Tests also require stardard-chunk and standard-audioop, the latter of which is not currently available from pypi.

@Kangie where is chunk being used? Can't see any references to it in the repo.

Yes, looks like audioop might be a problem. I can see one place it gets used:

It came up when trying to update the gentoo package, possibly an indirect dependency via either sunau or aifc that's come up trying to invoke them in the test suite?

Sorry in the middle of a huge system update, but I did stash the ebuild I was using where I added it as a dependency:

https://github.com/Kangie/gentoo/blob/283c7d14199454342c7be9e4ab00650f15cb5583/dev-python/audioread/audioread-3.0.1.ebuild

Perhaps it's worth using https://github.com/AbstractUmbra/audioop for that instead.

Honestly if we're up to 3 or 4 "dead batteries" I wonder if it's better to find proper replacements for all of them. These were dropped for a reason after all and this is just a band-aid fix.

Honestly if we're up to 3 or 4 "dead batteries" I wonder if it's better to find proper replacements for all of them. These were dropped for a reason after all and this is just a band-aid fix.

That does sound reasonable. However, personally I don't know enough about low-level audio file modification to be able to contribute much to that endeavour.

Tests also require stardard-chunk and standard-audioop, the latter of which is not currently available from pypi.

I've dug into this a bit further. Both standard-aifc and standard-sunau include audioop-lts as a dependency. (Sorry, I should've realised that was the case - looking at the commit history, I was the one who added it 🤦 ).

Running the tests in github actions appears to work for me: https://github.com/lucas42/audioread/actions/runs/11599693327/job/32298399085 (there's some warnings thrown, but no errors).

@Kangie are you running tests and/or pulling in dependencies differently to how it's configured in the github workflow in this repo?

I don't see any reason to re-implement these codecs, especially since the code has MIT-compatible licenses.
Could the code be included wholesale in the audioread repo?

I believe these are the relevant files:
https://github.com/python/cpython/blob/3.11/Modules/audioop.c
https://github.com/python/cpython/blob/3.11/Modules/clinic/audioop.c.h
https://github.com/python/cpython/blob/3.11/Lib/sunau.py
https://github.com/python/cpython/blob/3.11/Lib/aifc.py

It might not be hard to convert audioop to pure python for PyPy compatibility.

@kamocat do you know anyone who'd be willing to convert audioop to pure python? I don't have enough knowledge of audio encoding to do it myself.

Regarding the files you link to for sunau and aifc, that's where the standard-aifc and standard-sunau packages got their source from. I don't have a strong opinion about whether they get included through pip or included in this repo. I guess it comes down to which will get the best ongoing support (at the moment, neither option seems like it'll see much love).
I picked the option with the simplest changeset in the hope it wouldn't be too controversial.

You're right. Using the standard-* options should be fine, as long as it sticks around.
It seems the reason they were dropped in PEP- 954 is that they were considered to be "old, obsolete file formats". Nothing about them was broken or expected to break. These file formats aren't going to change.

Another option is to use FFmpeg to read the AIFF and SUN AU formats.