Pass metrics to xgcm.Grid by default.

pop_tools/xgcm_util.py

@@ -100,17 +100,39 @@ def relabel_pop_dims(ds):
         if coord in ds_new.coords:
             ds_new = ds_new.drop_vars(coord)
     if 'z_w_top' in ds_new.dims and 'z_w' in ds_new.dims:
-        ds_new = ds_new.drop('z_w_top').rename({'z_w': 'z_w_top'})
+        ds_new = ds_new.drop_vars('z_w_top').rename({'z_w': 'z_w_top'})
     return ds_new
 
 
-def to_xgcm_grid_dataset(ds, **kwargs):
+def get_metrics(ds):
+    """Finds metrics variables present in `ds`, returns a dict that can be passed to xgcm."""
+    metrics = {
+        ('X',): ['DXU', 'DXT'],  # X distances
+        ('Y',): ['DYU', 'DYT'],  # Y distances
+        ('Z',): ['DZU', 'DZT'],  # Z distances
+        ('X', 'Y'): ['UAREA', 'TAREA'],  # Areas
+    }
+    # filter to variables that are present
+    new_metrics = {}
+    for axis, names in metrics.items():
+        new_names = [name for name in names if name in ds]
+        if new_names:
+            new_metrics[axis] = new_names
+    return new_metrics
+
+
+def to_xgcm_grid_dataset(ds, metrics='detect', **kwargs):
     """Modify POP model output to be compatible with xgcm.
 
     Parameters
     ----------
     ds : xarray.Dataset
         An xarray Dataset
+    metrics : {"detect"} or dict, optional
+        Dictionary providing metrics to the `xgcm.Grid` contructor.
+        If ``"detect"``, will autodetect metrics that are present by searching for
+        variables named DXU, DXT, DYU, DYT, DZU, DZT, UAREA, TAREA.
+        If None, no metrics will be assigned.
     kwargs:
        Additional keyword arguments are passed through to `xgcm.Grid` class.
 
@@ -204,5 +226,7 @@ def to_xgcm_grid_dataset(ds, **kwargs):
             """to_xgcm_grid_dataset() function requires the `xgcm` package. \nYou can install it via PyPI or Conda"""
         )
     ds_new = relabel_pop_dims(ds)
-    grid = xgcm.Grid(ds_new, **kwargs)
+    if metrics == 'detect':
+        metrics = get_metrics(ds_new)
+    grid = xgcm.Grid(ds_new, metrics=metrics, **kwargs)
     return grid, ds_new

tests/test_xgcm_util.py

@@ -33,7 +33,7 @@
     ],
 )
 def test_to_xgcm_grid_dataset(ds, old_spatial_coords, axes):
-    grid, ds_new = pop_tools.to_xgcm_grid_dataset(ds, metrics=None)
+    grid, ds_new = pop_tools.to_xgcm_grid_dataset(ds)
     assert isinstance(grid, xgcm.Grid)
     assert set(axes) == set(grid.axes.keys())
     new_spatial_coords = ['nlon_u', 'nlat_u', 'nlon_t', 'nlat_t']
@@ -49,4 +49,52 @@ def test_to_xgcm_grid_dataset_missing_xgcm():
         with mock.patch.dict(sys.modules, {'xgcm': None}):
             filepath = DATASETS.fetch('tend_zint_100m_Fe.nc')
             ds = xr.open_dataset(filepath)
-            _, _ = pop_tools.to_xgcm_grid_dataset(ds, metrics=None)
+            _, _ = pop_tools.to_xgcm_grid_dataset(ds)
+
+
+def test_set_metrics():
+    from pop_tools.xgcm_util import get_metrics
+
+    ds = xr.Dataset({'DXU': 1, 'DYT': 2, 'DZT': 3})
+    actual = get_metrics(ds)
+    expected = {('X',): ['DXU'], ('Y',): ['DYT'], ('Z',): ['DZT']}
+    assert actual == expected
+
+    assert not get_metrics(xr.Dataset({}))
+
+
+def test_metrics_assignment_no_metrics():
+    grid, _ = pop_tools.to_xgcm_grid_dataset(ds_c)
+    assert not grid._metrics
+
+
+def get_metrics(grid):
+    return {
+        tuple(sorted(key)): [metric.name for metric in metrics]
+        for key, metrics in grid._metrics.items()
+    }
+
+
+@pytest.mark.parametrize('ds', [ds_a, ds_b])
+def test_metrics_assignment(ds):
+    grid, _ = pop_tools.to_xgcm_grid_dataset(ds)
+    expected = {
+        ('X',): ['DXU', 'DXT'],  # X distances
+        ('Y',): ['DYU', 'DYT'],  # Y distances
+        ('X', 'Y'): ['UAREA', 'TAREA'],  # Areas
+    }
+
+    if 'S_FLUX_ROFF_VSF' in ds:
+        expected[('X', 'Y')] = ['TAREA']
+        expected[('X',)] = ['DXU']
+
+    actual = get_metrics(grid)
+    assert actual == expected
+
+    grid, _ = pop_tools.to_xgcm_grid_dataset(ds, metrics=None)
+    assert not grid._metrics
+
+    expected = {('X',): ['DXU']}
+    grid, _ = pop_tools.to_xgcm_grid_dataset(ds, metrics={'X': ['DXU']})
+    actual = get_metrics(grid)
+    assert actual == expected