Clickify cleanup

[o-smirnov]

@landmanbester please check if this works with PFB.

I've removed the pass_missing_as_none logic from clickify_parameters(), as upon closer investigation, this is what click now does always and anyways, except when dealing with some multiply-valued parameters (that is, Lists with a policy of repeat: repeat, in which case a missing parameter is passed as an empty tuple). I don't see any way to change this behaviour. The new code checks for UNSET defaults, and doesn't give click a default value for them, which results in a None being passed to the callable in the absence of the parameter. I think this behaviour is harmless and unlikely to trip us up, except when you have something perverted like

def func(a: float=3):
  print(a)

with a schema that doesn't mention the default from the function signature:

a:
  dtype: float
  required: false

...which results in func(a=None) when the command-line doesn't specify --a. This is because neither stimela nor click knows anything about the signature actually, and has no choice but to trust what the schema says.

Note that pass_missing_as_none is still honoured by stimela itself when invoking a cab (i.e. is the difference between having nothing in **kw, or having a None value in **kw). Given your love for **kw, I think this is what you expect, right?

[o-smirnov]

Run tests/scabha_tests/test_clickify.py to test this behaviour (and feel free to extend).

[landmanbester]

This seems ok to me. It will require handling optional lists differently in pfb but that should be doable. More worrying is the numpy dependency here

stimela/pyproject.toml

Line 18 in 36d1abb

numpy = "^1.20"

Not sure if this relates to the current setuptools issues but even with setuptools[core] I get the following conflicts

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
codex-africanus 0.4.1 requires numpy<3.0,>=2.0, but you have numpy 1.26.4 which is incompatible.
dask-ms 0.2.23 requires numpy>=2.0.0, but you have numpy 1.26.4 which is incompatible.

What does stimela use numpy for anyway?

[o-smirnov]

Oh fsck numpy, I was only using it in one place, and I can work around without it. I've taken the dependency out, try again.

It will require handling optional lists differently in pfb but that should be doable.

Well nothing's changed there -- it seems that click was passing optional lists as () all along, even with pass_missing_as_none in effect. So if you depend on optional_list is None and don't cope with empty lists, your CLI has always been slightly broken. But checking for an empty optional list or None with if optional_list: is probably harmless to do in most cases, right? Unless "empty list" and "None" are two distinct logic cases for you, which would be a bit perverse.

[landmanbester]

Seems to be working for me. Empty lists are still parsed as None for some reason though so this does not really change anything in pfb. I guess it's because I don't have a repeat policy set. Note I will be mostly AFK until I get back from leave in early Jan

[o-smirnov]

OK well if @Athanaseus approves, I'll merge this and do a point release, so he can get his breizorro PR in as well.