.append('"', state); |
|
utils.catchup(node.value.range[0], state); |
|
traverse(node.value, path, state); |
|
return false; |
|
} |
|
|
|
visitProperty.test = function(node) { |
|
return node.type === Syntax.Property && |
|
node.key.type === Syntax.Identifier && |
|
!node.method && |
|
!node.shorthand && |
|
!node.computed && |
|
reserverdWordsHelper.isES3ReservedWord(node.key.name); |
|
}; |
|
|
|
function visitMemberExpression(traverse, node, path, state) { |
|
traverse(node.object, path, state); |
|
utils.catchup(node.property.range[0] - 1, state); |
|
utils.append('[', state); |
|
utils.catchupWhiteSpace(node.property.range[0], state); |
|
utils.append('"', state); |
|
utils.catchup(node.property.range[1], state); |
|
utils.append('"]', state); |
|
return false; |
|
} |
|
|
|
visitMemberExpression.test = function(node) { |
|
return node.type === Syntax.MemberExpression && |
|
node.property.type === Syntax.Identifier && |
|
reserverdWordsHelper.isES3ReservedWord(node.property.name); |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitProperty, |
|
visitMemberExpression |
|
]; |
|
|
|
},{"../src/utils":23,"./reserved-words-helper":34,"esprima-fb":9}],36:[function(_dereq_,module,exports){ |
|
var esprima = _dereq_('esprima-fb'); |
|
var utils = _dereq_('../src/utils'); |
|
|
|
var Syntax = esprima.Syntax; |
|
|
|
function _isFunctionNode(node) { |
|
return node.type === Syntax.FunctionDeclaration |
|
|| node.type === Syntax.FunctionExpression |
|
|| node.type === Syntax.ArrowFunctionExpression; |
|
} |
|
|
|
function visitClassProperty(traverse, node, path, state) { |
|
utils.catchup(node.range[0], state); |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitClassProperty.test = function(node, path, state) { |
|
return node.type === Syntax.ClassProperty; |
|
}; |
|
|
|
function visitTypeAlias(traverse, node, path, state) { |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitTypeAlias.test = function(node, path, state) { |
|
return node.type === Syntax.TypeAlias; |
|
}; |
|
|
|
function visitTypeCast(traverse, node, path, state) { |
|
path.unshift(node); |
|
traverse(node.expression, path, state); |
|
path.shift(); |
|
|
|
utils.catchup(node.typeAnnotation.range[0], state); |
|
utils.catchupWhiteOut(node.typeAnnotation.range[1], state); |
|
return false; |
|
} |
|
visitTypeCast.test = function(node, path, state) { |
|
return node.type === Syntax.TypeCastExpression; |
|
}; |
|
|
|
function visitInterfaceDeclaration(traverse, node, path, state) { |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitInterfaceDeclaration.test = function(node, path, state) { |
|
return node.type === Syntax.InterfaceDeclaration; |
|
}; |
|
|
|
function visitDeclare(traverse, node, path, state) { |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitDeclare.test = function(node, path, state) { |
|
switch (node.type) { |
|
case Syntax.DeclareVariable: |
|
case Syntax.DeclareFunction: |
|
case Syntax.DeclareClass: |
|
case Syntax.DeclareModule: |
|
return true; |
|
} |
|
return false; |
|
}; |
|
|
|
function visitFunctionParametricAnnotation(traverse, node, path, state) { |
|
utils.catchup(node.range[0], state); |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitFunctionParametricAnnotation.test = function(node, path, state) { |
|
return node.type === Syntax.TypeParameterDeclaration |
|
&& path[0] |
|
&& _isFunctionNode(path[0]) |
|
&& node === path[0].typeParameters; |
|
}; |
|
|
|
function visitFunctionReturnAnnotation(traverse, node, path, state) { |
|
utils.catchup(node.range[0], state); |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitFunctionReturnAnnotation.test = function(node, path, state) { |
|
return path[0] && _isFunctionNode(path[0]) && node === path[0].returnType; |
