Use default pydantic model for testing

Previously we changed the pydantic models used in testing to be
the compatibility models that understand v1 and v2. This is fine
but means that any external users using pydantic models as
butler datasets break if they are still using v1.

Change the test model to use whatever is the native pydantic
and then the YAML formatter has been modified to support
both APIs. The JSON formatter already supported both.

python/lsst/daf/butler/formatters/yaml.py

@@ -152,15 +152,26 @@ def _toBytes(self, inMemoryDataset: Any) -> bytes:
         This will fail for data structures that have complex python classes
         without a registered YAML representer.
         """
+        converted = False
         if hasattr(inMemoryDataset, "model_dump") and hasattr(inMemoryDataset, "model_dump_json"):
-            # Pydantic-like model if both dump() and json() exist.
+            # Pydantic v2-like model if both model_dump() and model_json()
+            # exist.
             with contextlib.suppress(Exception):
                 inMemoryDataset = inMemoryDataset.model_dump()
+                converted = True
+
+        if not converted and hasattr(inMemoryDataset, "dict") and hasattr(inMemoryDataset, "json"):
+            # Pydantic v1-like model if both dict() and json() exist.
+            with contextlib.suppress(Exception):
+                inMemoryDataset = inMemoryDataset.dict()
+                converted = True
+
+        if not converted:
+            if dataclasses.is_dataclass(inMemoryDataset):
+                inMemoryDataset = dataclasses.asdict(inMemoryDataset)
+            elif hasattr(inMemoryDataset, "_asdict"):
+                inMemoryDataset = inMemoryDataset._asdict()
 
-        if dataclasses.is_dataclass(inMemoryDataset):
-            inMemoryDataset = dataclasses.asdict(inMemoryDataset)
-        elif hasattr(inMemoryDataset, "_asdict"):
-            inMemoryDataset = inMemoryDataset._asdict()
         unsafe_dump = self.writeParameters.get("unsafe_dump", False)
         if unsafe_dump:
             serialized = yaml.dump(inMemoryDataset)

python/lsst/daf/butler/tests/_examplePythonTypes.py

@@ -44,7 +44,7 @@
 from typing import TYPE_CHECKING, Any
 
 from lsst.daf.butler import StorageClass, StorageClassDelegate
-from lsst.daf.butler._compat import _BaseModelCompat
+from pydantic import BaseModel
 
 if TYPE_CHECKING:
     from lsst.daf.butler import Butler, Datastore, FormatterFactory
@@ -264,7 +264,7 @@ def makeFromDict(cls, exportDict: dict[str, list | dict | None]) -> MetricsExamp
         return cls(exportDict["summary"], exportDict["output"], data)
 
 
-class MetricsExampleModel(_BaseModelCompat):
+class MetricsExampleModel(BaseModel):
     """A variant of `MetricsExample` based on model."""
 
     summary: dict[str, Any] | None = None
@@ -274,7 +274,12 @@ class MetricsExampleModel(_BaseModelCompat):
     @classmethod
     def from_metrics(cls, metrics: MetricsExample) -> MetricsExampleModel:
         """Create a model based on an example."""
-        return cls.model_validate(metrics.exportAsDict())
+        d = metrics.exportAsDict()
+        # Assume pydantic v2 but fallback to v1
+        try:
+            return cls.model_validate(d)
+        except AttributeError:
+            return cls.parse_obj(d)
 
 
 @dataclasses.dataclass