]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/downloader/hls.py
Prepare to release.
[youtubedl] / youtube_dl / downloader / hls.py
index 2a775bf0023f7ddc09507d66ab660a8dd97d19b2..fd304527e854e768038880bf785d84f211f4471b 100644 (file)
 from __future__ import unicode_literals
 
-import os
 import re
-import subprocess
-import sys
+import binascii
+try:
+    from Crypto.Cipher import AES
+    can_decrypt_frag = True
+except ImportError:
+    can_decrypt_frag = False
 
-from .common import FileDownloader
 from .fragment import FragmentFD
+from .external import FFmpegFD
 
-from ..compat import compat_urlparse
-from ..postprocessor.ffmpeg import FFmpegPostProcessor
+from ..compat import (
+    compat_urllib_error,
+    compat_urlparse,
+    compat_struct_pack,
+)
 from ..utils import (
-    encodeArgument,
-    encodeFilename,
-    sanitize_open,
-    handle_youtubedl_headers,
+    parse_m3u8_attributes,
+    update_url_query,
 )
 
 
-class HlsFD(FileDownloader):
-    def real_download(self, filename, info_dict):
-        url = info_dict['url']
-        self.report_destination(filename)
-        tmpfilename = self.temp_name(filename)
-
-        ffpp = FFmpegPostProcessor(downloader=self)
-        if not ffpp.available:
-            self.report_error('m3u8 download detected but ffmpeg or avconv could not be found. Please install one.')
-            return False
-        ffpp.check_version()
-
-        args = [ffpp.executable, '-y']
-
-        if info_dict['http_headers'] and re.match(r'^https?://', url):
-            # Trailing \r\n after each HTTP header is important to prevent warning from ffmpeg/avconv:
-            # [http @ 00000000003d2fa0] No trailing CRLF found in HTTP header.
-            headers = handle_youtubedl_headers(info_dict['http_headers'])
-            args += [
-                '-headers',
-                ''.join('%s: %s\r\n' % (key, val) for key, val in headers.items())]
-
-        args += ['-i', url, '-c', 'copy']
-        if self.params.get('hls_use_mpegts', False):
-            args += ['-f', 'mpegts']
-        else:
-            args += ['-f', 'mp4', '-bsf:a', 'aac_adtstoasc']
-
-        args = [encodeArgument(opt) for opt in args]
-        args.append(encodeFilename(ffpp._ffmpeg_filename_argument(tmpfilename), True))
-
-        self._debug_cmd(args)
-
-        proc = subprocess.Popen(args, stdin=subprocess.PIPE)
-        try:
-            retval = proc.wait()
-        except KeyboardInterrupt:
-            # subprocces.run would send the SIGKILL signal to ffmpeg and the
-            # mp4 file couldn't be played, but if we ask ffmpeg to quit it
-            # produces a file that is playable (this is mostly useful for live
-            # streams). Note that Windows is not affected and produces playable
-            # files (see https://github.com/rg3/youtube-dl/issues/8300).
-            if sys.platform != 'win32':
-                proc.communicate(b'q')
-            raise
-        if retval == 0:
-            fsize = os.path.getsize(encodeFilename(tmpfilename))
-            self.to_screen('\r[%s] %s bytes' % (args[0], fsize))
-            self.try_rename(tmpfilename, filename)
-            self._hook_progress({
-                'downloaded_bytes': fsize,
-                'total_bytes': fsize,
-                'filename': filename,
-                'status': 'finished',
-            })
-            return True
-        else:
-            self.to_stderr('\n')
-            self.report_error('%s exited with code %d' % (ffpp.basename, retval))
-            return False
-
-
-class NativeHlsFD(FragmentFD):
-    """ A more limited implementation that does not require ffmpeg """
+class HlsFD(FragmentFD):
+    """ A limited implementation that does not require ffmpeg """
 
     FD_NAME = 'hlsnative'
 