|
}; |
|
|
|
function visitOptionalFunctionParameterAnnotation(traverse, node, path, state) { |
|
utils.catchup(node.range[0] + node.name.length, state); |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitOptionalFunctionParameterAnnotation.test = function(node, path, state) { |
|
return node.type === Syntax.Identifier |
|
&& node.optional |
|
&& path[0] |
|
&& _isFunctionNode(path[0]); |
|
}; |
|
|
|
function visitTypeAnnotatedIdentifier(traverse, node, path, state) { |
|
utils.catchup(node.typeAnnotation.range[0], state); |
|
utils.catchupWhiteOut(node.typeAnnotation.range[1], state); |
|
return false; |
|
} |
|
visitTypeAnnotatedIdentifier.test = function(node, path, state) { |
|
return node.type === Syntax.Identifier && node.typeAnnotation; |
|
}; |
|
|
|
function visitTypeAnnotatedObjectOrArrayPattern(traverse, node, path, state) { |
|
utils.catchup(node.typeAnnotation.range[0], state); |
|
utils.catchupWhiteOut(node.typeAnnotation.range[1], state); |
|
return false; |
|
} |
|
visitTypeAnnotatedObjectOrArrayPattern.test = function(node, path, state) { |
|
var rightType = node.type === Syntax.ObjectPattern |
|
|| node.type === Syntax.ArrayPattern; |
|
return rightType && node.typeAnnotation; |
|
}; |
|
|
|
/** |
|
* Methods cause trouble, since esprima parses them as a key/value pair, where |
|
* the location of the value starts at the method body. For example |
|
* { bar(x:number,...y:Array<number>):number {} } |
|
* is parsed as |
|
* { bar: function(x: number, ...y:Array<number>): number {} } |
|
* except that the location of the FunctionExpression value is 40-something, |
|
* which is the location of the function body. This means that by the time we |
|
* visit the params, rest param, and return type organically, we've already |
|
* catchup()'d passed them. |
|
*/ |
|
function visitMethod(traverse, node, path, state) { |
|
path.unshift(node); |
|
traverse(node.key, path, state); |
|
|
|
path.unshift(node.value); |
|
traverse(node.value.params, path, state); |
|
node.value.rest && traverse(node.value.rest, path, state); |
|
node.value.returnType && traverse(node.value.returnType, path, state); |
|
traverse(node.value.body, path, state); |
|
|
|
path.shift(); |
|
|
|
path.shift(); |
|
return false; |
|
} |
|
|
|
visitMethod.test = function(node, path, state) { |
|
return (node.type === "Property" && (node.method || node.kind === "set" || node.kind === "get")) |
|
|| (node.type === "MethodDefinition"); |
|
}; |
|
|
|
function visitImportType(traverse, node, path, state) { |
|
utils.catchupWhiteOut(node.range[1], state); |
|
return false; |
|
} |
|
visitImportType.test = function(node, path, state) { |
|
return node.type === 'ImportDeclaration' |
|
&& node.isType; |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitClassProperty, |
|
visitDeclare, |
|
visitImportType, |
|
visitInterfaceDeclaration, |
|
visitFunctionParametricAnnotation, |
|
visitFunctionReturnAnnotation, |
|
visitMethod, |
|
visitOptionalFunctionParameterAnnotation, |
|
visitTypeAlias, |
|
visitTypeCast, |
|
visitTypeAnnotatedIdentifier, |
|
visitTypeAnnotatedObjectOrArrayPattern |
|
]; |
|
|
|
},{"../src/utils":23,"esprima-fb":9}],37:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013-2015, Facebook, Inc. |
|
* All rights reserved. |
|
* |
|
* This source code is licensed under the BSD-style license found in the |
|
* LICENSE file in the root directory of this source tree. An additional grant |
|
* of patent rights can be found in the PATENTS file in the same directory. |
|
*/ |
|
/*global exports:true*/ |
|
'use strict'; |
|
var Syntax = _dereq_('jstransform').Syntax; |
|
var utils = _dereq_('jstransform/src/utils'); |
|
|
|
function renderJSXLiteral(object, isLast, state, start, end) { |
