mirror of
https://codeberg.org/polarisfm/youtube-dl
synced 2024-11-16 06:34:31 +01:00
36e6f62cd0
Attributes aren't unicode objects, so they couldn't be directly used in info_dict fields (for example '--write-description' doesn't work with bytes).
392 lines
13 KiB
Python
392 lines
13 KiB
Python
from __future__ import division, unicode_literals
|
|
|
|
import base64
|
|
import io
|
|
import itertools
|
|
import os
|
|
import time
|
|
|
|
from .fragment import FragmentFD
|
|
from ..compat import (
|
|
compat_etree_fromstring,
|
|
compat_urlparse,
|
|
compat_urllib_error,
|
|
compat_urllib_parse_urlparse,
|
|
)
|
|
from ..utils import (
|
|
encodeFilename,
|
|
sanitize_open,
|
|
struct_pack,
|
|
struct_unpack,
|
|
xpath_text,
|
|
)
|
|
|
|
|
|
class FlvReader(io.BytesIO):
|
|
"""
|
|
Reader for Flv files
|
|
The file format is documented in https://www.adobe.com/devnet/f4v.html
|
|
"""
|
|
|
|
# Utility functions for reading numbers and strings
|
|
def read_unsigned_long_long(self):
|
|
return struct_unpack('!Q', self.read(8))[0]
|
|
|
|
def read_unsigned_int(self):
|
|
return struct_unpack('!I', self.read(4))[0]
|
|
|
|
def read_unsigned_char(self):
|
|
return struct_unpack('!B', self.read(1))[0]
|
|
|
|
def read_string(self):
|
|
res = b''
|
|
while True:
|
|
char = self.read(1)
|
|
if char == b'\x00':
|
|
break
|
|
res += char
|
|
return res
|
|
|
|
def read_box_info(self):
|
|
"""
|
|
Read a box and return the info as a tuple: (box_size, box_type, box_data)
|
|
"""
|
|
real_size = size = self.read_unsigned_int()
|
|
box_type = self.read(4)
|
|
header_end = 8
|
|
if size == 1:
|
|
real_size = self.read_unsigned_long_long()
|
|
header_end = 16
|
|
return real_size, box_type, self.read(real_size - header_end)
|
|
|
|
def read_asrt(self):
|
|
# version
|
|
self.read_unsigned_char()
|
|
# flags
|
|
self.read(3)
|
|
quality_entry_count = self.read_unsigned_char()
|
|
# QualityEntryCount
|
|
for i in range(quality_entry_count):
|
|
self.read_string()
|
|
|
|
segment_run_count = self.read_unsigned_int()
|
|
segments = []
|
|
for i in range(segment_run_count):
|
|
first_segment = self.read_unsigned_int()
|
|
fragments_per_segment = self.read_unsigned_int()
|
|
segments.append((first_segment, fragments_per_segment))
|
|
|
|
return {
|
|
'segment_run': segments,
|
|
}
|
|
|
|
def read_afrt(self):
|
|
# version
|
|
self.read_unsigned_char()
|
|
# flags
|
|
self.read(3)
|
|
# time scale
|
|
self.read_unsigned_int()
|
|
|
|
quality_entry_count = self.read_unsigned_char()
|
|
# QualitySegmentUrlModifiers
|
|
for i in range(quality_entry_count):
|
|
self.read_string()
|
|
|
|
fragments_count = self.read_unsigned_int()
|
|
fragments = []
|
|
for i in range(fragments_count):
|
|
first = self.read_unsigned_int()
|
|
first_ts = self.read_unsigned_long_long()
|
|
duration = self.read_unsigned_int()
|
|
if duration == 0:
|
|
discontinuity_indicator = self.read_unsigned_char()
|
|
else:
|
|
discontinuity_indicator = None
|
|
fragments.append({
|
|
'first': first,
|
|
'ts': first_ts,
|
|
'duration': duration,
|
|
'discontinuity_indicator': discontinuity_indicator,
|
|
})
|
|
|
|
return {
|
|
'fragments': fragments,
|
|
}
|
|
|
|
def read_abst(self):
|
|
# version
|
|
self.read_unsigned_char()
|
|
# flags
|
|
self.read(3)
|
|
|
|
self.read_unsigned_int() # BootstrapinfoVersion
|
|
# Profile,Live,Update,Reserved
|
|
flags = self.read_unsigned_char()
|
|
live = flags & 0x20 != 0
|
|
# time scale
|
|
self.read_unsigned_int()
|
|
# CurrentMediaTime
|
|
self.read_unsigned_long_long()
|
|
# SmpteTimeCodeOffset
|
|
self.read_unsigned_long_long()
|
|
|
|
self.read_string() # MovieIdentifier
|
|
server_count = self.read_unsigned_char()
|
|
# ServerEntryTable
|
|
for i in range(server_count):
|
|
self.read_string()
|
|
quality_count = self.read_unsigned_char()
|
|
# QualityEntryTable
|
|
for i in range(quality_count):
|
|
self.read_string()
|
|
# DrmData
|
|
self.read_string()
|
|
# MetaData
|
|
self.read_string()
|
|
|
