Kevacoin source tree
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.
 
 
 
 
 
 

303 lines
8.6 KiB

#!/usr/bin/env python
#
# linearize-data.py: Construct a linear, no-fork version of the chain.
#
# Copyright (c) 2013-2014 The Bitcoin Core developers
# Distributed under the MIT software license, see the accompanying
# file COPYING or http://www.opensource.org/licenses/mit-license.php.
#
from __future__ import print_function, division
import json
import struct
import re
import os
import os.path
import base64
import httplib
import sys
import hashlib
import datetime
import time
from collections import namedtuple
settings = {}
def uint32(x):
return x & 0xffffffffL
def bytereverse(x):
return uint32(( ((x) << 24) | (((x) << 8) & 0x00ff0000) |
(((x) >> 8) & 0x0000ff00) | ((x) >> 24) ))
def bufreverse(in_buf):
out_words = []
for i in range(0, len(in_buf), 4):
word = struct.unpack('@I', in_buf[i:i+4])[0]
out_words.append(struct.pack('@I', bytereverse(word)))
return ''.join(out_words)
def wordreverse(in_buf):
out_words = []
for i in range(0, len(in_buf), 4):
out_words.append(in_buf[i:i+4])
out_words.reverse()
return ''.join(out_words)
def calc_hdr_hash(blk_hdr):
hash1 = hashlib.sha256()
hash1.update(blk_hdr)
hash1_o = hash1.digest()
hash2 = hashlib.sha256()
hash2.update(hash1_o)
hash2_o = hash2.digest()
return hash2_o
def calc_hash_str(blk_hdr):
hash = calc_hdr_hash(blk_hdr)
hash = bufreverse(hash)
hash = wordreverse(hash)
hash_str = hash.encode('hex')
return hash_str
def get_blk_dt(blk_hdr):
members = struct.unpack("<I", blk_hdr[68:68+4])
nTime = members[0]
dt = datetime.datetime.fromtimestamp(nTime)
dt_ym = datetime.datetime(dt.year, dt.month, 1)
return (dt_ym, nTime)
def get_block_hashes(settings):
blkindex = []
f = open(settings['hashlist'], "r")
for line in f:
line = line.rstrip()
blkindex.append(line)
print("Read " + str(len(blkindex)) + " hashes")
return blkindex
def mkblockmap(blkindex):
blkmap = {}
for height,hash in enumerate(blkindex):
blkmap[hash] = height
return blkmap
# Block header and extent on disk
BlockExtent = namedtuple('BlockExtent', ['fn', 'offset', 'inhdr', 'blkhdr', 'size'])
class BlockDataCopier:
def __init__(self, settings, blkindex, blkmap):
self.settings = settings
self.blkindex = blkindex
self.blkmap = blkmap
self.inFn = 0
self.inF = None
self.outFn = 0
self.outsz = 0
self.outF = None
self.outFname = None
self.blkCountIn = 0
self.blkCountOut = 0
self.lastDate = datetime.datetime(2000, 1, 1)
self.highTS = 1408893517 - 315360000
self.timestampSplit = False
self.fileOutput = True
self.setFileTime = False
self.maxOutSz = settings['max_out_sz']
if 'output' in settings:
self.fileOutput = False
if settings['file_timestamp'] != 0:
self.setFileTime = True
if settings['split_timestamp'] != 0:
self.timestampSplit = True
# Extents and cache for out-of-order blocks
self.blockExtents = {}
self.outOfOrderData = {}
self.outOfOrderSize = 0 # running total size for items in outOfOrderData
def writeBlock(self, inhdr, blk_hdr, rawblock):
blockSizeOnDisk = len(inhdr) + len(blk_hdr) + len(rawblock)
if not self.fileOutput and ((self.outsz + blockSizeOnDisk) > self.maxOutSz):
self.outF.close()
if self.setFileTime:
os.utime(outFname, (int(time.time()), highTS))
self.outF = None
self.outFname = None
self.outFn = self.outFn + 1
self.outsz = 0
(blkDate, blkTS) = get_blk_dt(blk_hdr)
if self.timestampSplit and (blkDate > self.lastDate):
print("New month " + blkDate.strftime("%Y-%m") + " @ " + hash_str)
lastDate = blkDate
if outF:
outF.close()
if setFileTime:
os.utime(outFname, (int(time.time()), highTS))
self.outF = None
self.outFname = None
self.outFn = self.outFn + 1
self.outsz = 0
if not self.outF:
if self.fileOutput:
outFname = self.settings['output_file']
else:
outFname = os.path.join(self.settings['output'], "blk%05d.dat" % self.outFn)
print("Output file " + outFname)
self.outF = open(outFname, "wb")
self.outF.write(inhdr)
self.outF.write(blk_hdr)
self.outF.write(rawblock)
self.outsz = self.outsz + len(inhdr) + len(blk_hdr) + len(rawblock)
self.blkCountOut = self.blkCountOut + 1
if blkTS > self.highTS:
self.highTS = blkTS
if (self.blkCountOut % 1000) == 0:
print('%i blocks scanned, %i blocks written (of %i, %.1f%% complete)' %
