Removed raise ValueError from previous commit

It seems that np.min/max works in place of nanmin/nanmax for datetime types for numpy < 1.18, see https://github.com/pydata/xarray/pull/3924/files

doc/whats-new.rst

@@ -76,8 +76,7 @@ Bug fixes
 - Fix :py:func:`xarray.apply_ufunc` with ``vectorize=True`` and ``exclude_dims`` (:issue:`3890`).
   By `Mathias Hauser <https://github.com/mathause>`_.
 - Fix `KeyError` when doing linear interpolation to an nd `DataArray`
-  that contains NaNs (:pull:`4233`). Requires `numpy` version `1.18`
-  or newer to work with datetime indices.
+  that contains NaNs (:pull:`4233`).
   By `Jens Svensmark <https://github.com/jenssss>`_
 
 Documentation

xarray/core/missing.py

@@ -554,13 +554,10 @@ def _localize(var, indexes_coords):
         if np.issubdtype(new_x.dtype, np.datetime64) and LooseVersion(
             np.__version__
         ) < LooseVersion("1.18"):
-            if new_x.isnull().any():
-                raise ValueError(
-                    "numpy 1.18 or newer required to use interp with datetime/ timedelta array containing missing values"
-                )
-            else:
-                minval = np.min(new_x.values)
-                maxval = np.max(new_x.values)
+            # np.nanmin/max changed behaviour for datetime types in numpy 1.18,
+            # see https://github.com/pydata/xarray/pull/3924/files
+            minval = np.min(new_x.values)
+            maxval = np.max(new_x.values)
         else:
             minval = np.nanmin(new_x.values)
             maxval = np.nanmax(new_x.values)