Update References with pointer to target object

[jmuhlich]

Adds a .ref to all Reference subclasses that returns the referenced model
object. There is an underlying ref_ field to support this that is only filled
in by the root OME object's constructor. A future API addition will be
created to support proper creation of References.

Closes #27

[jmuhlich]

I just realized I didn't write any tests for this. I understand if you want to hold off accepting this until I get around to that.

[codecov-commenter]

Codecov Report

Merging #34 into master will decrease coverage by 0.06%.
The diff coverage is 97.63%.

@@            Coverage Diff             @@
##           master      #34      +/-   ##
==========================================
- Coverage   95.20%   95.13%   -0.07%     
==========================================
  Files           1        1              
  Lines         396      473      +77     
==========================================
+ Hits          377      450      +73     
- Misses         19       23       +4     

Impacted Files	Coverage Δ
src/ome_autogen.py	95.13% <97.63%> (-0.07%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a9d5fc4...beab9ee. Read the comment docs.

[tlambert03]

just noticing that we're building a fair amount of string literals in ome_autogen
for some of these added methods, like the class overrides here, how would you feel about putting the autogenerated code at ome_types._model, then creating a new ome_types.model module (checked into source), that is mostly a passthrough module, but reimplements a few of the classes with additional methods... eg:

#ome_types/model/ome.py
from ome_types._model.ome import OME as _OME

class OME(_OME):
    def __post_init_post_parse__(self: Any, *args: Any) -> None:
        ids = collect_ids(self)
        references = collect_references(self)
        for ref_id, ref in references.items():
            ref.ref_ = weakref.ref(ids[ref_id])

thoughts?

[jmuhlich]

Good idea. Or maybe the reverse, where the bases are committed and the generated classes extend those. Another option might be to write “loose” functions in a committed module and bind them to the classes rather than using inheritance. We’ll have to see which works best.

[jmuhlich]

What do you think about this PR as it stands? Maybe I add some tests and if you're happy with the API and approach we can merge it, and the next PR will work on moving all this fixed code into a separate module.

[tlambert03]

sorry for the delay, I'm happy working with this approach. Couple basic tests would be good and then we can merge

[tlambert03]

just tried to write a very basic test for this:

def test_collect_ref(model):
    from ome_types import from_xml
    from ome_types.util import collect_references

    ome = from_xml(TESTING_DIR / "data" / "filter.ome.xml", model.OME)
    expected = {
        "Dichroic:1",
        "Filter:5",
        "Filter:1",
        "Filter:Dichroic:2",
        "FilterSet:2",
        "Filter:6",
        "Instrument:0",
        "Objective:0:0",
        "Filter:2",
        "Filter:3",
        "Filter:4",
    }
    assert set(collect_references(ome)) == expected

that "expected" set is what I get when I run that exact code interactively... but for some reason when I run it in the test, set(collect_references(ome)) is an empty set... any idea what's going on there?

[jmuhlich]

The model fixture imports the model module as test_model0, so the isinstance(value, Reference) check in collect_references never returns True because test_model0.reference.Reference is not the same as ome_types.model.reference.Reference. I tried running with --nogen but then I got an error from a circular import for some reason.

[tlambert03]

ah, ok. thanks, will see what I can do to clean up that testing arrangement.

[tlambert03]

since most of the testing difficulty here is unrelated to this PR, and has more to do with the way I setup testing a dynamically generated import, it's fine with me if you want to just merge this PR when you're happy with it, and we can create an issue to add tests later