Make checking function accessible from python interpreter

[aulemahal]

Hi all!

Maybe I should have opened an issue before pushing a PR ? Anyway, this took me only a few minutes.

In a goal of running check_cmip7_packing from inside a python script without having to make a system call and parse the result, I suggest here taking the core checking functionality and making it a standalone function that can be imported.

Reading errors are still managed externally and parsing errors are passed along using a custom error class. The function returns a boolean and a string detailing the reason for the result.

I'm not sure if this was needed, but without this change the check_cmip7_packing script was usable even if cmip7_repack was not installed in a python environment. I tried to preserve this possibility here with a try..except when importing.

Tested by running one two files (one passing and one failing) from inside an interpreter:

from cmip7_repack import check_cmip7_packing

check_cmip7_packing(filename)

and from the CLI. Also ran the same "tests" from the CLI outside the environment (pyfive available, but cmip7_repack not installed).

For my use case, this is the minimal importable stuff I need, but it could be interesting to also have access to the functions in cmip7_repack that compute the metadata blocksize and chunksizes, etc. I could maybe suggest another PR if there's interest.

What do you guys think ? Is this useful ?

[jesusff]

Pascal, this is maybe related to these functions which already exist in the quality checker
https://github.com/ESGF/cc-plugin-wcrp/blob/54398a09d208e2d8141d24087e3c734df241dea6/checks/format_checks/check_internal_packing.py

The code is becoming hard to maintain as copies and forks are proliferating to be able to tweak the default hard-coded behaviour. A flexible central code that others can just invoke with different parameters would be simpler

[aulemahal]

Ah yes indeed, I didn't see that.
I thought this repo was the central code which is why I suggested this here!

[valeriupredoi]

@aulemahal Pascal, many thanks, mate! As noted above, this is the source repo for the CMIP7 repack checker, but it's not where it's being used by ESGF; your implementation makes sense in order to add a bit more reach via Python API, and I support it, but I am also CCing my colleagues @davidhassell and @zequihg50 that are the actual developers of the package - I am devops, and from my packaging/operational PoV it's fine, but we'd need a few more tests from you 🍺

[valeriupredoi]

ping @davidhassell and @zequihg50 gents, what say you about this implement?