Jonas Kittner

Yes, the kitql solution only works w/o compression. For now disabling it solved the problem for me. I subscribed to the other issue as well.

#346 should resolve those 3.11 test failures - it looks like it captures the rasterio warnings?

There are no 3.9 wheels for crick on pypi. It tries to install version 0.0.6 which is compiled with numpy < 2 and hence incompatible with numpy >= 2. ```...

macos is missing the postgres binaries: `pg_config` (or they are not on `PATH`) that's why it fails compiling `psycopg2` and building a wheel. At least widnows and 3.9 works now...

Mhm, the test failures seem unrelated?! ``` E ACTUAL: array(3608.971706) E DESIRED: array(3608.9712, dtype=float32) ``` Looks like a floating point glitch? Maybe the tolerances need adjusting? Or it used to...

Upon further investigation this is caused by numpy >=2.3.0 - the mean in our case is calculated via crick. Since crick uses the C-API it's likely a change there -...

Mhm, not sure why it does not like `Iterable[_Choice | str]` This seems to work (explicitly defining unions of ` tuple[str, ...] | list[str]`), but it is quite ugly... (not...