Make subtitle language selection more powerful #6264
Conversation
|
I wish you had discussed your idea before trying to implement it. The behavior of existing options should not be changed. I have another idea for how to address these without breaking current compat. |
Could you elaborate? |
Yeah I was thinking about the design and implementation was easy enough to do quickly, so no worries if this does not get accepted.
Sounds good. My only feedback for the design is to allow selecting the default behavior (
I tried preserving backward compatibility. But then there needs to be a way to have "simple merge" as is the current default and then be able to choose "no merging" to download both normal+auto or "special merging" (see PR description for the terminology). Thus two operators. maybe Looking forward to your design. |
pukkandan
force-pushed
the
master
branch
2 times, most recently
from
ee280c7 to
7aeda6c
Compare
|
Closing this for now. |
Description of your pull request and other information
--sub-langsshould allow selecting one or more subtitle languages of the following types: normal, auto, both, normal if available else auto. Currently--sub-langsdoes not support all such combinations (see #2262). For example, current behavior is to merge: a normal sub is selected if present otherwise the auto sub. Let's call this a simple merge. There is currently no way to download both normal and auto sub for the same lang.When
--sub-langsis not specified, the current default behavior (see my PR #6240) is to select the first matching sub in the following order. Let's call it a special merge.enen. This is to handle cases where subs are named asen-USoren-qlPKC2UN_YU.enenensubs were found.Note that this does not allow the ability to download only
ensubs. Thus, ifensubs are missing, it will download an arbitrary (first) non-en lang.This PR makes language selection much more flexible be making the following major changes.
auto-(e.g.,en->auto-en). The main reason is to enable selecting auto subs separately from the normal subs (e.g.,en.*vsauto-en.*). A secondary benefit is that auto subs are named differently on disk, which make is easy to distinguish them from normal subs just from the file name (also see handle subs and auto-subs separately #2262 (comment)).--sub-langsand introduces a new special merging operator#<lang>to selectively enable merging as per the default behavior described above, with the change that if<lang>is not found, no other lang will be arbitrarily selected.en/frwill downloadenif available, elsefrif available, else nothingI'm giving a comprehensive set of examples to show the new behavior of
--sub-langs. The table assumes that subs are written to disk only. See #630 (comment) for corresponding embedding options.--write-subs--write-auto-subs--write-subs --write-auto-subs--write-subs --write-auto-subs --sub-langs 'all'enandauto-en--write-subs --write-auto-subs --sub-langs '#all'enif present elseauto-en, same as existingall--write-subs --write-auto-subs --sub-langs 'allnorm'--write-subs --write-auto-subs --sub-langs 'allauto'--write-subs --write-auto-subs --sub-langs 'en'ennormal subs--write-subs --write-auto-subs --sub-langs 'auto-en'enauto subs--write-subs --write-auto-subs --sub-langs 'en,auto-en'ennormal and auto subs--write-subs --write-auto-subs --sub-langs 'en/auto-en'ennormal sub if present else auto subsen--write-subs --write-auto-subs --sub-langs '#en'ensub after special merge--write-subs --write-auto-subs --sub-langs 'en,ja,auto-ja,es'--write-subs --write-auto-subs --sub-langs 'en.*,auto-en.*'--write-subs --write-auto-subs --sub-langs 'allnorm,auto-en-orig/auto-en'en-origauto sub if present elseenauto sub--write-subs --write-auto-subs --sub-langs 'allnorm,auto-.*-orig'*-origauto sub--write-subs --write-auto-subs --sub-langs 'en/auto-en/fr/auto-fr'frwill only be selected ifenis not present--write-subs --write-auto-subs --sub-langs '#en/fr'The above design was chosen to make the code and options simpler, but it breaks existing behavior. Backward compatibility can be maintained at the cost of more complex code and likely needs 2 operators instead of 1. To summarize the break in behavior:
--write-subs --write-auto-subs--write-subs --write-auto-subs --sub-langs 'all'--sub-langs '#all'for old behavior.--write-subs --write-auto-subs --sub-langs 'en'enis selectedenis selected. Use--sub-langs 'en/orig-en'for old behavior.TODOs
Let me know if this is an acceptable way to proceed. Once we come to a decision on the main design: renaming auto subs,
#and/operators and the general code, I can work on the TODOs.Fixes #2262 #4178
Template
Before submitting a pull request make sure you have:
In order to be accepted and merged into yt-dlp each piece of code must be in public domain or released under Unlicense. Check one of the following options:
What is the purpose of your pull request?