Multi period for HLS #83

mariocynicys · 2021-07-20T07:46:37Z

This PR adds the support for multi-period concatenation for HLS, using #EXT-X-DISCONTINUITY between periods.

The new file m3u8_concater.py has the classes used to parse the m3u8 files and reduce multiple files of them into one concatenated file.

HLSConcater is just an API that is exposed to the PeriodConcatNode for abstraction.

MasterPlaylist is a class representing a master hls playlist, the parsing part happens in the __init__ method, creates a MediaPlaylist object for each media(video,text,audio) playlist found in this master playlist, the stream attributes for each playlist is passed to the MediaPlaylist object and stored in it.

MediaPlaylist is a class representing an hls media playlist, the playlist gets parsed in the __init__ method, the biggest part of the playlist is saved mostly unchanged in the self.content variable, but the media segment paths get updated to point to the new relative path of the media segment.

The playlist concatenation methods are the static methods in the MediaPlaylist class (concat_sub, concat_aud, concat_vid, concat_aud_only(to handle the audio-only stream cases)).

No period concatenation is performed if an inconsistent media in periods is detected(e.g. one period has video and one does not).

Things that need addressing:
1- When Shaka Player gets an HLS master playlist containing unsupported and supported audio codecs, it breaks, this happens with:

audio-only playlists with and without concatenation,
video+audio playlists that have been concatenated, but does't happen for the un-concatenated playlists with the same set of codecs, this may have to do with the discontinuities inserted between periods.

2- When Shaka Packager receives an audio-only period to package, it normally creates a stream variant(stream-inf) for each media(audio) playlist generated, but when this audio-only input has one subtitles playlist along with the audio, Shaka-Packager doesn't create the stream-inf playlists, for the current implementation, this will create an error in the HLS concater, we can escape this period as a solution if it has no stream-infs, periods with no stream-inf we can't know its bitrates and codecs and isn't playable in Shaka Player.

adding period concatenation for dash, no tests are done yet.

pully

joeyparrish

I still need to read m3u8_concater.py, but I haven't been able to make time yet. Wanted to go ahead with feedback on the other files.

joeyparrish · 2021-07-20T20:32:24Z

streamer/periodconcat_node.py

+              "\tperiods with other periods that have video that is for the concatenation\n"
+              "\tto be performed successfully.\n")


English nit: This sentence is too long. Please end this sentence after "other periods that have video." If you want to say that this is necessary for concatenation, make that a second sentence.

joeyparrish · 2021-07-20T20:37:28Z

streamer/periodconcat_node.py

+        # Overwrite the start and the stop methods.
+        setattr(self, 'start', lambda: None)
+        setattr(self, 'stop', lambda _=None: None)


I don't like overwriting these methods. How about you set a flag like self._fatal_error = True and then throw a RuntimeError at the top of _thread_single_pass if it's set?

ah, sure,
Can we not error when we see the flag?, i think of supporting multiple inputs even if it's not going to be concatenated.
Can we just set the status to finished in the _thread_single_pass.

We don't intend to support it in the player, so we don't intend to support that scenario here, either. I think it would be best to guide people toward more reasonable, widely-supported structures in media, until there's a compelling use-case for us to broaden support. We haven't heard one yet.

What i meant is that the user can use one streamer instance to process multiple inputs, instead of instantiating multiple instances of streamer, for LIVE for example, since there is no concatenation support, i just didn't append the concater node here, so the user can push multiple live streams from the same streamer instance.

oh i think i miss understood you, ok let's raise an error till someone requests not to.

joeyparrish · 2021-07-20T20:39:50Z

streamer/periodconcat_node.py

+    """Concatenates multiple HLS playlists using #EXT-X-DISCONTINUITY."""
+
+    # Initialize the HLS concater with a sample Master HLS playlist and
+    # the output direcotry of the concatenated playlists.


typo: directory

joeyparrish · 2021-07-20T20:41:42Z

streamer/periodconcat_node.py

+
+    # Initialize the HLS concater with a sample Master HLS playlist and
+    # the output direcotry of the concatenated playlists.
+    hls_concater = HLSConcater(os.path.join(self._packager_nodes[0].output_dir,


