Improve performance of dataset indexing (pydata#11276)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Deepak Cherian <dcherian@users.noreply.github.com>

Author: 3 people

doc/whats-new.rst

@@ -102,7 +102,25 @@ Breaking Changes
 
 Deprecations
 ~~~~~~~~~~~~
-
+- Following pandas, on xarray's
+  :py:class:`~xarray.core.accessor_dt.DatetimeAccessor`,
+  :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.daysinmonth`
+  is deprecated in favor of
+  :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.days_in_month`;
+  :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.dayofweek` and
+  :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.weekday` are deprecated
+  in favor of :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.day_of_week`;
+  and :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.dayofyear` is
+  deprecated in favor of
+  :py:attr:`~xarray.core.accessor_dt.DatetimeAccessor.day_of_year`
+  (:issue:`11268`, :pull:`11270`). By `Spencer Clark
+  <https://github.com/spencerkclark>`_.
+- Following pandas, on xarray's :py:class:`~xarray.CFTimeIndex`,
+  :py:attr:`~xarray.CFTimeIndex.dayofweek` and
+  :py:attr:`~xarray.CFTimeIndex.dayofyear` are deprecated in favor of
+  :py:attr:`~xarray.CFTimeIndex.day_of_week` and
+  :py:attr:`~xarray.CFTimeIndex.day_of_year`, respectively (:pull:`11270`).
+  By `Spencer Clark <https://github.com/spencerkclark>`_.
 
 Bug Fixes
 ~~~~~~~~~

xarray/core/dataset_utils.py

@@ -67,8 +67,8 @@ def _get_virtual_variable(
         dim_sizes = {}
 
     if key in dim_sizes:
-        data = pd.Index(range(dim_sizes[key]), name=key)
-        variable = IndexVariable((key,), data)
+        data = pd.RangeIndex(dim_sizes[key], name=key)
+        variable = IndexVariable((key,), data, fastpath=True)
         return key, key, variable
 
     if not isinstance(key, str):

xarray/core/indexes.py

@@ -807,7 +807,9 @@ def create_variables(
             encoding = None
 
         data = PandasIndexingAdapter(self.index, dtype=self.coord_dtype)
-        var = IndexVariable(self.dim, data, attrs=attrs, encoding=encoding)
+        var = IndexVariable(
+            self.dim, data, attrs=attrs, encoding=encoding, fastpath=True
+        )
         return {name: var}
 
     def to_pandas_index(self) -> pd.Index:
@@ -829,6 +831,9 @@ def isel(
             # scalar indexer: drop index
             return None
 
+        if isinstance(indxr, slice) and indxr == slice(None):
+            return self
+
         return self._replace(self.index[indxr])  # type: ignore[index,unused-ignore]
 
     def sel(
@@ -2164,7 +2169,10 @@ def _apply_indexes_fast(indexes: Indexes[Index], args: Mapping[Any, Any], func:
             new_index = getattr(index, func)(index_args)
             if new_index is not None:
                 new_indexes.update(dict.fromkeys(index_vars, new_index))
-                new_index_vars = new_index.create_variables(index_vars)
+                if new_index is index:
+                    new_index_vars = index_vars
+                else:
+                    new_index_vars = new_index.create_variables(index_vars)
                 new_index_variables.update(new_index_vars)
             else:
                 for k in index_vars:

xarray/core/variable.py

@@ -634,6 +634,18 @@ def _broadcast_indexes(self, key):
             positions.
         """
         key = self._item_key_to_tuple(key)  # key is a tuple
+        # Fast path: key is already a tuple of the right length with only
+        # ints and slices (the common case from Variable.isel)
+        if (
+            isinstance(key, tuple)
+            and len(key) == self.ndim
+            and all(
+                not isinstance(k, bool) and isinstance(k, BASIC_INDEXING_TYPES)
+                for k in key
+            )
+        ):
+            return self._broadcast_indexes_basic(key)
+
         # key is a tuple of full size
         key = indexing.expanded_indexer(key, self.ndim)
         # Convert a scalar Variable to a 0d-array