|
var lines = object.value.split(/\r\n|\n|\r/); |
|
|
|
if (start) { |
|
utils.append(start, state); |
|
} |
|
|
|
var lastNonEmptyLine = 0; |
|
|
|
lines.forEach(function(line, index) { |
|
if (line.match(/[^ \t]/)) { |
|
lastNonEmptyLine = index; |
|
} |
|
}); |
|
|
|
lines.forEach(function(line, index) { |
|
var isFirstLine = index === 0; |
|
var isLastLine = index === lines.length - 1; |
|
var isLastNonEmptyLine = index === lastNonEmptyLine; |
|
|
|
// replace rendered whitespace tabs with spaces |
|
var trimmedLine = line.replace(/\t/g, ' '); |
|
|
|
// trim whitespace touching a newline |
|
if (!isFirstLine) { |
|
trimmedLine = trimmedLine.replace(/^[ ]+/, ''); |
|
} |
|
if (!isLastLine) { |
|
trimmedLine = trimmedLine.replace(/[ ]+$/, ''); |
|
} |
|
|
|
if (!isFirstLine) { |
|
utils.append(line.match(/^[ \t]*/)[0], state); |
|
} |
|
|
|
if (trimmedLine || isLastNonEmptyLine) { |
|
utils.append( |
|
JSON.stringify(trimmedLine) + |
|
(!isLastNonEmptyLine ? ' + \' \' +' : ''), |
|
state); |
|
|
|
if (isLastNonEmptyLine) { |
|
if (end) { |
|
utils.append(end, state); |
|
} |
|
if (!isLast) { |
|
utils.append(', ', state); |
|
} |
|
} |
|
|
|
// only restore tail whitespace if line had literals |
|
if (trimmedLine && !isLastLine) { |
|
utils.append(line.match(/[ \t]*$/)[0], state); |
|
} |
|
} |
|
|
|
if (!isLastLine) { |
|
utils.append('\n', state); |
|
} |
|
}); |
|
|
|
utils.move(object.range[1], state); |
|
} |
|
|
|
function renderJSXExpressionContainer(traverse, object, isLast, path, state) { |
|
// Plus 1 to skip `{`. |
|
utils.move(object.range[0] + 1, state); |
|
utils.catchup(object.expression.range[0], state); |
|
traverse(object.expression, path, state); |
|
|
|
if (!isLast && object.expression.type !== Syntax.JSXEmptyExpression) { |
|
// If we need to append a comma, make sure to do so after the expression. |
|
utils.catchup(object.expression.range[1], state, trimLeft); |
|
utils.append(', ', state); |
|
} |
|
|
|
// Minus 1 to skip `}`. |
|
utils.catchup(object.range[1] - 1, state, trimLeft); |
|
utils.move(object.range[1], state); |
|
return false; |
|
} |
|
|
|
function quoteAttrName(attr) { |
|
// Quote invalid JS identifiers. |
|
if (!/^[a-z_$][a-z\d_$]*$/i.test(attr)) { |
|
return '"' + attr + '"'; |
|
} |
|
return attr; |
|
} |
|
|
|
function trimLeft(value) { |
|
return value.replace(/^[ ]+/, ''); |
|
} |
|
|
|
exports.renderJSXExpressionContainer = renderJSXExpressionContainer; |
|
exports.renderJSXLiteral = renderJSXLiteral; |
|
exports.quoteAttrName = quoteAttrName; |
|
exports.trimLeft = trimLeft; |
|
|
|
},{"jstransform":22,"jstransform/src/utils":23}],38:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013-2015, Facebook, Inc. |
|
* All rights reserved. |
|
* |
|
* This source code is licensed under the BSD-style license found in the |
|
* LICENSE file in the root directory of this source tree. An additional grant |
|
* of patent rights can be found in the PATENTS file in the same directory. |
|
*/ |
|
/*global exports:true*/ |
|
'use strict'; |
|
|
|
var Syntax = _dereq_('jstransform').Syntax; |
|
var utils = _dereq_('jstransform/src/utils'); |
|
|
|
var renderJSXExpressionContainer = |
|
_dereq_('./jsx').renderJSXExpressionContainer; |
|
var renderJSXLiteral = _dereq_('./jsx').renderJSXLiteral; |
|
var quoteAttrName = _dereq_('./jsx').quoteAttrName; |
|
|
|
var trimLeft = _dereq_('./jsx').trimLeft; |
|
|
|
/** |
|
* Customized desugar processor for React JSX. Currently: |
|
* |
|
* <X> </X> => React.createElement(X, null) |
|
* <X prop="1" /> => React.createElement(X, {prop: '1'}, null) |
|
* <X prop="2"><Y /></X> => React.createElement(X, {prop:'2'}, |
|
* React.createElement(Y, null) |
|
* ) |
|