It took me a minute to read and understand this join. Can you make it easier to read by assigning this path to a local variable with an explanatory name? Perhaps something like first_playlist_path?

joeyparrish · 2021-07-20T20:42:02Z

streamer/periodconcat_node.py

+                               self._output_dir)
+
+    for packager_node in self._packager_nodes:
+      hls_concater.add(os.path.join(packager_node.output_dir,


Same here. Please assign the path to a local var for clarity.

mariocynicys · 2021-08-16T17:57:06Z

streamer/transcoder_node.py

@@ -142,13 +142,13 @@ def start(self) -> None:

        args += [output_stream.ipc_pipe.write_end()]

-    env = {}
+    stderr = None


@joeyparrish The multiple nodes form ffmpeg overwrite TranscoderNode.log every time they try to write to it.
Here i removed the FFreport and redirected the stderr instead to a file that is open for appending.
This has a downside though, there will be no ffmpeg logs in the terminal for the sake of saving them to a file, only when debug_logs is set.

See my comment on the PackagerNode logs. Let's go back to using FFREPORT, so we get separate log verbosity for stderr (just progress) and log file (more verbose). But let's also change the file name to indicate the period.

joeyparrish · 2021-08-16T19:01:48Z

streamer/packager_node.py

@@ -112,7 +112,7 @@ def start(self) -> None:
      # Log by writing all Packager output to a file.  Unlike the logging
      # system in ffmpeg, this will stop any Packager output from getting to
      # the screen.
-      stdout = open('PackagerNode.log', 'w')
+      stdout = open('PackagerNode.log', 'a')


The log files could have a number added, based on the period, such as PackagerNode-1.log. That would be a nice solution to the problem of overwriting them, I think.

joeyparrish · 2021-08-16T19:04:04Z

streamer/transcoder_node.py

@@ -142,13 +142,13 @@ def start(self) -> None:

        args += [output_stream.ipc_pipe.write_end()]

-    env = {}
+    stderr = None


See my comment on the PackagerNode logs. Let's go back to using FFREPORT, so we get separate log verbosity for stderr (just progress) and log file (more verbose). But let's also change the file name to indicate the period.

joeyparrish

Just 700 lines left to review... :-)

joeyparrish · 2021-08-16T19:11:42Z

streamer/m3u8_concater.py

+    # Save common master playlist header, this will call
+    # MediaPlaylist.extract_header() to save the common
+    # media playlist header as well.
+    MasterPlaylist.extract_headers(sample_master_playlist_path)


It bothers me that these are stored statically on the class. It means that an application using the Python API can't create multiple independent controllers doing concatenation if the headers would need to be different.

What if, instead, the headers (both master & media headers) are stored on the HLSConcater instance, and passed to .write()?

joeyparrish · 2021-08-16T19:17:14Z

streamer/m3u8_concater.py

+
+  def write(self, dir_name: str) -> None:
+    file_path = os.path.join(dir_name, _unquote(self.stream_info['URI']))
+    with open(file_path, 'w') as media_playlist:


There are other places where "media_playlist" is a MediaPlaylist instance. Here, it's a file instance. That makes it confusing when searching the code for media_playlist.write(), because they mean two different things in these two contexts. Could you please rename this file instance to something else?

joeyparrish · 2021-08-16T19:18:32Z

streamer/m3u8_concater.py

+                       self._output_location,
+                       packager_node))
+
+  def concat(self) -> None:


I'm not sure why concat() and write() are separate steps in this class, or why the resulting master playlist is saved to a member variable. These should be combined, I think, and the master playlist should be a local variable.

Yeah, we can do this.

joeyparrish · 2021-08-16T19:36:30Z

streamer/m3u8_concater.py

+          var_playlists.append(media_playlist)
+        else:
+          # TODO: We need a case for CLOSED-CAPTIONS(CC).
+          raise RuntimeError("TYPE={} is not regonized.".format(stream_type))  


typo: recognized

joeyparrish · 2021-08-16T19:38:11Z

streamer/m3u8_concater.py

+    master_hls.playlists.extend(
+        MediaPlaylist.concat_sub(all_txt_playlists, durations))
+
+    if all(not MediaPlaylist.inf_is_vid(inf_pl) for inf_pl in all_var_playlists):


I'm having trouble reading this. What's "inf" short for? Variable names make a big difference in readability and maintainability.

