Fix error when plotting empty transects

Fixes #119

src/emsarray/conventions/_base.py

@@ -268,21 +268,19 @@ def bind(self) -> None:
                 "cannot assign a new convention.")
         state.bind_convention(self)
 
+    @abc.abstractmethod
     def _get_data_array(self, data_array: DataArrayOrName) -> xarray.DataArray:
         """
         Utility to help get a data array for this dataset.
         If a string is passed in, the matching data array is fetched from the dataset.
-        If a data array is passed in, it is inspected to ensure the dimensions match
+        If a data array is passed in,
+        it is inspected to ensure the surface dimensions align
         before being returned as-is.
 
         This is useful for methods that support being passed either
         the name of a data array or a data array instance.
         """
-        if isinstance(data_array, xarray.DataArray):
-            utils.check_data_array_dimensions_match(self.dataset, data_array)
-            return data_array
-        else:
-            return self.dataset[data_array]
+        pass
 
     @cached_property
     def time_coordinate(self) -> xarray.DataArray:
@@ -1011,6 +1009,9 @@ def animate_on_figure(
         if coordinate is None:
             # Assume the user wants to plot along the time axis by default.
             coordinate = self.get_time_name()
+        if isinstance(coordinate, xarray.DataArray):
+            utils.check_data_array_dimensions_match(
+                self.dataset, coordinate, coordinate.dims)
 
         coordinate = self._get_data_array(coordinate)
 
@@ -1787,6 +1788,23 @@ def get_grid_kind(self, data_array: xarray.DataArray) -> GridKind:
                 return kind
         raise ValueError("Unknown grid kind")
 
+    def _get_data_array(self, data_array: DataArrayOrName) -> xarray.DataArray:
+        if isinstance(data_array, xarray.DataArray):
+            grid_kind = self.get_grid_kind(data_array)
+            grid_dimensions = self.grid_dimensions[grid_kind]
+            for dimension in grid_dimensions:
+                # The data array already has matching dimension names
+                # as we found the grid kind using `Convention.get_grid_kind()`.
+                if self.dataset.sizes[dimension] != data_array.sizes[dimension]:
+                    raise ValueError(
+                        f"Mismatched dimension {dimension!r}, "
+                        "dataset has size {self.dataset.sizes[dimension]} but "
+                        "data array has size {data_array.sizes[dimension]}!"
+                    )
+            return data_array
+        else:
+            return self.dataset[data_array]
+
     @abc.abstractmethod
     def unpack_index(self, index: Index) -> Tuple[GridKind, Sequence[int]]:
         """Convert a native index in to a grid kind and dimension indices.

src/emsarray/transect.py

@@ -188,10 +188,17 @@ def transect_dataset(self) -> xarray.Dataset:
             dims=(depth_dimension, 'bounds'),
         )
         distance_bounds = xarray.DataArray(
-            data=[
-                [segment.start_distance, segment.end_distance]
-                for segment in self.segments
-            ],
+            data=numpy.fromiter(
+                (
+                    [segment.start_distance, segment.end_distance]
+                    for segment in self.segments
+                ),
+                # Be explicit here, to handle the case when len(self.segments) == 0.
+                # This happens when the transect line does not intersect the dataset.
+                # This will result in an empty transect plot.
+                count=len(self.segments),
+                dtype=numpy.dtype((float, 2)),
+            ),
             dims=('index', 'bounds'),
             attrs={
                 'long_name': 'Distance along transect',
@@ -762,8 +769,15 @@ def _plot_on_figure(
 
         cmap = colormaps[cmap].copy()
         cmap.set_bad(ocean_floor_colour)
-        collection = self.make_poly_collection(
-            cmap=cmap, clim=(numpy.nanmin(data_array), numpy.nanmax(data_array)))
+
+        if data_array.size != 0:
+            clim = (numpy.nanmin(data_array), numpy.nanmax(data_array))
+        else:
+            # An empty data array happens when the transect line does not
+            # intersect the dataset geometry.
+            clim = None
+
+        collection = self.make_poly_collection(cmap=cmap, clim=clim, edgecolor='face')
         axes.add_collection(collection)
 
         if bathymetry is not None:

src/emsarray/utils.py

@@ -427,7 +427,12 @@ def dimensions_from_coords(
     return dimensions
 
 
-def check_data_array_dimensions_match(dataset: xarray.Dataset, data_array: xarray.DataArray) -> None:
+def check_data_array_dimensions_match(
+    dataset: xarray.Dataset,
+    data_array: xarray.DataArray,
+    *,
+    dimensions: Optional[Sequence[Hashable]] = None,
+) -> None:
     """
     Check that the dimensions of a :class:`xarray.DataArray`
     match the dimensions of a :class:`xarray.Dataset`.
@@ -438,27 +443,38 @@ def check_data_array_dimensions_match(dataset: xarray.Dataset, data_array: xarra
 
     Parameters
     ----------
-    dataset
+    dataset : xarray.Dataset
         The dataset used as a reference
-    data_array
+    data_array : xarray.DataArray
         The data array to check the dimensions of
+    dimensions: list of Hashable, optional
+        The dimension names to check for equal sizes.
+        Optional, defaults to checking all dimensions on the data array.
 
     Raises
     ------
     ValueError
         Raised if the dimensions do not match
     """
-    for dimension, data_array_size in zip(data_array.dims, data_array.shape):
-        if dimension not in dataset.dims:
+    if dimensions is None:
+        dimensions = data_array.dims
+
+    for dimension in dimensions:
+        if dimension not in dataset.dims and dimension not in data_array.dims:
             raise ValueError(
-                f"Data array has unknown dimension {dimension} of size {data_array_size}"
+                f"Dimension {dimension!r} not present on either dataset or data array"
             )
-
+        elif dimension not in dataset.dims:
+            raise ValueError(f"Dataset does not have dimension {dimension!r}")
+        elif dimension not in data_array.dims:
+            raise ValueError(f"Data array does not have dimension {dimension!r}")
         dataset_size = dataset.sizes[dimension]
+        data_array_size = data_array.sizes[dimension]
+
         if data_array_size != dataset_size:
             raise ValueError(
                 "Dimension mismatch between dataset and data array: "
-                f"Dataset dimension {dimension} has size {dataset_size}, "
+                f"Dataset dimension {dimension!r} has size {dataset_size}, "
                 f"data array has size {data_array_size}"
             )
 

tests/conventions/test_base.py

@@ -44,6 +44,12 @@ class SimpleConvention(Convention[SimpleGridKind, SimpleGridIndex]):
     def check_dataset(cls, dataset: xarray.Dataset) -> Optional[int]:
         return None
 
+    def _get_data_array(self, data_array_or_name) -> xarray.DataArray:
+        if isinstance(data_array_or_name, str):
+            return self.dataset[data_array_or_name]
+        else:
+            return data_array_or_name
+
     @cached_property
     def shape(self) -> Tuple[int, int]:
         y, x = map(int, self.dataset['botz'].shape)

tests/test_transect.py

@@ -55,3 +55,45 @@ def test_plot(
 
     matplotlib.pyplot.savefig(tmp_path / 'plot.png')
     logger.info("Saved plot to %r", tmp_path / 'plot.png')
+
+
+@pytest.mark.matplotlib(mock_coast=True)
+@pytest.mark.tutorial
+def test_plot_no_intersection(
+    datasets: pathlib.Path,
+    tmp_path: pathlib.Path,
+):
+    """
+    Transects that do not intersect the dataset geometry need special handling.
+    This should produce an empty transect plot, which is better than raising an error.
+    """
+    dataset = emsarray.tutorial.open_dataset('gbr4')
+    temp = dataset['temp'].copy()
+    temp = temp.isel(time=-1)
+
+    # This line goes through the Bass Strait, no where near the GBR.
+    # Someone picked the wrong dataset...
+    line = shapely.LineString([
+        [142.097168, -39.206719],
+        [145.393066, -39.3088],
+        [149.798584, -39.172659],
+    ])
+    emsarray.transect.plot(
+        dataset, line, temp,
+        bathymetry=dataset['botz'])
+
+    figure = matplotlib.pyplot.gcf()
+    axes = figure.axes[0]
+    # This is assembled from the variable long_name and the time coordinate
+    assert axes.get_title() == 'Temperature\n2022-05-11T14:00'
+    # This is the long_name of the depth coordinate
+    assert axes.get_ylabel() == 'Z coordinate'
+    # This is made up
+    assert axes.get_xlabel() == 'Distance along transect'
+
+    colorbar = figure.axes[-1]
+    # This is the variable units
+    assert colorbar.get_ylabel() == 'degrees C'
+
+    matplotlib.pyplot.savefig(tmp_path / 'plot.png')
+    logger.info("Saved plot to %r", tmp_path / 'plot.png')