Also as requested, pinging@Rogdham to review the Python code.

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phraseI have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

I am working on addressing some of the issues which will include the test failures

Thanks for making the requested changes!

@AA-Turner: please review the changes made to this pull request.

FWIW I believe I responded to all reviews, so I will add the buildbots to test this.

🤖 New build scheduled with the buildbot fleet by@emmatyping for commita12a031 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F133365%2Fmerge

If you want to schedule another build, you need to add the🔨 test-with-buildbots label again.

I've pushed a series of commits. Most are explanatory and fairly minor (docstring wording, etc). The one of most note is removing the exact type checks on{C,Dec}ompressionParamater. I think it is overly restrictivie to check the exact type of the enum, and I can't think of a similar example where we do this. Users who want the validation of not conflating compression and decompression can use a staic type checker.

I've pushed this for speed, given time, but if you feel strongly I'm happy for you to revert.

A

@AA-Turner The commits look fine other than the last one. It is very important to check the type of a CompressionParameter vs DecompressionParameter because the C enum variants have overlapping values (compression_level == window_log_max)

You can see in the tests this could lead to confusing errors:

Traceback (most recent call last):  File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_zstd.py", line 1424, in test_init_bad_mode    ZstdFile(io.BytesIO(), 'rb',    ~~~~~~~~^^^^^^^^^^^^^^^^^^^^             options={CompressionParameter.compression_level:5})             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/compression/zstd/_zstdfile.py", line 91, in __init__    raw = _streams.DecompressReader(        self._fp,    ...<3 lines>...        options=options,    )  File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/compression/_common/_streams.py", line 53, in __init__    self._decompressor = self._decomp_factory(**self._decomp_args)                         ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^_zstd.ZstdError: Error when setting zstd decompression parameter "window_log_max", it should 10 <= value <= 31, provided value is 5. (zstd v1.5.5, 64-bit build)

I think I will revert the last commit but keep the others.

It is very important to check the type of a CompressionParameter vs DecompressionParameter because the C enum variants have overlapping values

Yes, but I think the risk of this actually hitting users is low? For example, the examplar test failure you provided is very clear:CompressionParameter.compression_level is provided as a key, and the error message says 'Error when setting zstddecompression parameter "window_log_max"' (emph. added).typeshed could annotate theoptions dict asMapping[CompressionParameter, int] which would solve the problem for users of static type checkers. Users who pass nonsense or invalid values will always get undefined behaviour or an error. I think it's better to apply the consenting adults principle here that we see elsewhere in the language. For example, the current code doesn't let me pass anint as the key directly.

I think it's fine to keep for now as we can discuss further and change this during the betas -- this particular item isn't impacted by feature freeze as it's entirely private.

A

For example, the examplar test failure you provided is very clear: CompressionParameter.compression_level is provided as a key, and the error message says 'Error when setting zstddecompression parameter "window_log_max"' (emph. added).

I think we view the clarify of the error differently! To me I would expect users to not understand why they are getting an error about a decompression parameter when passing a compression parameter. I think perhaps a compromise would be to check the type if it is a (De)CompressionParameter, but otherwise just allow ints without checking. If you're passing an int, you're on your own!

@AA-Turner

For example, the current code doesn't let me pass an int as the key directly.

After writing this:

I think perhaps a compromise would be to check the type if it is a (De)CompressionParameter, but otherwise just allow ints without checking. If you're passing an int, you're on your own!

I just checked and we only check if it isCompressionParameter for decompression, orDecompressionParameter for compression. So the compromise is actually the current behavior. So you can pass an int if you'd like:

>>>from compression.zstdimport*>>> compress(b'testtesttest',options={100:5})b'(\xb5/\xfd \x0ca\x00\x00testtesttest'

Thanks everyone, and especially thank you@Rogdham for the valuable feedback & context from pyzstd.

Hear, hear! And thank you@AA-Turner for your suggestions and contributions! Thank you to all of the reviewers!