|
segments_count = self.read_unsigned_char()
|
|
segments = []
|
|
for i in range(segments_count):
|
|
box_size, box_type, box_data = self.read_box_info()
|
|
assert box_type == b'asrt'
|
|
segment = FlvReader(box_data).read_asrt()
|
|
segments.append(segment)
|
|
fragments_run_count = self.read_unsigned_char()
|
|
fragments = []
|
|
for i in range(fragments_run_count):
|
|
box_size, box_type, box_data = self.read_box_info()
|
|
assert box_type == b'afrt'
|
|
fragments.append(FlvReader(box_data).read_afrt())
|
|
|
|
return {
|
|
'segments': segments,
|
|
'fragments': fragments,
|
|
'live': live,
|
|
}
|
|
|
|
def read_bootstrap_info(self):
|
|
total_size, box_type, box_data = self.read_box_info()
|
|
assert box_type == b'abst'
|
|
return FlvReader(box_data).read_abst()
|
|
|
|
|
|
def read_bootstrap_info(bootstrap_bytes):
|
|
return FlvReader(bootstrap_bytes).read_bootstrap_info()
|
|
|
|
|
|
def build_fragments_list(boot_info):
|
|
""" Return a list of (segment, fragment) for each fragment in the video """
|
|
res = []
|
|
segment_run_table = boot_info['segments'][0]
|
|
fragment_run_entry_table = boot_info['fragments'][0]['fragments']
|
|
first_frag_number = fragment_run_entry_table[0]['first']
|
|
fragments_counter = itertools.count(first_frag_number)
|
|
for segment, fragments_count in segment_run_table['segment_run']:
|
|
for _ in range(fragments_count):
|
|
res.append((segment, next(fragments_counter)))
|
|
|
|
if boot_info['live']:
|
|
res = res[-2:]
|
|
|
|
return res
|
|
|
|
|
|
def write_unsigned_int(stream, val):
|
|
stream.write(struct_pack('!I', val))
|
|
|
|
|
|
def write_unsigned_int_24(stream, val):
|
|
stream.write(struct_pack('!I', val)[1:])
|
|
|
|
|
|
def write_flv_header(stream):
|
|
"""Writes the FLV header to stream"""
|
|
# FLV header
|
|
stream.write(b'FLV\x01')
|
|
stream.write(b'\x05')
|
|
stream.write(b'\x00\x00\x00\x09')
|
|
stream.write(b'\x00\x00\x00\x00')
|
|
|
|
|
|
def write_metadata_tag(stream, metadata):
|
|
"""Writes optional metadata tag to stream"""
|
|
SCRIPT_TAG = b'\x12'
|
|
FLV_TAG_HEADER_LEN = 11
|
|
|
|
if metadata:
|
|
stream.write(SCRIPT_TAG)
|
|
write_unsigned_int_24(stream, len(metadata))
|
|
stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
|
|
stream.write(metadata)
|
|
write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
|
|
|
|
|
|
def _add_ns(prop):
|
|
return '{http://ns.adobe.com/f4m/1.0}%s' % prop
|
|
|
|
|
|
class F4mFD(FragmentFD):
|
|
"""
|
|
A downloader for f4m manifests or AdobeHDS.
|
|
"""
|
|
|
|
FD_NAME = 'f4m'
|
|
|
|
def _get_unencrypted_media(self, doc):
|
|
media = doc.findall(_add_ns('media'))
|
|
if not media:
|
|
self.report_error('No media found')
|
|
for e in (doc.findall(_add_ns('drmAdditionalHeader')) +
|
|
doc.findall(_add_ns('drmAdditionalHeaderSet'))):
|
|
# If id attribute is missing it's valid for all media nodes
|
|
# without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
|
|
if 'id' not in e.attrib:
|
|
self.report_error('Missing ID in f4m DRM')
|
|
media = list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
|
|
'drmAdditionalHeaderSetId' not in e.attrib,
|
|
media))
|
|
if not media:
|
|
self.report_error('Unsupported DRM')
|
|
return media
|
|
|
|
def _get_bootstrap_from_url(self, bootstrap_url):
|
|
bootstrap = self.ydl.urlopen(bootstrap_url).read()
|
|
return read_bootstrap_info(bootstrap)
|
|
|
|
def _update_live_fragments(self, bootstrap_url, latest_fragment):
|
|
fragments_list = []
|
|
retries = 30
|
|
while (not fragments_list) and (retries > 0):
|
|
boot_info = self._get_bootstrap_from_url(bootstrap_url)
|
|
fragments_list = build_fragments_list(boot_info)
|
|
fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
|
|
if not fragments_list:
|
|
# Retry after a while
|
|
time.sleep(5.0)
|
|
retries -= 1
|
|
|
|
if not fragments_list:
|
|
self.report_error('Failed to update fragments')
|
|
|
|
return fragments_list
|
|
|
|
def _parse_bootstrap_node(self, node, base_url):
