BUG: Ignore division by 0 when merging empty dataframes (#17776) (#17846)

Author: yeemey

doc/source/whatsnew/v0.21.1.txt

@@ -120,7 +120,7 @@ Reshaping
 - Error message in ``pd.merge_asof()`` for key datatype mismatch now includes datatype of left and right key (:issue:`18068`)
 - Bug in ``pd.concat`` when empty and non-empty DataFrames or Series are concatenated (:issue:`18178` :issue:`18187`)
 - Bug in ``DataFrame.filter(...)`` when :class:`unicode` is passed as a condition in Python 2 (:issue:`13101`)
--
+- Bug when merging empty DataFrames when ``np.seterr(divide='raise')`` is set (:issue:`17776`)
 
 Numeric
 ^^^^^^^

pandas/core/reshape/merge.py

@@ -1529,7 +1529,8 @@ def _get_join_keys(llab, rlab, shape, sort):
     rkey = stride * rlab[0].astype('i8', subok=False, copy=False)
 
     for i in range(1, nlev):
-        stride //= shape[i]
+        with np.errstate(divide='ignore'):
+            stride //= shape[i]
         lkey += llab[i] * stride
         rkey += rlab[i] * stride
 

pandas/tests/reshape/test_merge.py

@@ -864,6 +864,12 @@ def test_validation(self):
         result = merge(left, right, on=['a', 'b'], validate='1:1')
         assert_frame_equal(result, expected_multi)
 
+    def test_merge_two_empty_df_no_division_error(self):
+        # GH17776, PR #17846
+        a = pd.DataFrame({'a': [], 'b': [], 'c': []})
+        with np.errstate(divide='raise'):
+            merge(a, a, on=('a', 'b'))
+
 
 def _check_merge(x, y):
     for how in ['inner', 'left', 'outer']: