[heads up] The default compression level of gzip.GZipFile object is changed to 6 in Python 3.15

[befeleme]

In Python 3.15, the default compression level is reduced to 6.

Changed in version 3.15: The default compression level was reduced to 6 (down from 9). It is the default level used by most compression tools and a better tradeoff between speed and performance.

DeterministicGzipFile inherits from that class, and sets the compression level to 9 (two places with the default).

nibabel/nibabel/_compression.py

Line 85 in db76723

compresslevel: int = 9,

The tests for this object started failing on the nonmatching checksums with Python 3.15.0a6 in Fedora Linux, where I test the readiness of packages to support new Python versions.

This can either be mitigated by matching the compression level to Python's if the class is supposed to closely follow upstream, or increase the compression level of the gzip's GZipFile instance in the test:

nibabel/nibabel/tests/test_openers.py

Line 369 in db76723

with GzipFile(filename='', mode='wb', fileobj=fobj, mtime=0) as gzobj:

I tested the first approach locally on my machine, and the tests passed.

[effigies]

I suppose I would lean toward using whatever the default is for the version of Python that is running. Used to be, we'd just use **kwargs or say compresslevel=None and then only include it as a kwarg if someone passes it explicitly. I'm not 100% on how well those strategies work with type checking.

If neither of those is easy to do in a way that satisfies a type-checker, you could set compresslevel: int = -1 and then just not pass it as a kwarg if the value is -1.