* <div /> => React.createElement("div", null) |
|
*/ |
|
|
|
/** |
|
* Removes all non-whitespace/parenthesis characters |
|
*/ |
|
var reNonWhiteParen = /([^\s\(\)])/g; |
|
function stripNonWhiteParen(value) { |
|
return value.replace(reNonWhiteParen, ''); |
|
} |
|
|
|
var tagConvention = /^[a-z]|\-/; |
|
function isTagName(name) { |
|
return tagConvention.test(name); |
|
} |
|
|
|
function visitReactTag(traverse, object, path, state) { |
|
var openingElement = object.openingElement; |
|
var nameObject = openingElement.name; |
|
var attributesObject = openingElement.attributes; |
|
|
|
utils.catchup(openingElement.range[0], state, trimLeft); |
|
|
|
if (nameObject.type === Syntax.JSXNamespacedName && nameObject.namespace) { |
|
throw new Error('Namespace tags are not supported. ReactJSX is not XML.'); |
|
} |
|
|
|
// We assume that the React runtime is already in scope |
|
utils.append('React.createElement(', state); |
|
|
|
if (nameObject.type === Syntax.JSXIdentifier && isTagName(nameObject.name)) { |
|
utils.append('"' + nameObject.name + '"', state); |
|
utils.move(nameObject.range[1], state); |
|
} else { |
|
// Use utils.catchup in this case so we can easily handle |
|
// JSXMemberExpressions which look like Foo.Bar.Baz. This also handles |
|
// JSXIdentifiers that aren't fallback tags. |
|
utils.move(nameObject.range[0], state); |
|
utils.catchup(nameObject.range[1], state); |
|
} |
|
|
|
utils.append(', ', state); |
|
|
|
var hasAttributes = attributesObject.length; |
|
|
|
var hasAtLeastOneSpreadProperty = attributesObject.some(function(attr) { |
|
return attr.type === Syntax.JSXSpreadAttribute; |
|
}); |
|
|
|
// if we don't have any attributes, pass in null |
|
if (hasAtLeastOneSpreadProperty) { |
|
utils.append('React.__spread({', state); |
|
} else if (hasAttributes) { |
|
utils.append('{', state); |
|
} else { |
|
utils.append('null', state); |
|
} |
|
|
|
// keep track of if the previous attribute was a spread attribute |
|
var previousWasSpread = false; |
|
|
|
// write attributes |
|
attributesObject.forEach(function(attr, index) { |
|
var isLast = index === attributesObject.length - 1; |
|
|
|
if (attr.type === Syntax.JSXSpreadAttribute) { |
|
// Close the previous object or initial object |
|
if (!previousWasSpread) { |
|
utils.append('}, ', state); |
|
} |
|
|
|
// Move to the expression start, ignoring everything except parenthesis |
|
// and whitespace. |
|
utils.catchup(attr.range[0], state, stripNonWhiteParen); |
|
// Plus 1 to skip `{`. |
|
utils.move(attr.range[0] + 1, state); |
|
utils.catchup(attr.argument.range[0], state, stripNonWhiteParen); |
|
|
|
traverse(attr.argument, path, state); |
|
|
|
utils.catchup(attr.argument.range[1], state); |
|
|
|
// Move to the end, ignoring parenthesis and the closing `}` |
|
utils.catchup(attr.range[1] - 1, state, stripNonWhiteParen); |
|
|
|
if (!isLast) { |
|
utils.append(', ', state); |
|
} |
|
|
|
utils.move(attr.range[1], state); |
|
|
|
previousWasSpread = true; |
|
|
|
return; |
|
} |
|
|
|
// If the next attribute is a spread, we're effective last in this object |
|
if (!isLast) { |
|
isLast = attributesObject[index + 1].type === Syntax.JSXSpreadAttribute; |
|
} |
|
|
|
if (attr.name.namespace) { |
|
throw new Error( |
|
'Namespace attributes are not supported. ReactJSX is not XML.'); |
|
} |
|
var name = attr.name.name; |
|
|
|
utils.catchup(attr.range[0], state, trimLeft); |
|
|
|
if (previousWasSpread) { |
|
utils.append('{', state); |
|
} |
|
|
|
utils.append(quoteAttrName(name), state); |
|
utils.append(': ', state); |
|
|
|
if (!attr.value) { |
|
state.g.buffer += 'true'; |
|
state.g.position = attr.name.range[1]; |
|
if (!isLast) { |
|
utils.append(', ', state); |
|
} |
|
} else { |