+    @staticmethod
+    def can_download(manifest, info_dict):
+        UNSUPPORTED_FEATURES = (
+            r'#EXT-X-KEY:METHOD=(?!NONE|AES-128)',  # encrypted streams [1]
+            # r'#EXT-X-BYTERANGE',  # playlists composed of byte ranges of media files [2]
+
+            # Live streams heuristic does not always work (e.g. geo restricted to Germany
+            # http://hls-geo.daserste.de/i/videoportal/Film/c_620000/622873/format,716451,716457,716450,716458,716459,.mp4.csmil/index_4_av.m3u8?null=0)
+            # r'#EXT-X-MEDIA-SEQUENCE:(?!0$)',  # live streams [3]
+
+            # This heuristic also is not correct since segments may not be appended as well.
+            # Twitch vods of finished streams have EXT-X-PLAYLIST-TYPE:EVENT despite
+            # no segments will definitely be appended to the end of the playlist.
+            # r'#EXT-X-PLAYLIST-TYPE:EVENT',  # media segments may be appended to the end of
+            #                                 # event media playlists [4]
+
+            # 1. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.4
+            # 2. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.2
+            # 3. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.2
+            # 4. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.5
+        )
+        check_results = [not re.search(feature, manifest) for feature in UNSUPPORTED_FEATURES]
+        is_aes128_enc = '#EXT-X-KEY:METHOD=AES-128' in manifest
+        check_results.append(can_decrypt_frag or not is_aes128_enc)
+        check_results.append(not (is_aes128_enc and r'#EXT-X-BYTERANGE' in manifest))
+        check_results.append(not info_dict.get('is_live'))
+        return all(check_results)
+
     def real_download(self, filename, info_dict):
         man_url = info_dict['url']
         self.to_screen('[%s] Downloading m3u8 manifest' % self.FD_NAME)
-        manifest = self.ydl.urlopen(man_url).read()
 
-        s = manifest.decode('utf-8', 'ignore')
-        fragment_urls = []
+        urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
+        man_url = urlh.geturl()
+        s = urlh.read().decode('utf-8', 'ignore')
+
+        if not self.can_download(s, info_dict):
+            if info_dict.get('extra_param_to_segment_url'):
+                self.report_error('pycrypto not found. Please install it.')
+                return False
+            self.report_warning(
+                'hlsnative has detected features it does not support, '
+                'extraction will be delegated to ffmpeg')
+            fd = FFmpegFD(self.ydl, self.params)
+            for ph in self._progress_hooks:
+                fd.add_progress_hook(ph)
+            return fd.real_download(filename, info_dict)
+
+        def is_ad_fragment(s):
+            return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s or
+                    s.startswith('#UPLYNK-SEGMENT') and s.endswith(',ad'))
+
+        media_frags = 0
+        ad_frags = 0
+        ad_frag_next = False
         for line in s.splitlines():
             line = line.strip()
-            if line and not line.startswith('#'):
-                segment_url = (
-                    line
-                    if re.match(r'^https?://', line)
-                    else compat_urlparse.urljoin(man_url, line))
-                fragment_urls.append(segment_url)
-                # We only download the first fragment during the test
-                if self.params.get('test', False):
-                    break
+            if not line:
+                continue
+            if line.startswith('#'):
+                if is_ad_fragment(line):
+                    ad_frags += 1
+                    ad_frag_next = True
+                continue
+            if ad_frag_next:
+                ad_frag_next = False
+                continue
+            media_frags += 1
 
         ctx = {
             'filename': filename,
-            'total_frags': len(fragment_urls),
+            'total_frags': media_frags,
+            'ad_frags': ad_frags,
         }
 
         self._prepare_and_start_frag_download(ctx)
 