Since "inf_pl" is an element from the list "all_var_playlists", could this fairly be called "var_playlist" instead?

Oh, sorry about that, i just renamed all the infs to vars today but overlooked this one.
I was calling a playlist that is a stream variant(#EXT-X-STREAM-INF) an inf playlist.

joeyparrish · 2021-08-16T19:39:40Z

streamer/m3u8_concater.py

+            durations))
+    else:
+      master_hls.playlists.extend(
+          MediaPlaylist.concat_aud(all_aud_playlists))


Why don't you need the duration for audio?

For text streams we use the durations to substitute for missing text streamer(in the worst case) in some periods, but ideally we don't need the durations anywhere else other than with stream variants, because the durations are used to calculated the average bitrate of the concatenated playlist.

joeyparrish · 2021-08-16T19:50:54Z

streamer/m3u8_concater.py

+        outstream
+        .get_single_seg_file()
+        .write_end(): outstream


Will these fit on one line? I think I would find it easier to read that way.

joeyparrish · 2021-08-16T19:51:24Z

streamer/m3u8_concater.py

+        outstream
+        .get_media_seg_file()
+        .write_end()
+        .replace('$Number$', '1'): outstream


Same here. Does this fit on one line?

But this does not, can we make an exception for this :), ends at col.87

mariocynicys · 2021-08-16T20:27:49Z

streamer/m3u8_concater.py

+        else:
+          # If the no playlist were found for this period,
+          # Create a time gap filled with dummy data for the period's duration.
+          dummy = ',\ndata:text/plain,NO {} SUBTITLES\n'.format(lang.upper())


I haven't actually tested whether this renders on the screen normally or not, i have a problem with subtitles in HLS, their streams are being dropped in chrome and firefox too. Will try to fix it and report back.

I think this should be an empty WEBVTT file:

data:text/vtt;charset=utf-8,WEBVTT%0A%0A

The WEBVTT header, followed by two newlines, with the correct MIME type and character set.

If you want, you could add an actual comment in the WEBVTT:

data:text/vtt;charset=utf-8,WEBVTT%0A%0ANOTE%20No%20{}%20subtitles

Or put a comment into the HLS playlist above the URI instead.

It's weird, though, and maybe hacky, to put the command and newline before the URI as part of this string. That would more logically belong to the EXTINF string.

I think it's better to leave it a blank string as you stated. Thanks!

It's weird, though, and maybe hacky, to put the comma and newline before the URI as part of this string. That would more logically belong to the EXTINF string.

Was fighting with line lengths T_T

mariocynicys · 2021-08-16T20:37:06Z

streamer/m3u8_concater.py

+            aud_playlist_options = [codec_lang_division[codec][lang][0] for
+                                    lang in langs
+                                    if len(codec_lang_division[codec][lang])]
+            sub_lang = MediaPlaylist._fit_missing_lang(aud_playlist_options,


This substitution is not as big as the one you do in the player(based on channel count, bitrate, sample rate, etc..).
It is only based on the language only for simplicity, so if some language has no playlists at all with any channel layout, we query _fit_missing_lang() with the all the media playlists we have for that period and a language to fit upon, the method will return a best fit based on the language only.

Sounds good. You should be able to count on similar channels for the same reason as similar resolutions in video: everything went through the same pipeline config.

joeyparrish

Still reviewing concatenation of audio and video, but I've read and commented on the rest now.

joeyparrish · 2021-08-16T20:23:21Z

streamer/m3u8_concater.py

+            self.content += ',BYTERANGE=' + attribs['BYTERANGE']
+          self.content += '\n'
+        elif line.startswith(MediaPlaylist.HEADER_TAGS + ('#EXT-X-ENDLIST',)):
+          # Escape header and end-list tags.


