You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I wish adding additional formats didn't require monkeypatching, but in the meantime this is one of the ones that files containing it exist in the wild.
I chatted with @thatch a bit at the PyConUS sprints this year. Deflate64 is a proprietary compression format, but it is generated by a few tools and dotnet. It's not much better than Deflate, and newer tools should probably use a newer format (i.e. LZMA/Zstandard). The use case is mostly for analyzing existing files.
inflate64 is available for current Python versions and genereally seems to be more actively maintained than zipfile-deflate64, so it might be a better candidate for deflate64 support.
As Windows uses deflate64 by default for bigger files / archives I think this compression is much more common than thought.
Activity
emmatyping commentedon May 20, 2025
I chatted with @thatch a bit at the PyConUS sprints this year. Deflate64 is a proprietary compression format, but it is generated by a few tools and dotnet. It's not much better than Deflate, and newer tools should probably use a newer format (i.e. LZMA/Zstandard). The use case is mostly for analyzing existing files.
gpshead commentedon May 22, 2025
#113767 seems like it'd be a way for deflate64 PyPI package to add support without the monkeys.
quassy commentedon Aug 15, 2025
inflate64 is available for current Python versions and genereally seems to be more actively maintained than zipfile-deflate64, so it might be a better candidate for deflate64 support.
As Windows uses deflate64 by default for bigger files / archives I think this compression is much more common than thought.
emmatyping commentedon Oct 16, 2025
Hm, reading online it seems this is used when the zip file becomes >2GB, so this isn't super common but not uncommon either.