|
utils.move(attr.name.range[1], state); |
|
// Use catchupNewlines to skip over the '=' in the attribute |
|
utils.catchupNewlines(attr.value.range[0], state); |
|
if (attr.value.type === Syntax.Literal) { |
|
renderJSXLiteral(attr.value, isLast, state); |
|
} else { |
|
renderJSXExpressionContainer(traverse, attr.value, isLast, path, state); |
|
} |
|
} |
|
|
|
utils.catchup(attr.range[1], state, trimLeft); |
|
|
|
previousWasSpread = false; |
|
|
|
}); |
|
|
|
if (!openingElement.selfClosing) { |
|
utils.catchup(openingElement.range[1] - 1, state, trimLeft); |
|
utils.move(openingElement.range[1], state); |
|
} |
|
|
|
if (hasAttributes && !previousWasSpread) { |
|
utils.append('}', state); |
|
} |
|
|
|
if (hasAtLeastOneSpreadProperty) { |
|
utils.append(')', state); |
|
} |
|
|
|
// filter out whitespace |
|
var childrenToRender = object.children.filter(function(child) { |
|
return !(child.type === Syntax.Literal |
|
&& typeof child.value === 'string' |
|
&& child.value.match(/^[ \t]*[\r\n][ \t\r\n]*$/)); |
|
}); |
|
if (childrenToRender.length > 0) { |
|
var lastRenderableIndex; |
|
|
|
childrenToRender.forEach(function(child, index) { |
|
if (child.type !== Syntax.JSXExpressionContainer || |
|
child.expression.type !== Syntax.JSXEmptyExpression) { |
|
lastRenderableIndex = index; |
|
} |
|
}); |
|
|
|
if (lastRenderableIndex !== undefined) { |
|
utils.append(', ', state); |
|
} |
|
|
|
childrenToRender.forEach(function(child, index) { |
|
utils.catchup(child.range[0], state, trimLeft); |
|
|
|
var isLast = index >= lastRenderableIndex; |
|
|
|
if (child.type === Syntax.Literal) { |
|
renderJSXLiteral(child, isLast, state); |
|
} else if (child.type === Syntax.JSXExpressionContainer) { |
|
renderJSXExpressionContainer(traverse, child, isLast, path, state); |
|
} else { |
|
traverse(child, path, state); |
|
if (!isLast) { |
|
utils.append(', ', state); |
|
} |
|
} |
|
|
|
utils.catchup(child.range[1], state, trimLeft); |
|
}); |
|
} |
|
|
|
if (openingElement.selfClosing) { |
|
// everything up to /> |
|
utils.catchup(openingElement.range[1] - 2, state, trimLeft); |
|
utils.move(openingElement.range[1], state); |
|
} else { |
|
// everything up to </ sdflksjfd> |
|
utils.catchup(object.closingElement.range[0], state, trimLeft); |
|
utils.move(object.closingElement.range[1], state); |
|
} |
|
|
|
utils.append(')', state); |
|
return false; |
|
} |
|
|
|
visitReactTag.test = function(object, path, state) { |
|
return object.type === Syntax.JSXElement; |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitReactTag |
|
]; |
|
|
|
},{"./jsx":37,"jstransform":22,"jstransform/src/utils":23}],39:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013-2015, Facebook, Inc. |
|
* All rights reserved. |
|
* |
|
* This source code is licensed under the BSD-style license found in the |
|
* LICENSE file in the root directory of this source tree. An additional grant |
|
* of patent rights can be found in the PATENTS file in the same directory. |
|
*/ |
|
/*global exports:true*/ |
|
'use strict'; |
|
|
|
var Syntax = _dereq_('jstransform').Syntax; |
|
var utils = _dereq_('jstransform/src/utils'); |
|
|
|
function addDisplayName(displayName, object, state) { |
|
if (object && |
|
object.type === Syntax.CallExpression && |
|
object.callee.type === Syntax.MemberExpression && |
|
object.callee.object.type === Syntax.Identifier && |
|
object.callee.object.name === 'React' && |
|
object.callee.property.type === Syntax.Identifier && |
|
object.callee.property.name === 'createClass' && |
|
object.arguments.length === 1 && |
|
object.arguments[0].type === Syntax.ObjectExpression) { |
|
// Verify that the displayName property isn't already set |
|
var properties = object.arguments[0].properties; |