By "escape", do you mean "skip"?

joeyparrish · 2021-08-16T20:23:27Z

streamer/m3u8_concater.py

+          # Escape header and end-list tags.
+          pass
+        elif not line.startswith('#EXT'):
+          # Escape comments.


joeyparrish · 2021-08-16T20:24:48Z

streamer/m3u8_concater.py

+    """Get the audio and video codecs and other relavent stream features
+    from the matching OutputStream in the `streams_map`, this will be used
+    in the codec matching process in the concat_xxx() methods, but the codecs
+    that will be written in the final concatenateded master playlist will be


typo: concatenated

joeyparrish · 2021-08-16T20:27:28Z

streamer/m3u8_concater.py

+    output_stream: Optional[OutputStream] = None
+    lines = self.content.split('\n')
+    for i in range(len(lines)):
+      # Don't use #EXT-X-MAP to get the codec, because HLS specs says


I don't think EXT-X-MAP would ever contain codec information anyway. What did you mean by this comment?

def get_init_seg_file(self) -> Pipe: INIT_SEGMENT = { MediaType.AUDIO: 'audio_{language}_{channels}c_{bitrate}_{codec}_init.{format}', MediaType.VIDEO: 'video_{resolution_name}_{bitrate}_{codec}_init.{format}', MediaType.TEXT: 'text_{language}_init.{format}', } path_templ = INIT_SEGMENT[self.type].format(**self.features) return Pipe.create_file_pipe(path_templ, mode='w') def get_media_seg_file(self) -> Pipe: MEDIA_SEGMENT = { MediaType.AUDIO: 'audio_{language}_{channels}c_{bitrate}_{codec}_$Number$.{format}', MediaType.VIDEO: 'video_{resolution_name}_{bitrate}_{codec}_$Number$.{format}', MediaType.TEXT: 'text_{language}_$Number$.{format}', } path_templ = MEDIA_SEGMENT[self.type].format(**self.features) return Pipe.create_file_pipe(path_templ, mode='w') def get_single_seg_file(self) -> Pipe: SINGLE_SEGMENT = { MediaType.AUDIO: 'audio_{language}_{channels}c_{bitrate}_{codec}.{format}', MediaType.VIDEO: 'video_{resolution_name}_{bitrate}_{codec}.{format}', MediaType.TEXT: 'text_{language}.{format}', } path_templ = SINGLE_SEGMENT[self.type].format(**self.features) return Pipe.create_file_pipe(path_templ, mode='w')

#EXTM3U #EXT-X-VERSION:6 ## Generated with https://github.com/google/shaka-packager version c1f64e5-release #EXT-X-TARGETDURATION:4 #EXT-X-PLAYLIST-TYPE:VOD #EXT-X-MAP:URI="video_480p_1M_h264_init.mp4" #EXTINF:3.000, video_480p_1M_h264_1.mp4 #EXT-X-ENDLIST

We use the streams_map to get the matching output stream object, but when creating the streams map we only added the mapping from the methods get_single_seg_file() and get_media_seg_file(), in the segment_per_file case, we could have added get_init_seg_file() instead and checked the URI in the #EXT-X-MAP tag, this would yield to the same result, but i was skeptic about that in some cases packager might not output an #EXT-X-MAP tag since it's optional according to the HLS specs. Kept the comment as a reminder.

Ah, now I see. How about this, which would make more sense to me in a comment:

# Don't use the URIs from any tag to try to extract codec information. # We should not rely on the exact structure of file names for this. # Use stream_maps instead.

joeyparrish · 2021-08-16T20:31:43Z

streamer/m3u8_concater.py

+
+    output_stream: Optional[OutputStream] = None
+    lines = self.content.split('\n')
+    for i in range(len(lines)):


It looks like you're seeking to the first URI line in the file after EXTINF. Is that right?

This loop is so long, I think I could use a comment just before it, explaining what it's for.

Yes.
Ok, will do.

joeyparrish · 2021-08-16T20:46:48Z

streamer/m3u8_concater.py

+    """Returns a substitution language for a missing language by considering
+    the languages of the given variants.
+
+    Returns the argument `language` back only if no variants were given


What is this argument? A default? A goal/target? It's difficult to tell from the arguments, the name, or the docstring.