|
|
if node.text is None:
|
|
bootstrap_url = compat_urlparse.urljoin(
|
|
base_url, node.attrib['url'])
|
|
boot_info = self._get_bootstrap_from_url(bootstrap_url)
|
|
else:
|
|
bootstrap_url = None
|
|
bootstrap = base64.b64decode(node.text.encode('ascii'))
|
|
boot_info = read_bootstrap_info(bootstrap)
|
|
return (boot_info, bootstrap_url)
|
|
|
|
def real_download(self, filename, info_dict):
|
|
man_url = info_dict['url']
|
|
requested_bitrate = info_dict.get('tbr')
|
|
self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
|
|
urlh = self.ydl.urlopen(man_url)
|
|
man_url = urlh.geturl()
|
|
manifest = urlh.read()
|
|
|
|
doc = compat_etree_fromstring(manifest)
|
|
formats = [(int(f.attrib.get('bitrate', -1)), f)
|
|
for f in self._get_unencrypted_media(doc)]
|
|
if requested_bitrate is None:
|
|
# get the best format
|
|
formats = sorted(formats, key=lambda f: f[0])
|
|
rate, media = formats[-1]
|
|
else:
|
|
rate, media = list(filter(
|
|
lambda f: int(f[0]) == requested_bitrate, formats))[0]
|
|
|
|
base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
|
|
bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
|
|
boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, base_url)
|
|
live = boot_info['live']
|
|
metadata_node = media.find(_add_ns('metadata'))
|
|
if metadata_node is not None:
|
|
metadata = base64.b64decode(metadata_node.text.encode('ascii'))
|
|
else:
|
|
metadata = None
|
|
|
|
fragments_list = build_fragments_list(boot_info)
|
|
if self.params.get('test', False):
|
|
# We only download the first fragment
|
|
fragments_list = fragments_list[:1]
|
|
total_frags = len(fragments_list)
|
|
# For some akamai manifests we'll need to add a query to the fragment url
|
|
akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
|
|
|
|
ctx = {
|
|
'filename': filename,
|
|
'total_frags': total_frags,
|
|
}
|
|
|
|
self._prepare_frag_download(ctx)
|
|
|
|
dest_stream = ctx['dest_stream']
|
|
|
|
write_flv_header(dest_stream)
|
|
if not live:
|
|
write_metadata_tag(dest_stream, metadata)
|
|
|
|
base_url_parsed = compat_urllib_parse_urlparse(base_url)
|
|
|
|
self._start_frag_download(ctx)
|
|
|
|
frags_filenames = []
|
|
while fragments_list:
|
|
seg_i, frag_i = fragments_list.pop(0)
|
|
name = 'Seg%d-Frag%d' % (seg_i, frag_i)
|
|
query = []
|
|
if base_url_parsed.query:
|
|
query.append(base_url_parsed.query)
|
|
if akamai_pv:
|
|
query.append(akamai_pv.strip(';'))
|
|
if info_dict.get('extra_param_to_segment_url'):
|
|
query.append(info_dict['extra_param_to_segment_url'])
|
|
url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
|
|
frag_filename = '%s-%s' % (ctx['tmpfilename'], name)
|
|
try:
|
|
success = ctx['dl'].download(frag_filename, {'url': url_parsed.geturl()})
|
|
if not success:
|
|
return False
|
|
(down, frag_sanitized) = sanitize_open(frag_filename, 'rb')
|
|
down_data = down.read()
|
|
down.close()
|
|
reader = FlvReader(down_data)
|
|
while True:
|
|
_, box_type, box_data = reader.read_box_info()
|
|
if box_type == b'mdat':
|
|
dest_stream.write(box_data)
|
|
break
|
|
if live:
|
|
os.remove(encodeFilename(frag_sanitized))
|
|
else:
|
|
frags_filenames.append(frag_sanitized)
|
|
except (compat_urllib_error.HTTPError, ) as err:
|
|
if live and (err.code == 404 or err.code == 410):
|
|
# We didn't keep up with the live window. Continue
|
|
# with the next available fragment.
|
|
msg = 'Fragment %d unavailable' % frag_i
|
|
self.report_warning(msg)
|
|
fragments_list = []
|
|
else:
|
|
raise
|
|
|
|
if not fragments_list and live and bootstrap_url:
|
|
fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
|
|
total_frags += len(fragments_list)
|
|
if fragments_list and (fragments_list[0][1] > frag_i + 1):
|
|
msg = 'Missed %d fragments' % (fragments_list[0][1] - (frag_i + 1))
|
|
self.report_warning(msg)
|
|
|
|
self._finish_frag_download(ctx)
|
|
|
|
for frag_file in frags_filenames:
|
|
os.remove(encodeFilename(frag_file))
|
|
|
|
return True
|