Improvements based on feedback

* Better testing
* Clarify comment
* Handle other functions as well, like sum, min, max

Author: dschwoerer

doc/whats-new.rst

@@ -85,10 +85,11 @@ Breaking changes
   as positional, all others need to be passed are keyword arguments. This is part of the
   refactor to support external backends (:issue:`4309`, :pull:`4989`).
   By `Alessandro Amici <https://github.com/alexamici>`_.
-- :py:func:`mean` does not change the data if axis is None. This
-  ensures that Datasets where some variables do not have the averaged
-  dimensions are not accidentially changed (:issue:`4885`).
-  By `David Schwörer <https://github.com/dschwoerer>`_
+- Functions that are identities for 0d data return the unchanged data
+  if axis is empty. This ensures that Datasets where some variables do
+  not have the averaged dimensions are not accidentially changed
+  (:issue:`4885`, :pull:`5207`).  By `David Schwörer
+  <https://github.com/dschwoerer>`_
 
 Deprecations
 ~~~~~~~~~~~~

xarray/core/duck_array_ops.py

@@ -310,13 +310,21 @@ def _ignore_warnings_if(condition):
         yield
 
 
-def _create_nan_agg_method(name, dask_module=dask_array, coerce_strings=False):
+def _create_nan_agg_method(
+    name, dask_module=dask_array, coerce_strings=False, invariant_0d=False
+):
     from . import nanops
 
     def f(values, axis=None, skipna=None, **kwargs):
         if kwargs.pop("out", None) is not None:
             raise TypeError(f"`out` is not valid for {name}")
 
+        # The data is invariant in the case of 0d data, so do not
+        # change the data (and dtype)
+        # See https://github.com/pydata/xarray/issues/4885
+        if invariant_0d and axis == ():
+            return values
+
         values = asarray(values)
 
         if coerce_strings and values.dtype.kind in "SU":
@@ -354,28 +362,30 @@ def f(values, axis=None, skipna=None, **kwargs):
 # See ops.inject_reduce_methods
 argmax = _create_nan_agg_method("argmax", coerce_strings=True)
 argmin = _create_nan_agg_method("argmin", coerce_strings=True)
-max = _create_nan_agg_method("max", coerce_strings=True)
-min = _create_nan_agg_method("min", coerce_strings=True)
-sum = _create_nan_agg_method("sum")
+max = _create_nan_agg_method("max", coerce_strings=True, invariant_0d=True)
+min = _create_nan_agg_method("min", coerce_strings=True, invariant_0d=True)
+sum = _create_nan_agg_method("sum", invariant_0d=True)
 sum.numeric_only = True
 sum.available_min_count = True
 std = _create_nan_agg_method("std")
 std.numeric_only = True
 var = _create_nan_agg_method("var")
 var.numeric_only = True
-median = _create_nan_agg_method("median", dask_module=dask_array_compat)
+median = _create_nan_agg_method(
+    "median", dask_module=dask_array_compat, invariant_0d=True
+)
 median.numeric_only = True
-prod = _create_nan_agg_method("prod")
+prod = _create_nan_agg_method("prod", invariant_0d=True)
 prod.numeric_only = True
 prod.available_min_count = True
-cumprod_1d = _create_nan_agg_method("cumprod")
+cumprod_1d = _create_nan_agg_method("cumprod", invariant_0d=True)
 cumprod_1d.numeric_only = True
-cumsum_1d = _create_nan_agg_method("cumsum")
+cumsum_1d = _create_nan_agg_method("cumsum", invariant_0d=True)
 cumsum_1d.numeric_only = True
 unravel_index = _dask_or_eager_func("unravel_index")
 
 
-_mean = _create_nan_agg_method("mean")
+_mean = _create_nan_agg_method("mean", invariant_0d=True)
 
 
 def _datetime_nanmin(array):
@@ -537,11 +547,6 @@ def mean(array, axis=None, skipna=None, **kwargs):
     dtypes"""
     from .common import _contains_cftime_datetimes
 
-    # The mean over an empty axis shouldn't change the data
-    # See https://github.com/pydata/xarray/issues/4885
-    if axis == tuple():
-        return array
-
     array = asarray(array)
     if array.dtype.kind in "Mm":
         offset = _datetime_nanmin(array)

xarray/tests/test_duck_array_ops.py

@@ -26,7 +26,7 @@
     where,
 )
 from xarray.core.pycompat import dask_array_type
-from xarray.testing import assert_allclose, assert_equal
+from xarray.testing import assert_allclose, assert_equal, assert_identical
 
 from . import (
     arm_xfail,
@@ -373,14 +373,15 @@ def test_cftime_datetime_mean_dask_error():
         da.mean()
 
 
-def test_mean_dtype():
+def test_empty_axis_dtype():
     ds = Dataset()
     ds["pos"] = [1, 2, 3]
     ds["data"] = ("pos", "time"), [[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]
     ds["var"] = "pos", [2, 3, 4]
-    ds2 = ds.mean(dim="time")
-    assert all(ds2["var"] == ds["var"])
-    assert ds2["var"].dtype == ds["var"].dtype
+    assert_identical(ds.mean(dim="time")["var"], ds["var"])
+    assert_identical(ds.max(dim="time")["var"], ds["var"])
+    assert_identical(ds.min(dim="time")["var"], ds["var"])
+    assert_identical(ds.sum(dim="time")["var"], ds["var"])
 
 
 @pytest.mark.parametrize("dim_num", [1, 2])