(self.blkCountIn, self.blkCountOut, len(self.blkindex), 100.0 * self.blkCountOut / len(self.blkindex)))
def inFileName(self, fn):
return os.path.join(self.settings['input'], "blk%05d.dat" % fn)
def fetchBlock(self, extent):
'''Fetch block contents from disk given extents'''
with open(self.inFileName(extent.fn), "rb") as f:
f.seek(extent.offset)
return f.read(extent.size)
def copyOneBlock(self):
'''Find the next block to be written in the input, and copy it to the output.'''
extent = self.blockExtents.pop(self.blkCountOut)
if self.blkCountOut in self.outOfOrderData:
# If the data is cached, use it from memory and remove from the cache
rawblock = self.outOfOrderData.pop(self.blkCountOut)
self.outOfOrderSize -= len(rawblock)
else: # Otherwise look up data on disk
rawblock = self.fetchBlock(extent)
self.writeBlock(extent.inhdr, extent.blkhdr, rawblock)
def run(self):
while self.blkCountOut < len(self.blkindex):
if not self.inF:
fname = self.inFileName(self.inFn)
print("Input file " + fname)
try:
self.inF = open(fname, "rb")
except IOError:
print("Premature end of block data")
return
inhdr = self.inF.read(8)
if (not inhdr or (inhdr[0] == "\0")):
self.inF.close()
self.inF = None
self.inFn = self.inFn + 1
continue
inMagic = inhdr[:4]
if (inMagic != self.settings['netmagic']):
print("Invalid magic: " + inMagic.encode('hex'))
return
inLenLE = inhdr[4:]
su = struct.unpack("<I", inLenLE)
inLen = su[0] - 80 # length without header
blk_hdr = self.inF.read(80)
inExtent = BlockExtent(self.inFn, self.inF.tell(), inhdr, blk_hdr, inLen)
hash_str = calc_hash_str(blk_hdr)
if not hash_str in blkmap:
print("Skipping unknown block " + hash_str)
self.inF.seek(inLen, os.SEEK_CUR)
continue
blkHeight = self.blkmap[hash_str]
self.blkCountIn += 1
if self.blkCountOut == blkHeight:
# If in-order block, just copy
rawblock = self.inF.read(inLen)
self.writeBlock(inhdr, blk_hdr, rawblock)
# See if we can catch up to prior out-of-order blocks
while self.blkCountOut in self.blockExtents:
self.copyOneBlock()
else: # If out-of-order, skip over block data for now
self.blockExtents[blkHeight] = inExtent
if self.outOfOrderSize < self.settings['out_of_order_cache_sz']:
# If there is space in the cache, read the data
# Reading the data in file sequence instead of seeking and fetching it later is preferred,
# but we don't want to fill up memory
self.outOfOrderData[blkHeight] = self.inF.read(inLen)
self.outOfOrderSize += inLen
else: # If no space in cache, seek forward
self.inF.seek(inLen, os.SEEK_CUR)
print("Done (%i blocks written)" % (self.blkCountOut))
if __name__ == '__main__':
if len(sys.argv) != 2:
print("Usage: linearize-data.py CONFIG-FILE")
sys.exit(1)
f = open(sys.argv[1])
for line in f:
# skip comment lines
m = re.search('^\s*#', line)
if m:
continue
# parse key=value lines
m = re.search('^(\w+)\s*=\s*(\S.*)$', line)
if m is None:
continue
settings[m.group(1)] = m.group(2)
f.close()
if 'netmagic' not in settings:
settings['netmagic'] = 'f9beb4d9'
if 'genesis' not in settings:
settings['genesis'] = '000000000019d6689c085ae165831e934ff763ae46a2a6c172b3f1b60a8ce26f'
if 'input' not in settings:
settings['input'] = 'input'
if 'hashlist' not in settings:
settings['hashlist'] = 'hashlist.txt'
if 'file_timestamp' not in settings:
settings['file_timestamp'] = 0
if 'split_timestamp' not in settings:
settings['split_timestamp'] = 0
if 'max_out_sz' not in settings:
settings['max_out_sz'] = 1000L * 1000 * 1000
if 'out_of_order_cache_sz' not in settings:
settings['out_of_order_cache_sz'] = 100 * 1000 * 1000
settings['max_out_sz'] = long(settings['max_out_sz'])
settings['split_timestamp'] = int(settings['split_timestamp'])
settings['file_timestamp'] = int(settings['file_timestamp'])
settings['netmagic'] = settings['netmagic'].decode('hex')
settings['out_of_order_cache_sz'] = int(settings['out_of_order_cache_sz'])
if 'output_file' not in settings and 'output' not in settings:
print("Missing output file / directory")
sys.exit(1)
blkindex = get_block_hashes(settings)
blkmap = mkblockmap(blkindex)
if not settings['genesis'] in blkmap:
print("Genesis block not found in hashlist")
else:
BlockDataCopier(settings, blkindex, blkmap).run()