openmedialibrary_platform/Shared/lib/python2.7/site-packages/ox/file.py

321 lines
10 KiB
Python
Raw Normal View History

2013-10-11 17:28:32 +00:00
# -*- coding: utf-8 -*-
# vi:si:et:sw=4:sts=4:ts=4
# GPL 2008
from __future__ import division, with_statement
import os
import hashlib
import re
import shutil
import struct
import subprocess
import sqlite3
from ox.utils import json
__all__ = ['sha1sum', 'oshash', 'avinfo', 'makedirs']
EXTENSIONS = {
'audio': [
'aac', 'aif', 'aiff',
'flac', 'm4a', 'mp3', 'oga', 'ogg', 'wav', 'wma'
2013-10-11 17:28:32 +00:00
],
'image': [
'bmp', 'gif', 'jpeg', 'jpg', 'png', 'svg', 'webp'
],
'subtitle': [
'idx', 'srt', 'sub'
],
'video': [
'3gp',
'avi', 'divx', 'dv', 'flv', 'm2t', 'm4v', 'mkv', 'mov', 'mp4',
'mpeg', 'mpg', 'mts', 'ogm', 'ogv', 'rm', 'vob', 'webm', 'wmv',
'mod', 'tod', # http://en.wikipedia.org/wiki/MOD_and_TOD
'mxf'
2013-10-11 17:28:32 +00:00
],
}
def cmd(program):
local = os.path.expanduser('~/.ox/bin/%s' % program)
if os.path.exists(local):
program = local
return program
def _get_file_cache():
import ox.cache
path = ox.cache.cache_path()
if path.startswith('fs:'):
path = path[3:]
return os.path.join(path, 'files.sqlite')
2013-10-11 17:28:32 +00:00
def cache(filename, type='oshash'):
conn = sqlite3.connect(_get_file_cache(), timeout=10)
conn.text_factory = str
conn.row_factory = sqlite3.Row
if not cache.init:
c = conn.cursor()
c.execute('CREATE TABLE IF NOT EXISTS cache (path varchar(1024) unique, oshash varchar(16), sha1 varchar(42), size int, mtime int, info text)')
c.execute('CREATE INDEX IF NOT EXISTS cache_oshash ON cache (oshash)')
c.execute('CREATE INDEX IF NOT EXISTS cache_sha1 ON cache (sha1)')
conn.commit()
cache.init = True
c = conn.cursor()
c.execute('SELECT oshash, sha1, info, size, mtime FROM cache WHERE path = ?', (filename, ))
stat = os.stat(filename)
row = None
h = None
sha1 = None
info = ''
for row in c:
if stat.st_size == row['size'] and int(stat.st_mtime) == int(row['mtime']):
value = row[type]
if value:
if type == 'info':
value = json.loads(value)
return value
h = row['oshash']
sha1 = row['sha1']
info = row['info']
if type == 'oshash':
value = h = oshash(filename, cached=False)
elif type == 'sha1':
value = sha1 = sha1sum(filename, cached=False)
elif type == 'info':
value = avinfo(filename, cached=False)
info = json.dumps(value)
t = (filename, h, sha1, stat.st_size, int(stat.st_mtime), info)
with conn:
sql = u'INSERT OR REPLACE INTO cache values (?, ?, ?, ?, ?, ?)'
c.execute(sql, t)
return value
cache.init = None
def cleanup_cache():
conn = sqlite3.connect(_get_file_cache(), timeout=10)
conn.text_factory = str
conn.row_factory = sqlite3.Row
c = conn.cursor()
c.execute('SELECT path FROM cache')
paths = [r[0] for r in c]
for path in paths:
if not os.path.exists(path):
c.execute('DELETE FROM cache WHERE path = ?', (path, ))
conn.commit()
c.execute('VACUUM')
conn.commit()
def sha1sum(filename, cached=False):
if cached:
return cache(filename, 'sha1')
sha1 = hashlib.sha1()
with open(filename) as f:
for chunk in iter(lambda: f.read(128*sha1.block_size), ''):
sha1.update(chunk)
return sha1.hexdigest()
'''
os hash - http://trac.opensubtitles.org/projects/opensubtitles/wiki/HashSourceCodes
plus modification for files < 64k, buffer is filled with file data and padded with 0
'''
def oshash(filename, cached=True):
if cached:
return cache(filename, 'oshash')
try:
longlongformat = 'q' # long long
bytesize = struct.calcsize(longlongformat)
f = open(filename, "rb")
filesize = os.path.getsize(filename)
hash = filesize
if filesize < 65536:
for x in range(int(filesize/bytesize)):
buffer = f.read(bytesize)
(l_value,)= struct.unpack(longlongformat, buffer)
hash += l_value
hash = hash & 0xFFFFFFFFFFFFFFFF #to remain as 64bit number
else:
for x in range(int(65536/bytesize)):
buffer = f.read(bytesize)
(l_value,)= struct.unpack(longlongformat, buffer)
hash += l_value
hash = hash & 0xFFFFFFFFFFFFFFFF #to remain as 64bit number
f.seek(max(0,filesize-65536),0)
for x in range(int(65536/bytesize)):
buffer = f.read(bytesize)
(l_value,)= struct.unpack(longlongformat, buffer)