-        frags_filenames = []
-        for i, frag_url in enumerate(fragment_urls):
-            frag_filename = '%s-Frag%d' % (ctx['tmpfilename'], i)
-            success = ctx['dl'].download(frag_filename, {'url': frag_url})
-            if not success:
-                return False
-            down, frag_sanitized = sanitize_open(frag_filename, 'rb')
-            ctx['dest_stream'].write(down.read())
-            down.close()
-            frags_filenames.append(frag_sanitized)
+        fragment_retries = self.params.get('fragment_retries', 0)
+        skip_unavailable_fragments = self.params.get('skip_unavailable_fragments', True)
+        test = self.params.get('test', False)
+
+        extra_query = None
+        extra_param_to_segment_url = info_dict.get('extra_param_to_segment_url')
+        if extra_param_to_segment_url:
+            extra_query = compat_urlparse.parse_qs(extra_param_to_segment_url)
+        i = 0
+        media_sequence = 0
+        decrypt_info = {'METHOD': 'NONE'}
+        byte_range = {}
+        frag_index = 0
+        ad_frag_next = False
+        for line in s.splitlines():
+            line = line.strip()
+            if line:
+                if not line.startswith('#'):
+                    if ad_frag_next:
+                        ad_frag_next = False
+                        continue
+                    frag_index += 1
+                    if frag_index <= ctx['fragment_index']:
+                        continue
+                    frag_url = (
+                        line
+                        if re.match(r'^https?://', line)
+                        else compat_urlparse.urljoin(man_url, line))
+                    if extra_query:
+                        frag_url = update_url_query(frag_url, extra_query)
+                    count = 0
+                    headers = info_dict.get('http_headers', {})
+                    if byte_range:
+                        headers['Range'] = 'bytes=%d-%d' % (byte_range['start'], byte_range['end'])
+                    while count <= fragment_retries:
+                        try:
+                            success, frag_content = self._download_fragment(
+                                ctx, frag_url, info_dict, headers)
+                            if not success:
+                                return False
+                            break
+                        except compat_urllib_error.HTTPError as err:
+                            # Unavailable (possibly temporary) fragments may be served.
+                            # First we try to retry then either skip or abort.
+                            # See https://github.com/rg3/youtube-dl/issues/10165,
+                            # https://github.com/rg3/youtube-dl/issues/10448).
+                            count += 1
+                            if count <= fragment_retries:
+                                self.report_retry_fragment(err, frag_index, count, fragment_retries)
+                    if count > fragment_retries:
+                        if skip_unavailable_fragments:
+                            i += 1
+                            media_sequence += 1
+                            self.report_skip_fragment(frag_index)
+                            continue
+                        self.report_error(
+                            'giving up after %s fragment retries' % fragment_retries)
+                        return False
+                    if decrypt_info['METHOD'] == 'AES-128':
+                        iv = decrypt_info.get('IV') or compat_struct_pack('>8xq', media_sequence)
+                        decrypt_info['KEY'] = decrypt_info.get('KEY') or self.ydl.urlopen(
+                            self._prepare_url(info_dict, decrypt_info['URI'])).read()
+                        frag_content = AES.new(
+                            decrypt_info['KEY'], AES.MODE_CBC, iv).decrypt(frag_content)
+                    self._append_fragment(ctx, frag_content)
+                    # We only download the first fragment during the test
+                    if test:
+                        break
+                    i += 1
+                    media_sequence += 1
+                elif line.startswith('#EXT-X-KEY'):
+                    decrypt_url = decrypt_info.get('URI')
+                    decrypt_info = parse_m3u8_attributes(line[11:])
+                    if decrypt_info['METHOD'] == 'AES-128':
+                        if 'IV' in decrypt_info:
+                            decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
+                        if not re.match(r'^https?://', decrypt_info['URI']):
+                            decrypt_info['URI'] = compat_urlparse.urljoin(
+                                man_url, decrypt_info['URI'])
+                        if extra_query:
+                            decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
+                        if decrypt_url != decrypt_info['URI']:
+                            decrypt_info['KEY'] = None
+                elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
+                    media_sequence = int(line[22:])
+                elif line.startswith('#EXT-X-BYTERANGE'):
+                    splitted_byte_range = line[17:].split('@')
+                    sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
+                    byte_range = {
+                        'start': sub_range_start,
+                        'end': sub_range_start + int(splitted_byte_range[0]),
+                    }
+                elif is_ad_fragment(line):
+                    ad_frag_next = True
 
         self._finish_frag_download(ctx)
 
-        for frag_file in frags_filenames:
-            os.remove(encodeFilename(frag_file))
-
         return True