feat: Forecast wrapper for custom xarray datasets

src/extremeweatherbench/inputs.py

@@ -423,6 +423,55 @@ def _open_data_from_source(self) -> IncomingDataInput:
         )
 
 
+@dataclasses.dataclass
+class XarrayForecast(ForecastBase):
+    """Forecast class for datasets that were previously constructed and opened using xarray.
+
+    This class is intended for situations where the user has to manually prepare a dataset to
+    use in their evaluation. This can happen when the user is manually constructed such a
+    dataset from a collection of NetCDF or Zarr archives which need to be assembled into a
+    single, master dataset.
+
+    Attributes:
+        ds: The xarray dataset containing the forecast data.
+        source: The source of the data, defaults to "memory" for in-memory datasets.
+        name: The name of the input data source, defaults to "in-memory dataset".
+    """
+
+    #: The xarray dataset containing the forecast data. This is required for the class to be instantiated
+    #: because we inherit from ForecastBase, which has its own set of required attributes.
+    ds: Optional[xr.Dataset] = None  # type: ignore[assignment]
+    source: str = "memory"
+    name: str = "in-memory dataset"
+
+    def __post_init__(self):
+        """Validate that ds is provided and normalize None values to defaults.
+
+        This ensures backwards compatibility with the ForecastBase's __init__ behavior
+        where None values for variables and variable_mapping were converted to empty
+        containers. If the user does not provide a ds, we raise a ValueError.
+        """
+        if self.ds is None:
+            raise ValueError(
+                "The 'ds' parameter is required for XarrayForecast. "
+                "Please provide an xarray.Dataset."
+            )
+
+        # Convert None to empty containers for backwards compatibility
+        if self.variables is None:
+            object.__setattr__(self, "variables", [])
+        if self.variable_mapping is None:
+            object.__setattr__(self, "variable_mapping", {})
+
+    def _open_data_from_source(self) -> xr.Dataset:
+        """Open the input data from the source.
+
+        Returns:
+            The xarray dataset that was provided during initialization.
+        """
+        return self.ds
+
+
 @dataclasses.dataclass
 class TargetBase(InputBase):
     """An abstract base class for target data.