Optimize broadcasting (#230)

* Optimize broadcasting

xref pydata/xarray#7730

* reorder

* Fix tests

* Another optimization

* fixes

* fix

flox/xarray.py

@@ -257,8 +257,6 @@ def xarray_reduce(
             more_drop.update(idx_other_names)
     maybe_drop.update(more_drop)
 
-    ds = ds.drop_vars([var for var in maybe_drop if var in ds.variables])
-
     if dim is Ellipsis:
         if nby > 1:
             raise NotImplementedError("Multiple by are not allowed when dim is Ellipsis.")
@@ -275,17 +273,23 @@ def xarray_reduce(
     # broadcast to make sure grouper dimensions are present in the array.
     exclude_dims = tuple(d for d in ds.dims if d not in grouper_dims and d not in dim_tuple)
 
+    if any(d not in grouper_dims and d not in obj.dims for d in dim_tuple):
+        raise ValueError(f"Cannot reduce over absent dimensions {dim}.")
+
     try:
         xr.align(ds, *by_da, join="exact", copy=False)
     except ValueError as e:
         raise ValueError(
             "Object being grouped must be exactly aligned with every array in `by`."
         ) from e
 
-    ds_broad = xr.broadcast(ds, *by_da, exclude=exclude_dims)[0]
-
-    if any(d not in grouper_dims and d not in obj.dims for d in dim_tuple):
-        raise ValueError(f"Cannot reduce over absent dimensions {dim}.")
+    needs_broadcast = any(
+        not set(grouper_dims).issubset(set(variable.dims)) for variable in ds.data_vars.values()
+    )
+    if needs_broadcast:
+        ds_broad = xr.broadcast(ds, *by_da, exclude=exclude_dims)[0]
+    else:
+        ds_broad = ds
 
     dims_not_in_groupers = tuple(d for d in dim_tuple if d not in grouper_dims)
     if dims_not_in_groupers == tuple(dim_tuple) and not any(isbins):
@@ -305,6 +309,8 @@ def xarray_reduce(
         else:
             return result
 
+    ds = ds.drop_vars([var for var in maybe_drop if var in ds.variables])
+
     axis = tuple(range(-len(dim_tuple), 0))
 
     # Set expected_groups and convert to index since we need coords, sizes
@@ -432,7 +438,7 @@ def wrapper(array, *by, func, skipna, core_dims, **kwargs):
 
     # restore non-dim coord variables without the core dimension
     # TODO: shouldn't apply_ufunc handle this?
-    for var in set(ds_broad.variables) - set(ds_broad._indexes) - set(ds_broad.dims):
+    for var in set(ds_broad._coord_names) - set(ds_broad._indexes) - set(ds_broad.dims):
         if all(d not in ds_broad[var].dims for d in dim_tuple):
             actual[var] = ds_broad[var]
 

tests/test_xarray.py

@@ -343,6 +343,8 @@ def test_multi_index_groupby_sum(engine):
     expected = ds.sum("z")
     stacked = ds.stack(space=["x", "y"])
     actual = xarray_reduce(stacked, "space", dim="z", func="sum", engine=engine)
+    expected_xarray = stacked.groupby("space").sum("z")
+    assert_equal(expected_xarray, actual)
     assert_equal(expected, actual.unstack("space"))
 
     actual = xarray_reduce(stacked.foo, "space", dim="z", func="sum", engine=engine)