SQLite-based read-write lock

[leventov]

Draft for #307.

@gaborbernat @Yard1 please let me know how I should change the code from the API perspective. There are quite a bit very elaborate stuff that is hard to understand regarding thread-locals, etc.

[gaborbernat]

Please align comments to 120 charachters 🙇

[gaborbernat]

Prefer

from os import PathLike

[leventov]

@gaborbernat I've pushed a more "serious" attempt.

[gaborbernat]

Looks better, can you add tests please?

[leventov]

@gaborbernat what do you think about using SQLite's VFS interface directly, without the "dummy database" thing? https://github.com/rogerbinns/apsw provides bindings

[gaborbernat]

It's important that we have no dependencies. The only alternative could be to hide this functionality behind some extra, I do not know how widely available the bindings are in form of wheels.

…

On Mon, Feb 24, 2025, 21:58 Roman Leventov ***@***.***> wrote: @gaborbernat <https://github.com/gaborbernat> what do you think about using SQLite's VFS interface <https://www.sqlite.org/vfs.html> directly, without the "dummy database" thing? https://github.com/rogerbinns/apsw provides bindings — Reply to this email directly, view it on GitHub <#399 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFIQPULETXMCF2DNC5TAZD2RQA7HAVCNFSM6AAAAABXO35KTOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOBQGY4TIOBZG4> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: leventov]*leventov* left a comment (tox-dev/filelock#399) <#399 (comment)> @gaborbernat <https://github.com/gaborbernat> what do you think about using SQLite's VFS interface <https://www.sqlite.org/vfs.html> directly, without the "dummy database" thing? https://github.com/rogerbinns/apsw provides bindings — Reply to this email directly, view it on GitHub <#399 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFIQPULETXMCF2DNC5TAZD2RQA7HAVCNFSM6AAAAABXO35KTOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOBQGY4TIOBZG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[leventov]

@gaborbernat I've added tests.

If you want to finesse API (context managers vs. parameters to acquire_read() and acquire_write() methods; are read_lock() and write_lock() methods even needed? Are instance's default timeout and blocking needed?), docs, formatting, class/method naming, etc., could you please do it yourself to your liking?

[gaborbernat]

CI failing.

[leventov]

@gaborbernat I don't need this PR to be merged to use it later through filelock library. In fact, I submitted a PR only to attract code review and discussion. Myself, I'm going to use this code copied (and modified to an extent) in the codebase. So, I'm not invested in the merge per se.

What of these seem valid to you, and what seem like false positives?

src/filelock/_read_write.py:61: error: Returning Any from function declared to return "ReadWriteLock"  [no-any-return]
src/filelock/_read_write.py:61: error: "_ReadWriteLockMeta" has no attribute "get_lock"  [attr-defined]
src/filelock/_read_write.py:62: error: Returning Any from function declared to return "ReadWriteLock"  [no-any-return]
src/filelock/_read_write.py:67: error: Need type annotation for "_instances"  [var-annotated]
src/filelock/_read_write.py:77: error: Argument 2 for "super" not an instance of argument 1  [misc]
src/filelock/_read_write.py:91: error: Returning Any from function declared to return "ReadWriteLock"  [no-any-return]

Can the CI auto-fix from "ReadWriteLock" to ReadWriteLock itself cause some of these errors?

check / test pypy3.10 - windows-latest is test-suite related, perhaps even CI environment-related.

[gaborbernat]

Moved to draft until we fix the issues.