BUG: invalid rolling window on empty input (#21291)

doc/source/whatsnew/v0.23.1.txt

@@ -53,6 +53,7 @@ Groupby/Resample/Rolling
 - Bug in :func:`DataFrame.agg` where applying multiple aggregation functions to a :class:`DataFrame` with duplicated column names would cause a stack overflow (:issue:`21063`)
 - Bug in :func:`pandas.core.groupby.GroupBy.ffill` and :func:`pandas.core.groupby.GroupBy.bfill` where the fill within a grouping would not always be applied as intended due to the implementations' use of a non-stable sort (:issue:`21207`)
 - Bug in :func:`pandas.core.groupby.GroupBy.rank` where results did not scale to 100% when specifying ``method='dense'`` and ``pct=True``
+- Bug in :func:`pandas.DataFrame.rolling` and :func:`pandas.Series.rolling` which incorrectly accepted a 0 window size rather than raising (:issue:`21286`)
 
 Data-type specific
 ~~~~~~~~~~~~~~~~~~

pandas/core/window.py

@@ -602,8 +602,8 @@ def validate(self):
         if isinstance(window, (list, tuple, np.ndarray)):
             pass
         elif is_integer(window):
-            if window < 0:
-                raise ValueError("window must be non-negative")
+            if window <= 0:
+                raise ValueError("window must be > 0 ")
             try:
                 import scipy.signal as sig
             except ImportError:

pandas/tests/test_window.py

@@ -389,8 +389,8 @@ def test_constructor(self, which):
         c(window=2, min_periods=1, center=False)
 
         # GH 13383
-        c(0)
         with pytest.raises(ValueError):
+            c(0)
             c(-1)
 
         # not valid
@@ -409,7 +409,6 @@ def test_constructor_with_win_type(self, which):
         # GH 13383
         o = getattr(self, which)
         c = o.rolling
-        c(0, win_type='boxcar')
         with pytest.raises(ValueError):
             c(-1, win_type='boxcar')