|
var safe = properties.every(function(property) { |
|
var value = property.key.type === Syntax.Identifier ? |
|
property.key.name : |
|
property.key.value; |
|
return value !== 'displayName'; |
|
}); |
|
|
|
if (safe) { |
|
utils.catchup(object.arguments[0].range[0] + 1, state); |
|
utils.append('displayName: "' + displayName + '",', state); |
|
} |
|
} |
|
} |
|
|
|
/** |
|
* Transforms the following: |
|
* |
|
* var MyComponent = React.createClass({ |
|
* render: ... |
|
* }); |
|
* |
|
* into: |
|
* |
|
* var MyComponent = React.createClass({ |
|
* displayName: 'MyComponent', |
|
* render: ... |
|
* }); |
|
* |
|
* Also catches: |
|
* |
|
* MyComponent = React.createClass(...); |
|
* exports.MyComponent = React.createClass(...); |
|
* module.exports = {MyComponent: React.createClass(...)}; |
|
*/ |
|
function visitReactDisplayName(traverse, object, path, state) { |
|
var left, right; |
|
|
|
if (object.type === Syntax.AssignmentExpression) { |
|
left = object.left; |
|
right = object.right; |
|
} else if (object.type === Syntax.Property) { |
|
left = object.key; |
|
right = object.value; |
|
} else if (object.type === Syntax.VariableDeclarator) { |
|
left = object.id; |
|
right = object.init; |
|
} |
|
|
|
if (left && left.type === Syntax.MemberExpression) { |
|
left = left.property; |
|
} |
|
if (left && left.type === Syntax.Identifier) { |
|
addDisplayName(left.name, right, state); |
|
} |
|
} |
|
|
|
visitReactDisplayName.test = function(object, path, state) { |
|
return ( |
|
object.type === Syntax.AssignmentExpression || |
|
object.type === Syntax.Property || |
|
object.type === Syntax.VariableDeclarator |
|
); |
|
}; |
|
|
|
exports.visitorList = [ |
|
visitReactDisplayName |
|
]; |
|
|
|
},{"jstransform":22,"jstransform/src/utils":23}],40:[function(_dereq_,module,exports){ |
|
/*global exports:true*/ |
|
|
|
'use strict'; |
|
|
|
var es6ArrowFunctions = |
|
_dereq_('jstransform/visitors/es6-arrow-function-visitors'); |
|
var es6Classes = _dereq_('jstransform/visitors/es6-class-visitors'); |
|
var es6Destructuring = |
|
_dereq_('jstransform/visitors/es6-destructuring-visitors'); |
|
var es6ObjectConciseMethod = |
|
_dereq_('jstransform/visitors/es6-object-concise-method-visitors'); |
|
var es6ObjectShortNotation = |
|
_dereq_('jstransform/visitors/es6-object-short-notation-visitors'); |
|
var es6RestParameters = _dereq_('jstransform/visitors/es6-rest-param-visitors'); |
|
var es6Templates = _dereq_('jstransform/visitors/es6-template-visitors'); |
|
var es6CallSpread = |
|
_dereq_('jstransform/visitors/es6-call-spread-visitors'); |
|
var es7SpreadProperty = |
|
_dereq_('jstransform/visitors/es7-spread-property-visitors'); |
|
var react = _dereq_('./transforms/react'); |
|
var reactDisplayName = _dereq_('./transforms/reactDisplayName'); |
|
var reservedWords = _dereq_('jstransform/visitors/reserved-words-visitors'); |
|
|
|
/** |
|
* Map from transformName => orderedListOfVisitors. |
|
*/ |
|
var transformVisitors = { |
|
'es6-arrow-functions': es6ArrowFunctions.visitorList, |
|
'es6-classes': es6Classes.visitorList, |
|
'es6-destructuring': es6Destructuring.visitorList, |
|
'es6-object-concise-method': es6ObjectConciseMethod.visitorList, |
|
'es6-object-short-notation': es6ObjectShortNotation.visitorList, |
|
'es6-rest-params': es6RestParameters.visitorList, |
|
'es6-templates': es6Templates.visitorList, |
|
'es6-call-spread': es6CallSpread.visitorList, |
|
'es7-spread-property': es7SpreadProperty.visitorList, |
|
'react': react.visitorList.concat(reactDisplayName.visitorList), |
|
'reserved-words': reservedWords.visitorList |
|
}; |
|
|
|
var transformSets = { |
|
'harmony': [ |
|
'es6-arrow-functions', |
|
'es6-object-concise-method', |
