Fix to _get_nearest_indexer for pydata/xarray#3751

ci/deps/travis-36-cov.yaml

@@ -15,6 +15,7 @@ dependencies:
   # pandas dependencies
   - beautifulsoup4
   - botocore>=1.11
+  - cftime
   - cython>=0.29.13
   - dask
   - fastparquet>=0.3.2

pandas/core/indexes/base.py

@@ -3049,8 +3049,9 @@ def _get_nearest_indexer(self, target: "Index", limit, tolerance) -> np.ndarray:
         left_indexer = self.get_indexer(target, "pad", limit=limit)
         right_indexer = self.get_indexer(target, "backfill", limit=limit)
 
-        left_distances = np.abs(self[left_indexer] - target)
-        right_distances = np.abs(self[right_indexer] - target)
+        target = np.asarray(target)
+        left_distances = np.abs(self.values[left_indexer] - target)
+        right_distances = np.abs(self.values[right_indexer] - target)
 
         op = operator.lt if self.is_monotonic_increasing else operator.le
         indexer = np.where(

pandas/core/indexes/datetimelike.py

@@ -2,6 +2,7 @@
 Base and utility classes for tseries type pandas objects.
 """
 from datetime import datetime
+import operator
 from typing import Any, List, Optional, Union
 
 import numpy as np
@@ -957,3 +958,25 @@ def insert(self, loc, item):
             raise TypeError(
                 f"cannot insert {type(self).__name__} with incompatible label"
             ) from err
+
+    def _get_nearest_indexer(self, target: "Index", limit, tolerance) -> np.ndarray:
+        """
+        Get the indexer for the nearest index labels; requires an index with
+        values that can be subtracted from each other (e.g., not strings or
+        tuples).
+        """
+        left_indexer = self.get_indexer(target, "pad", limit=limit)
+        right_indexer = self.get_indexer(target, "backfill", limit=limit)
+
+        left_distances = np.abs(self[left_indexer] - target)
+        right_distances = np.abs(self[right_indexer] - target)
+
+        op = operator.lt if self.is_monotonic_increasing else operator.le
+        indexer = np.where(
+            op(left_distances, right_distances) | (right_indexer == -1),
+            left_indexer,
+            right_indexer,
+        )
+        if tolerance is not None:
+            indexer = self._filter_indexer_tolerance(target, indexer, tolerance)
+        return indexer

pandas/tests/test_downstream.py

@@ -47,6 +47,18 @@ def test_xarray(df):
     assert df.to_xarray() is not None
 
 
+def test_xarray_cftimeindex_nearest():
+    # https://github.com/pydata/xarray/issues/3751
+    cftime = import_module("cftime")  # noqa
+    xarray = import_module("xarray")  # noqa
+
+    times = xarray.cftime_range("0001", periods=2)
+    da = xarray.DataArray(range(2), coords=[("time", times)])
+    result = da.sel(time=cftime.DatetimeGregorian(2000, 1, 1), method="nearest")
+    expected = xarray.DataArray(1, coords={"time": cftime.DatetimeGregorian(1, 1, 2)})
+    xarray.testing.assert_identical(result, expected)
+
+
 def test_oo_optimizable():
     # GH 21071
     subprocess.check_call([sys.executable, "-OO", "-c", "import pandas"])