prepare for release?

[newville]

@auxym Would you be willing to release and push to PyPI? I would be willing to help, if that would be useful.

I might suggest moving src/zarr_sqlite to zarr_sqlite so that a development-mode from import zarr_sqilte import SqliteStore could work.

[auxym]

Yes, I am willing to do this, but I will wait for the zarr devs final decision on zarr-developers/zarr-python#3319.

Do you think there is are any modifications necessary to the schema (anything that could break file compatibility)? I was thinking of leaving an RFC open for a few days/weeks so other interested parties may chime in on a "final" schema. Any implementation/performance details (like async compatibility) could always be added later without breaking file-level compatibility

The current schema is identical to the zarr v2 SQLiteStore schema, which is an advantage (may be compatible with older files or other zarr-v2 implementations in other languages).

[newville]

@auxym Thanks -- I have to admit I have not done a lot of work with this code beyond the tests/examples. I was very much hoping that
zarr-developers/zarr-python#3319 would at the very least result in "yes, make a PR and let's start a discussion".

I was unpleasantly surprised to see SqliteStore dropped in Zarr3 just as I was ready to start using Zarr in production code - that's been delayed for a year now. I saw your Issue and the initial positive response, and thought "oh, great, that will get restored". I am shocked at the careless attitude to your efforts, basically ignoring your code for 9 months, and only coming back to looking at this when nudged. I feel like they really should all thank you and apologize for the neglect. So, let me clearly say Thank you very much for working on this, and I am very sorry to see the negative responses.

From looking at the code (but not really stress testing), I might suggest:
a) for sqlite3.threadsafety (zarr_sqlite.py#L135), you might try setting sqlite3.threadsafety = 3, which looks to be supported in Python 3.11 and later. It might be reasonable to simply require Python 3.11. for this module.
b) I have no opinion on the use of aiosqlite in zarr-developers/zarr-python#3319 (comment) . I don't really know anything about it. Have you any thoughts on this?

I suggest that an initial release might be clearly marked as such, with a request for comments, suggestions, collaborations. Maybe someone will have good suggestions.

Thanks again.