|
'es6-object-short-notation', |
|
'es6-classes', |
|
'es6-rest-params', |
|
'es6-templates', |
|
'es6-destructuring', |
|
'es6-call-spread', |
|
'es7-spread-property' |
|
], |
|
'es3': [ |
|
'reserved-words' |
|
], |
|
'react': [ |
|
'react' |
|
] |
|
}; |
|
|
|
/** |
|
* Specifies the order in which each transform should run. |
|
*/ |
|
var transformRunOrder = [ |
|
'reserved-words', |
|
'es6-arrow-functions', |
|
'es6-object-concise-method', |
|
'es6-object-short-notation', |
|
'es6-classes', |
|
'es6-rest-params', |
|
'es6-templates', |
|
'es6-destructuring', |
|
'es6-call-spread', |
|
'es7-spread-property', |
|
'react' |
|
]; |
|
|
|
/** |
|
* Given a list of transform names, return the ordered list of visitors to be |
|
* passed to the transform() function. |
|
* |
|
* @param {array?} excludes |
|
* @return {array} |
|
*/ |
|
function getAllVisitors(excludes) { |
|
var ret = []; |
|
for (var i = 0, il = transformRunOrder.length; i < il; i++) { |
|
if (!excludes || excludes.indexOf(transformRunOrder[i]) === -1) { |
|
ret = ret.concat(transformVisitors[transformRunOrder[i]]); |
|
} |
|
} |
|
return ret; |
|
} |
|
|
|
/** |
|
* Given a list of visitor set names, return the ordered list of visitors to be |
|
* passed to jstransform. |
|
* |
|
* @param {array} |
|
* @return {array} |
|
*/ |
|
function getVisitorsBySet(sets) { |
|
var visitorsToInclude = sets.reduce(function(visitors, set) { |
|
if (!transformSets.hasOwnProperty(set)) { |
|
throw new Error('Unknown visitor set: ' + set); |
|
} |
|
transformSets[set].forEach(function(visitor) { |
|
visitors[visitor] = true; |
|
}); |
|
return visitors; |
|
}, {}); |
|
|
|
var visitorList = []; |
|
for (var i = 0; i < transformRunOrder.length; i++) { |
|
if (visitorsToInclude.hasOwnProperty(transformRunOrder[i])) { |
|
visitorList = visitorList.concat(transformVisitors[transformRunOrder[i]]); |
|
} |
|
} |
|
|
|
return visitorList; |
|
} |
|
|
|
exports.getVisitorsBySet = getVisitorsBySet; |
|
exports.getAllVisitors = getAllVisitors; |
|
exports.transformVisitors = transformVisitors; |
|
|
|
},{"./transforms/react":38,"./transforms/reactDisplayName":39,"jstransform/visitors/es6-arrow-function-visitors":24,"jstransform/visitors/es6-call-spread-visitors":25,"jstransform/visitors/es6-class-visitors":26,"jstransform/visitors/es6-destructuring-visitors":27,"jstransform/visitors/es6-object-concise-method-visitors":28,"jstransform/visitors/es6-object-short-notation-visitors":29,"jstransform/visitors/es6-rest-param-visitors":30,"jstransform/visitors/es6-template-visitors":31,"jstransform/visitors/es7-spread-property-visitors":33,"jstransform/visitors/reserved-words-visitors":35}],41:[function(_dereq_,module,exports){ |
|
/** |
|
* Copyright 2013-2015, Facebook, Inc. |
|
* All rights reserved. |
|
* |
|
* This source code is licensed under the BSD-style license found in the |
|
* LICENSE file in the root directory of this source tree. An additional grant |
|
* of patent rights can be found in the PATENTS file in the same directory. |
|
*/ |
|
|
|
'use strict'; |
|
/*eslint-disable no-undef*/ |
|
var Buffer = _dereq_('buffer').Buffer; |
|
|
|
function inlineSourceMap(sourceMap, sourceCode, sourceFilename) { |
|
// This can be used with a sourcemap that has already has toJSON called on it. |
|
// Check first. |
|
var json = sourceMap; |
|
if (typeof sourceMap.toJSON === 'function') { |
|
json = sourceMap.toJSON(); |
|
} |
|
json.sources = [sourceFilename]; |
|
json.sourcesContent = [sourceCode]; |
|
var base64 = Buffer(JSON.stringify(json)).toString('base64'); |
|
return '//# sourceMappingURL=data:application/json;base64,' + base64; |
|
} |
|
|
|
module.exports = inlineSourceMap; |
|
|
|
},{"buffer":3}]},{},[1])(1) |
|
}); |