This is the argument passed to the method, it's like, if you don't provide me with variants i can choose form to replace the missing language, i will just return this missing language back to you.

This will only happen for when no text input in any language what so ever is present for some period, this means we will fill this period with the dummy subtitles for the period's duration to keep the subtitles of the next period(if there is) in sync.
This method will return the same missing language back and that will indicate that we failed to find a substitution for this language.

It's some fancy way to handle this though, i can just change it in the concat_sub method itself.

joeyparrish · 2021-08-16T20:50:10Z

streamer/m3u8_concater.py

+      # the best fit is not the same as the base of the original language
+      # OR the candidate is a regional variant).
+      if language_base == candidate_base:
+        if language_base != best_fit_base or candidate_is_reg:


So any regional variant is better than ... anything else? I'm not sure I understand this clause.

(regional variant > base-only variant) das ist gut.

(regional variant > another regional variant with the same base) they should be equal, but the stability isn't a concern, the user should have provided stitchable subtitles in the first place >:O.

I wouldn't say it's not stable though, it is like stable from the end, so always the regional variant at the end wins the slot.

joeyparrish · 2021-08-16T20:53:59Z

streamer/m3u8_concater.py

+    stream_name = 'stream_' + str(MediaPlaylist.current_stream_index)
+    MediaPlaylist.current_stream_index += 1
+
+    return _quote(stream_name), _quote(stream_name + '.m3u8')


Again I'm surprised by the tuple of quoted strings. How about a dictionary indicating what they are for? If you return a dict with attributes to add to a tag, then the quoted string values would make more sense. And you could still easily use .update() with the return value.

That would be absolutely great.

joeyparrish · 2021-08-16T20:57:14Z

streamer/m3u8_concater.py

+        else:
+          # If the no playlist were found for this period,
+          # Create a time gap filled with dummy data for the period's duration.
+          dummy = ',\ndata:text/plain,NO {} SUBTITLES\n'.format(lang.upper())


I think this should be an empty WEBVTT file:

data:text/vtt;charset=utf-8,WEBVTT%0A%0A

The WEBVTT header, followed by two newlines, with the correct MIME type and character set.

If you want, you could add an actual comment in the WEBVTT:

data:text/vtt;charset=utf-8,WEBVTT%0A%0ANOTE%20No%20{}%20subtitles

Or put a comment into the HLS playlist above the URI instead.

It's weird, though, and maybe hacky, to put the command and newline before the URI as part of this string. That would more logically belong to the EXTINF string.

joeyparrish · 2021-08-16T21:02:31Z

streamer/m3u8_concater.py

+          dummy = ',\ndata:text/plain,NO {} SUBTITLES\n'.format(lang.upper())
+          # Break the period's duration into target duration count and remains,
+          # because the target duration might be a prime number, this might
+          # cause an accumulated rounding error.


This is fine, BTW.

Do you mean the error?
Python generates a very long float number when doing a normal float division, i think accumulating these fractions to one #EXTINF tag might reduce the error to some extent.

I mean some small accumulated rounding error here is fine. I think you're doing the right thing, and I think it's not worth the effort to do more until we have evidence that it's necessary.

joeyparrish

That's everything! Thanks for being patient with me. I know it has taken me all day (my day, well into your night) to get this review completed.

joeyparrish · 2021-08-16T23:23:26Z

streamer/m3u8_concater.py

+      # Initialize a mapping between video codecs and a list of resolutions available.
+      codec_division: Dict[VideoCodec, List[MediaPlaylist]] = {}
+      for codec in codecs:
+        codec_division[codec] = []


No need to change this in this PR, but just as a tip, you might consider if collections.defaultdict might be a useful tool for this pattern.

joeyparrish · 2021-08-16T23:26:48Z

streamer/m3u8_concater.py

+        codec_division[codec].sort(key=lambda pl: pl.resolution)
+        for i, resolution in enumerate(sorted(resolutions)):
+          division[codec][resolution].append(
+              # Append the ith resolution if found, else, append


So you're not comparing the actual resolutions? I'm thinking about this abstractly here at first, with something like:

[360p, 720p, 1080p] + [720p, 1080p, 4k]

Going by index, you'd end up with something misaligned like:

[360p + 720p, 720p + 1080p, 1080p + 4k]

Instead of something like:

[360p + 720p, 720p + 720p, 1080p + 1080p, 1080p + 4k]

At least in Player, we have to contend with these scenarios. But I guess here, we don't, since the bottom rungs of the output from Streamer will always align since they came from the same pipeline config, and we only cut off the top rungs based on the input res. Is that accurate?

