1
0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-12-23 08:26:50 +00:00

able to download first part of the video, but fail in the left part

This commit is contained in:
Celthi 2015-12-11 17:48:40 +08:00
parent 51094b1b08
commit fdf01663d1

View File

@ -1,6 +1,7 @@
# coding: utf-8 # coding: utf-8
from __future__ import unicode_literals from __future__ import unicode_literals
import json
import base64 import base64
from .common import InfoExtractor from .common import InfoExtractor
@ -84,17 +85,18 @@ class YoukuIE(InfoExtractor):
return bytes(s) return bytes(s)
sid, token = yk_t( sid, token = yk_t(
b'becaf9be', base64.b64decode(data2['ep'].encode('ascii')) b'becaf9be', base64.b64decode(data2['security']['encrypt_string'].encode('ascii'))
).decode('ascii').split('_') ).decode('ascii').split('_')
# get oip # get oip
oip = data2['ip'] oip = data2['security']['ip']
# get fileid # get fileid
string_ls = list( string_ls = list(
'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ/\:._-1234567890') 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ/\:._-1234567890')
shuffled_string_ls = [] shuffled_string_ls = []
seed = data1['seed'] seed = data1.get('seed')
seed = 0
N = len(string_ls) N = len(string_ls)
for ii in range(N): for ii in range(N):
seed = (seed * 0xd3 + 0x754f) % 0x10000 seed = (seed * 0xd3 + 0x754f) % 0x10000
@ -103,15 +105,19 @@ class YoukuIE(InfoExtractor):
del string_ls[idx] del string_ls[idx]
fileid_dict = {} fileid_dict = {}
for format in data1['streamtypes']: for stream in data1['stream']:
streamfileid = [ format = stream.get('stream_type')
int(i) for i in data1['streamfileids'][format].strip('*').split('*')] #streamfileid = [
fileid = ''.join( # int(i) for i in data1['stream']['streamfileids']#[format].strip('*').split('*')]
[shuffled_string_ls[i] for i in streamfileid]) #fileid = ''.join(
fileid_dict[format] = fileid[:8] + '%s' + fileid[10:] # [shuffled_string_ls[i] for i in streamfileid])
fileid = stream['stream_fileid']
fileid_dict[format] = fileid
#fileid_dict[format] = fileid[:8] + '%s' + fileid[10:]
def get_fileid(format, n): def get_fileid(format, n):
fileid = fileid_dict[format] % hex(int(n))[2:].upper().zfill(2) #fileid = fileid_dict[format] % hex(int(n))[2:].upper().zfill(2)
fileid = fileid_dict[format]
return fileid return fileid
# get ep # get ep
@ -126,15 +132,16 @@ class YoukuIE(InfoExtractor):
# generate video_urls # generate video_urls
video_urls_dict = {} video_urls_dict = {}
for format in data1['streamtypes']: for stream in data1['stream']:
format = stream.get('stream_type')
video_urls = [] video_urls = []
for dt in data1['segs'][format]: for dt in stream['segs']:
n = str(int(dt['no'])) n = str(int(dt['size']))
param = { param = {
'K': dt['k'], 'K': dt['key'],
'hd': self.get_hd(format), 'hd': self.get_hd(format),
'myp': 0, 'myp': 0,
'ts': dt['seconds'], 'ts': dt['total_milliseconds_video'],
'ypp': 0, 'ypp': 0,
'ctype': 12, 'ctype': 12,
'ev': 1, 'ev': 1,
@ -161,7 +168,8 @@ class YoukuIE(InfoExtractor):
'hd2': '2', 'hd2': '2',
'hd3': '3', 'hd3': '3',
'3gp': '0', '3gp': '0',
'3gphd': '1' '3gphd': '1',
'flvhd': '0'
} }
return hd_id_dict[fm] return hd_id_dict[fm]
@ -172,7 +180,8 @@ class YoukuIE(InfoExtractor):
'hd2': 'flv', 'hd2': 'flv',
'hd3': 'flv', 'hd3': 'flv',
'3gp': 'flv', '3gp': 'flv',
'3gphd': 'mp4' '3gphd': 'mp4',
'flvhd': 'flv'
} }
return ext_dict[fm] return ext_dict[fm]
@ -180,6 +189,7 @@ class YoukuIE(InfoExtractor):
_dict = { _dict = {
'3gp': 'h6', '3gp': 'h6',
'3gphd': 'h5', '3gphd': 'h5',
'flvhd': 'h4',
'flv': 'h4', 'flv': 'h4',
'mp4': 'h3', 'mp4': 'h3',
'hd2': 'h2', 'hd2': 'h2',
@ -204,8 +214,9 @@ class YoukuIE(InfoExtractor):
req.add_header('Ytdl-request-proxy', cn_verification_proxy) req.add_header('Ytdl-request-proxy', cn_verification_proxy)
raw_data = self._download_json(req, video_id, note=note) raw_data = self._download_json(req, video_id, note=note)
jsonDumpIn = json.dumps(raw_data,indent = 1)
return raw_data['data']['security'] return raw_data['data']
video_password = self._downloader.params.get('videopassword', None) video_password = self._downloader.params.get('videopassword', None)
@ -236,7 +247,8 @@ class YoukuIE(InfoExtractor):
msg += ': ' + error msg += ': ' + error
raise ExtractorError(msg) raise ExtractorError(msg)
title = data1['title'] #title = data1['title']
title = data1['video']['title']
# generate video_urls_dict # generate video_urls_dict
video_urls_dict = self.construct_video_urls(data1, data2) video_urls_dict = self.construct_video_urls(data1, data2)
@ -248,10 +260,11 @@ class YoukuIE(InfoExtractor):
'formats': [], 'formats': [],
# some formats are not available for all parts, we have to detect # some formats are not available for all parts, we have to detect
# which one has all # which one has all
} for i in range(max(len(v) for v in data1['segs'].values()))] } for i in range(max(len(v) for v in data1['stream']))]
for fm in data1['streamtypes']: for stream in data1['stream']:
fm = stream.get('stream_type')
video_urls = video_urls_dict[fm] video_urls = video_urls_dict[fm]
for video_url, seg, entry in zip(video_urls, data1['segs'][fm], entries): for video_url, seg, entry in zip(video_urls, stream['segs'], entries):
entry['formats'].append({ entry['formats'].append({
'url': video_url, 'url': video_url,
'format_id': self.get_format_name(fm), 'format_id': self.get_format_name(fm),