✅ Add regression tests for hydra

Author: jemrobinson

tests/config/test_hydra.py

@@ -0,0 +1,106 @@
+import inspect
+from importlib.resources import files
+
+import pytest
+from hydra import compose, initialize_config_dir
+from hydra.core.global_hydra import GlobalHydra
+from omegaconf import DictConfig
+
+from icenet_mp.cli.hydra import hydra_adaptor
+
+
+class TestHydraConfigLoading:
+    """Regression tests for icenet-mp config composition via hydra."""
+
+    CONFIG_DIR = str(files("icenet_mp.config"))
+
+    def setup_method(self) -> None:
+        GlobalHydra.instance().clear()
+
+    def teardown_method(self) -> None:
+        GlobalHydra.instance().clear()
+
+    def load_config(self, overrides: list[str] | None = None) -> DictConfig:
+        """Compose the sample config from the icenet-mp config directory with any overrides."""
+        with initialize_config_dir(config_dir=self.CONFIG_DIR, version_base=None):
+            return compose(config_name="sample", overrides=overrides or [])
+
+    def test_sample_config_has_expected_top_level_keys(self) -> None:
+        cfg = self.load_config()
+        for key in ("data", "model", "train", "loss", "predict", "evaluate"):
+            assert key in cfg, f"Key '{key}' missing from composed config"
+
+    def test_model_group_overridden_by_sample(self) -> None:
+        # sample.yaml uses `override /model: quick_test`, replacing the base default
+        cfg = self.load_config()
+        assert cfg.model.name == "quick_test"
+        assert cfg.model._target_ == "icenet_mp.models.EncodeProcessDecode"
+
+    def test_loss_defaults_resolved_from_base(self) -> None:
+        cfg = self.load_config()
+        assert cfg.loss._target_ == "torch.nn.HuberLoss"
+        assert cfg.loss.delta == pytest.approx(0.5)
+
+    def test_scalar_override_applied(self) -> None:
+        cfg = self.load_config(overrides=["loss.delta=1.0"])
+        assert cfg.loss.delta == pytest.approx(1.0)
+
+    def test_config_group_override_swaps_loss(self) -> None:
+        cfg = self.load_config(overrides=["loss=mse"])
+        assert cfg.loss._target_ == "torch.nn.MSELoss"
+
+
+class TestHydraAdaptor:
+    """Regression tests for icenet-mp's hydra_adaptor signature rewriter."""
+
+    def test_signature_rewriting(self) -> None:
+        def fn(config: DictConfig) -> None:
+            pass
+
+        params = inspect.signature(hydra_adaptor(fn)).parameters
+        assert "config" not in params
+        assert "config_name" in params
+        assert "overrides" in params
+
+    def test_preserves_positional_params(self) -> None:
+        def fn(x: int, config: DictConfig) -> None:
+            del x, config
+
+        params = list(inspect.signature(hydra_adaptor(fn)).parameters)
+        assert "x" in params
+        assert "config" not in params
+
+    def test_preserves_keyword_only_params(self) -> None:
+        def fn(config: DictConfig, *, flag: bool = False) -> None:
+            del config, flag
+
+        params = inspect.signature(hydra_adaptor(fn)).parameters
+        assert "flag" in params
+        assert params["flag"].kind == inspect.Parameter.KEYWORD_ONLY
+
+    def test_preserves_name_and_doc(self) -> None:
+        def fn(config: DictConfig) -> None:
+            """My docstring."""
+
+        assert hydra_adaptor(fn).__name__ == "fn"
+        assert hydra_adaptor(fn).__doc__ == "My docstring."
+
+    def test_wrapped_function_receives_dictconfig(self) -> None:
+        received: list[DictConfig] = []
+
+        def fn(config: DictConfig) -> None:
+            received.append(config)
+
+        hydra_adaptor(fn)(config_name="sample")  # type: ignore[arg-type]
+        assert len(received) == 1
+        assert isinstance(received[0], DictConfig)
+
+    def test_wrapped_function_forwards_kwargs(self) -> None:
+        received: list[tuple] = []
+
+        def fn(x: int, config: DictConfig) -> None:
+            received.append((x, config))
+
+        hydra_adaptor(fn)(x=42, config_name="sample")  # type: ignore[arg-type]
+        assert received[0][0] == 42
+        assert isinstance(received[0][1], DictConfig)