Could you explain this in the comment a little? (If I understood correctly, add to the comment that the lower resolutions will always align because the outputs shared a pipeline config, so this indexing is always safe.)

Yes, i am using this perk of the Streamer to make the matching easier, actually i had to do it similar to what you do in the Player with the audio channel matching prior to the 'channel_layout as an input feature' PR. I was starting with a dictionary full on Nones for every slot and start filling it appropriately with the channel count it should get for every period.

joeyparrish · 2021-08-16T23:29:12Z

streamer/m3u8_concater.py

+          media_playlist.write(dir_name)
+          # We don't write the URI in the attributes of a stream
+          # variant playlist.  Pop out the URI.
+          uri = _unquote(media_playlist.stream_info.pop('URI'))


This is really cool. I had no idea you could "pop" from a dictionary in Python.

joeyparrish · 2021-08-16T23:31:42Z

streamer/m3u8_concater.py

+      for codec in codecs:
+        for lang in langs:
+          # If this language for this codec in this period has no media playlists
+          # with a for any channel layout, this means that the language itself


typo: maybe remove "with a"? Guessing at your original intent.

joeyparrish · 2021-08-16T23:32:18Z

streamer/m3u8_concater.py

+            aud_playlist_options = [codec_lang_division[codec][lang][0] for
+                                    lang in langs
+                                    if len(codec_lang_division[codec][lang])]
+            sub_lang = MediaPlaylist._fit_missing_lang(aud_playlist_options,


Sounds good. You should be able to count on similar channels for the same reason as similar resolutions in video: everything went through the same pipeline config.

joeyparrish · 2021-08-16T23:34:39Z

streamer/m3u8_concater.py

+                      durations: List[float]) -> List['MediaPlaylist']:
+    """Concatenates audio only periods with other audio only periods."""
+
+    # Pair audio streams with their equivalent stream variants


I honestly think it feels wrong that Packager spits out a "variant" that pairs with an audio group of itself, but I haven't spent the time digging through the spec to decide if it's required by the spec. And Packager isn't the only one, and Player already has to handle it. In any case, you've done well with this detail.

joeyparrish · 2021-08-23T18:46:22Z

streamer/m3u8_concater.py

+    output_stream: Optional[OutputStream] = None
+    lines = self.content.split('\n')
+    for i in range(len(lines)):
+      # Don't use #EXT-X-MAP to get the codec, because HLS specs says


Ah, now I see. How about this, which would make more sense to me in a comment:

# Don't use the URIs from any tag to try to extract codec information. # We should not rely on the exact structure of file names for this. # Use stream_maps instead.

joeyparrish · 2021-08-23T18:48:33Z

streamer/m3u8_concater.py

-    return str(band), str(math.ceil(avg_band/sum(durations)))
+    return {
+        'BANDWIDTH': str(band),
+        'AVERAGE-BANDWIDTH': str(math.ceil(avg_band/sum(durations)))


Wow. I'm really surprised to check the spec and find that these values must be integers. Nice attention to detail!

I probably would have just written out a float, which would have worked fine in a JS player and may have broken some other player implementation.

joeyparrish · 2021-08-23T18:51:32Z

streamer/m3u8_concater.py

+          dummy = ',\ndata:text/plain,NO {} SUBTITLES\n'.format(lang.upper())
+          # Break the period's duration into target duration count and remains,
+          # because the target duration might be a prime number, this might
+          # cause an accumulated rounding error.


