How can the API be used to select the best stream object from a list that contains multiple stream objects？

[xueli12]

Description

I initially retrieved the list of stream objects - video_streams via DASHStream.parse_manifest.

Next, I filtered the video_streams, and chose the stream objects that fit the stream.video_representation.mimeType == "video/mp2t" criteria to form the filtered_streams.Is it possible to carry out this filtering operation using an API?

I attempted to select the stream object with the highest video_representation.bandwidth from the filtered_streams, but I used my own method to do this.Can the operation of filtering be done using the streamlink API? For object, if I put in filtered_streams and best as the parameters, can the API return the best_stream? Which specific API do I use and how does it work?

The following is the code I used.

import streamlink
from streamlink_cli.output import FileOutput
from streamlink_cli.main import open_stream
from streamlink_cli.streamrunner import StreamRunner
from streamlink.stream import StreamIO, HTTPStream, HLSStream, DASHStream

url = f"https://apis.naver.com/neonplayer/vodplay/v2/playback/{video_id}?key={in_key}"

 session = streamlink.session.Streamlink({
        'stream-segment-timeout': 60
    })
video_streams = DASHStream.parse_manifest(session, url, headers={"Accept": "application/dash+xml"})

for name in video_streams:
       print(f"Stream name: {name}, Stream object: {video_streams[name]},Stream mimeType: {video_streams[name].video_representation.mimeType}")

filtered_streams = {name: stream for name, stream in video_streams.items() if stream.video_representation.mimeType == "video/mp2t"}

best_stream_name = max(filtered_streams, key=lambda name: filtered_streams[name].video_representation.bandwidth,
                           default=None)
best_stream = filtered_streams[best_stream_name] if best_stream_name else None

stream_writer(best_stream, filename, output_path)

[bastimeyer]

Why are you doing this custom stuff when the chzzk plugin has just been merged into master?
https://github.com/streamlink/streamlink/commits/fef6fd66d1a4499cadd13c4ae76ffe1d8f427805/

Also, streamlink_cli is not a public interface. You can't rely on its stability. So whatever you're doing with it, you're on your own and won't receive any support in third party implementations.

DASHStream.parse_manifest() returns a dict[str, DASHStream] object where the keys are the "quality names" defined by the DASH implementation (limiting - see below). DASHStream instances have a video_representation and an audio_representation which both have different attributes. If you want to get the "best" stream from the returned dict manually, then you have to read those attributes and do your own sorting:
https://streamlink.github.io/api/stream.html#streamlink.stream.DASHStream.parse_manifest

Plugin.streams() however is the actual public interface for getting an ordered/ranked dict of streams, according to the stream "weights" that can optionally be overridden by the plugins themselves:
https://streamlink.github.io/api/plugin.html#streamlink.plugin.Plugin.streams

This whole design will be reworked at some point in the future, because it was developed during the early livestreamer days and has never been improved upon. It's very limiting today. Sorting streams via a public interface is tied to the plugins systems.
#4902

[xueli12]

Thank you very much for your suggestions. I really haven't noticed that the chzzk plugin has already been merged before.The main reason for me writing this code is that I need to visit [https://api.chzzk.naver.com/service/v2/videos/{video_id}](https://api.chzzk.naver.com/service/v2/videos/%7Bvideo_id%7D) to get some information about the video in advance for the filename, and when getting this information, I can also get the mpd link. However, not all of this filename information can be obtained through streamlink.If using the chzzk plugin, the chzzk plugin will request [https://api.chzzk.naver.com/service/v2/videos/{video_id}](https://api.chzzk.naver.com/service/v2/videos/%7Bvideo_id%7D) one more time during the execution process. In order to reduce this request, I want to directly push the mpd link to the streamlink API for processing.
Following your advice, I used the sorting_excludes parameter of Plugin.streams() to temporarily solve the problem of filtering video streams.