feat(serializer): add snapshot regex value matcher and bypass custom repr helper #791
Conversation
poetry.lock
Outdated
There was a problem hiding this comment.
Possibly a mismatch in poetry versions?
There was a problem hiding this comment.
Ran inv install again, jic.
| def bypass_custom_repr(cls, data: Any) -> "tuple[Any, ...]": | ||
| attr_names = list(cls.object_attrs(data)) | ||
| return collections.namedtuple(data.__class__.__name__, attr_names)( | ||
| **{prop: getattr(data, prop) for prop in attr_names} | ||
| ) |
There was a problem hiding this comment.
I was thinking about how the namedtuple serialization is exactly what we want when it came to #784.
Regarding the name of the method, this does not really "bypass" custom repr, it's just a utility for constructing a named tuple based on an object's attrs. If it's "bypassing" some serialization, I'd expect it to somehow "communicate" with the serializer itself. I.e. a model where it sets a marker on an object so that the serializer knows to handle it differently.
Instead, I'd rename this to more explicitly be a utility that constructs a named tuple from a custom object. Perhaps def make_named_tuple or def create_named_tuple / def convert_obj_to_named_tuple.
If the idea is to use this in conjunction with the replacer to provide a more elegant solution to the msgspec serializer issue, we'd want a mechanism to apply the replacer to all snapshots. It gets verbose having to specify the replacer per use.
Another way I was thinking of approaching the problem in the long run was to expose some static "registry" / map of custom class types -> serializers to register with the class / or to override.
e.g.
class Serializer(AmberDataSerializer):
custom_classes = { msgspec.Struct: some_serializer }
There was a problem hiding this comment.
Changed the name to object_as_named_tuple_class.
It gets verbose having to specify the replacer per use.
True, the replacer is a low hanging fruit that allows both functionality I wanted to add here. I like the custom_classes idea, it could even be used in the __matcher assertion property.
Another option would be to allow a change to the _matcher property that won't be reverted after each assertion.
@pytest.fixture
def snapshot_bypass_my_custom_repr(snapshot):
snapshot._matcher = path_type(
types=(MyCustomReprClass,),
replacer=lambda data, _: AmberDataSerializer.object_as_named_tuple(data),
)
return snapshot
def test_snapshot_object_as_named_tuple_class(snapshot_bypass_my_custom_repr):
assert MyCustomReprClass() == snapshot_bypass_my_custom_repr(
exclude=props("prop1"),
)
# [4.2.0](v4.1.1...v4.2.0) (2023-08-21) ### Features * **serializer:** add snapshot regex value matcher and bypass custom repr helper ([#791](#791)) ([3ac2ce8](3ac2ce8))
|
🎉 This PR is included in version 4.2.0 🎉 The release is available on:
Your semantic-release bot 📦🚀 |
Description
bypass_custom_reprto help with serializing objects with custom third party classreprimplementationsreplacerparameter to thepath_typematcher signature to allow custom replacement actions on matched property valuespath_valueto support regex match replacement on data valuesregexflag from the matcher to simplify the api, since the matcher and string compare serve equivalent functionsRelated Issues
Checklist
Additional Comments
No additional comments