Fixing case where multiple fields have the same field name

[matt035343]

Fixes #270

Fixing case where multiple fields have the same field name by inverting the map holding the field name overrides

[github-actions]

Coverage Report

File	Stmts	Miss	Cover	Missing
dataclasses_json
cfg.py	51	4	92%	80, 84–86
core.py	238	10	96%	38–41, 51, 64, 66, 81, 83, 172, 200
mm.py	202	30	85%	33–36, 42–45, 53–56, 62–65, 88, 161–162, 167, 171, 175, 180, 184, 188, 196, 202, 207, 216, 221, 226, 235, 244–251
stringcase.py	25	3	88%	59, 76, 97
undefined.py	143	2	99%	24, 38
utils.py	128	36	72%	11–24, 44–49, 60–64, 74, 99–100, 108–109, 124–132, 158, 177, 202
tests
entities.py	226	2	99%	229, 235
test_annotations.py	81	42	48%	50–67, 78–102, 106–122
test_api.py	142	4	97%	88, 99, 139–140
test_str_subclass.py	22	1	95%	9
test_union.py	98	10	90%	87–94, 108–115
TOTAL	2502	144	94%

Tests	Skipped	Failures	Errors	Time
293	3 💤	0 ❌	0 🔥	6.945s ⏱️

[george-zubrienko]

I'm not sure, but is this a good test case? Like, declaring myJsonField as a regular dict will not trigger the fixed behaviour first place?

[matt035343]

@george-zubrienko I am not sure I understand. Are you suggesting to use a dataclass instead of a dict?

[george-zubrienko]

I mean the MyDataClass can be defined like this:

class MyDataClass:
    myJsonField: dict
    other_field: str = field(metadata=config(field_name="otherField"))
    
    @property
    def first(self) -> str:
        return self.myJsonField.get('first', '')

    @property
    def second(self) -> str:
        return self.myJsonField.get('second', '')

[george-zubrienko]

Just IMO, the way the test case is defined, dataclass does not reflect the actual data structure being read, but rather relies on some rather artificial deserializing behaviour, which for example in v1 warrants a custom serializer for the class. If this is something people actually do in their user code, I'd like to see examples. I am near 100% certain we will not support these kind of cases in v1 and force people to write custom serializer, thus I'd be wary of using these as test cases for v0 as well.

[matt035343]

@george-zubrienko This is PR spawned from #270. Indeed, it is a funky usecase, however, the core here is that two dataclass-json field names maps from the same encoded JSON field name

[matt035343]

@george-zubrienko and I discussed that it is actually a bad practice do like #270 does. It will only work one way (decoding only, not encoding). Instead, use the approach @george-zubrienko outlined above.

I will discontinue this PR and make a new one that throws an exception instead.