I mean some small accumulated rounding error here is fine. I think you're doing the right thing, and I think it's not worth the effort to do more until we have evidence that it's necessary.

joeyparrish · 2021-09-03T16:53:37Z

Just back from vacation. Will try to take another pass at review today if at all possible.

joeyparrish · 2021-09-03T17:14:11Z

streamer/m3u8_concater.py

-          if remains:
-            concat_txt_playlist.content += '#EXTINF:' + str(remains) + ',\n'
-            concat_txt_playlist.content += dummy
+          # If the no playlist were found for this period, we create at time gap


typo: at => a

mariocynicys added 29 commits June 9, 2021 09:04

multi-input & multi-period concat proposal

f3d5d93

doc string for inputs_list and single_inputs

d562720

adding period concatenation for dash

8fa9720

adding period concatenation for dash, no tests are done yet.

adding period concat to LIVE for DASH

e806d66

using regex for iso8601 duration parsing

f239954

reverting unrelated changes

0ce17f6

limiting to 2 conflicting/required-exclusive fields

3f0aaef

using BaseURL to specify the nested period directories

71e50b0

minor edits

6b51750

minor edits2

5a91827

minor edits3

9c5d4c1

minor edits4

e9cc4da

so to not to stuck Running if one packager node errors

17ca8c2

adding tests

95d7510

using getDuration() instead of seekRange()

d1dd0cf

style code

11ea3a2

text and video streams concatenation for hls

ab8574e

substitution for missing langs + audio concatenation

7f781b2

add the language attribute only if it's not undefined

079213a

support for audio-only period concatenation

9ffece7

minor edits

d8c90a6

removing support for non-existing audio or video for some periods

395b3cc

no need for this conditional

85c63ea

we should break after we find the nearest channel

3d987ea

cleaning

109a485

more cleaning

1125784

Merge branch 'multi-period' into HLS-merge-conflicts

18d6ae5

Merge pull request #2 from meryacine/HLS-merge-conflicts

a80c398

pully

to pass the assertion for the subtitle playlists

efdf6ea

joeyparrish requested changes Jul 20, 2021

View reviewed changes

mariocynicys added 6 commits August 14, 2021 15:49

amend to me

31922b7

open files for appending so we can have the logs from all the nodes

c2575de

1-indexing the period dircotry names

b367aae

no need to store the output_stream to the media playlist

5bbcf39

some clarifications, comments, doc strings, and assertions

0cdafc1

merging with master for a clear diff

9000caa

mariocynicys commented Aug 16, 2021

View reviewed changes

joeyparrish requested changes Aug 16, 2021

View reviewed changes

fixing typos

794225a

joeyparrish requested changes Aug 16, 2021

View reviewed changes

var instead of inf + spliting a long line

3cf34c8

mariocynicys commented Aug 16, 2021

View reviewed changes

forgot to raise it O_O

d1bd891

joeyparrish requested changes Aug 16, 2021

View reviewed changes

mariocynicys added 4 commits August 17, 2021 19:04

addressing the requested changes, revision-1

b1956da

addressing the requested changes, revision-2

f36ec02

indexing the trancoder and packager nodes

c1118ee

fix indexing

8cd1a38

joeyparrish requested changes Aug 23, 2021

View reviewed changes

mariocynicys added 2 commits September 2, 2021 11:05

omitting the count and remaining variables

1d7f98b

resolving conflicts

0539d10

joeyparrish requested changes Sep 3, 2021

View reviewed changes

typos :(

9a142dc

joeyparrish approved these changes Sep 3, 2021

View reviewed changes

joeyparrish merged commit a8ec6fc into shaka-project:master Sep 3, 2021

github-actions bot added the status: archived Archived and locked; will not be updated label Jul 25, 2023

github-actions bot locked as resolved and limited conversation to collaborators Jul 25, 2023

		"\tperiods with other periods that have video that is for the concatenation\n"
		"\tto be performed successfully.\n")

Multi period for HLS #83

Multi period for HLS #83

Conversation

mariocynicys commented Jul 20, 2021 • edited Loading

joeyparrish left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

joeyparrish left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mariocynicys Aug 16, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

joeyparrish left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

joeyparrish left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

joeyparrish commented Sep 3, 2021

Choose a reason for hiding this comment

mariocynicys commented Jul 20, 2021 •

edited

Loading

mariocynicys Aug 16, 2021 •

edited

Loading