hash += l_value
hash = hash & 0xFFFFFFFFFFFFFFFF
f.close()
returnedhash = "%016x" % hash
return returnedhash
except(IOError):
return "IOError"
def avinfo(filename, cached=True):
if cached:
return cache(filename, 'info')
if os.path.getsize(filename):
ffmpeg2theora = cmd('ffmpeg2theora')
p = subprocess.Popen([ffmpeg2theora], stdout=subprocess.PIPE, stderr=subprocess.PIPE)
info, error = p.communicate()
version = info.split('\n')[0].split(' - ')[0].split(' ')[-1]
if version < '0.27':
raise EnvironmentError('version of ffmpeg2theora needs to be 0.27 or later, found %s' % version)
p = subprocess.Popen([ffmpeg2theora, '--info', filename],
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
info, error = p.communicate()
try:
info = json.loads(info)
except:
#remove metadata, can be broken
reg = re.compile('"metadata": {.*?},', re.DOTALL)
info = re.sub(reg, '', info)
info = json.loads(info)
if 'video' in info:
for v in info['video']:
if not 'display_aspect_ratio' in v and 'width' in v:
v['display_aspect_ratio'] = '%d:%d' % (v['width'], v['height'])
v['pixel_aspect_ratio'] = '1:1'
return info
return {'path': filename, 'size': 0}
def ffprobe(filename):
p = subprocess.Popen([
cmd('ffprobe'),
'-show_format',
'-show_streams',
'-print_format',
'json',
'-i', filename
], stdout=subprocess.PIPE, stderr=subprocess.PIPE)
info, error = p.communicate()
ffinfo = json.loads(info)
def fix_value(key, value):
if key == 'r_frame_rate':
value = value.replace('/', ':')
elif key == 'bit_rate':
value = float(value) / 1000
elif key == 'duration':
value = float(value)
elif key == 'size':
value = int(value)
return value
info = {}
for key in ('duration', 'size', 'bit_rate'):
info[{
'bit_rate': 'bitrate'
}.get(key, key)] = fix_value(key, ffinfo['format'][key])
info['audio'] = []
info['video'] = []
info['metadata'] = ffinfo['format'].get('tags', {})
for s in ffinfo['streams']:
tags = s.pop('tags', {})
for t in tags:
info['metadata'][t] = tags[t]
if s.get('codec_type') in ('audio', 'video'):
stream = {}
keys = [
'codec_name',
'width',
'height',
'bit_rate',
'index',
'display_aspect_ratio',
'sample_rate',
'channels',
]
if s['codec_type'] == 'video':
keys += [
'sample_aspect_ratio',
'r_frame_rate',
'pix_fmt',
]
for key in keys:
if key in s:
stream[{
'codec_name': 'codec',
'bit_rate': 'bitrate',
'index': 'id',
'r_frame_rate': 'framerate',
'sample_rate': 'samplerate',
'pix_fmt': 'pixel_format',
}.get(key, key)] = fix_value(key, s[key])
info[s['codec_type']].append(stream)
else:
pass
#print s
for v in info['video']:
if not 'display_aspect_ratio' in v and 'width' in v:
v['display_aspect_ratio'] = '%d:%d' % (v['width'], v['height'])
v['pixel_aspect_ratio'] = '1:1'
info['oshash'] = oshash(filename)
info['path'] = os.path.basename(filename)
return info
def makedirs(path):
if not os.path.exists(path):
try:
os.makedirs(path)
except OSError, e:
if e.errno != 17:
raise
def copy_file(source, target, verbose=False):
if verbose:
print 'copying', source, 'to', target
write_path(target)
shutil.copyfile(source, target)
def read_file(file, verbose=False):
if verbose:
print 'reading', file
f = open(file)
data = f.read()
f.close()
return data
def read_json(file, verbose=False):
if verbose:
print 'reading', file
with open(file) as fd:
data = json.load(fd)
return data
def write_file(file, data, verbose=False):
if verbose:
print 'writing', file
write_path(file)
f = open(file, 'w')
f.write(data)
f.close()
return len(data)
def write_image(file, image, verbose=False):
if verbose:
print 'writing', file
write_path(file)
image.save(file)
def write_json(file, data, ensure_ascii=True, indent=0, sort_keys=False, verbose=False):
data = json.dumps(data, ensure_ascii=ensure_ascii, indent=indent, sort_keys=sort_keys)
write_file(file, data if ensure_ascii else data.encode('utf-8'), verbose=verbose)
def write_link(source, target, verbose=False):
if verbose:
print 'linking', source, 'to', target
write_path(target)
if os.path.exists(target):
os.unlink(target)
os.symlink(source, target)
def write_path(file):
path = os.path.split(file)[0]
if path and not os.path.exists(path):
os.makedirs(path)