BUG: NaN should have pct rank of NaN

Closes gh-22519.

Author: gfyoung

doc/source/whatsnew/v0.23.5.txt

@@ -23,6 +23,9 @@ Fixed Regressions
 - Constructing a DataFrame with an index argument that wasn't already an
  instance of :class:`~pandas.core.Index` was broken in `4efb39f
  <https://github.com/pandas-dev/pandas/commit/4efb39f01f5880122fa38d91e12d217ef70fad9e>`_ (:issue:`22227`).
+- Calling :meth:`DataFrameGroupBy.rank` and :meth:`SeriesGroupBy.rank` with empty groups
+ and ``pct=True`` was raising a ``ZeroDivisionError`` in `c1068d9
+ <https://github.com/pandas-dev/pandas/commit/c1068d9d242c22cb2199156f6fb82eb5759178ae>`_ (:issue:`22519`)
 -
 -
 

pandas/_libs/groupby_helper.pxi.in

@@ -584,7 +584,10 @@ def group_rank_{{name}}(ndarray[float64_t, ndim=2] out,
 
  if pct:
  for i in range(N):
- out[i, 0] = out[i, 0] / grp_sizes[i, 0]
+ if out[i, 0] != out[i, 0] or out[i, 0] == NAN:
+ out[i, 0] = NAN
+ else:
+ out[i, 0] = out[i, 0] / grp_sizes[i, 0]
 {{endif}}
 {{endfor}}
 

pandas/tests/groupby/test_rank.py

@@ -1,7 +1,7 @@
 import pytest
 import numpy as np
 import pandas as pd
-from pandas import DataFrame, concat
+from pandas import DataFrame, Series, concat
 from pandas.util import testing as tm
 
 
@@ -273,3 +273,20 @@ def test_rank_naoption_raises(ties_method, ascending, na_option, pct, vals):
  df.groupby('key').rank(method=ties_method,
  ascending=ascending,
  na_option=na_option, pct=pct)
+
+
+def test_rank_empty_group():
+ # see gh-22519
+ column = "A"
+ df = DataFrame({
+ "A": [0, 1, 0],
+ "B": [1., np.nan, 2.]
+ })
+
+ result = df.groupby(column).B.rank(pct=True)
+ expected = Series([0.5, np.nan, 1.0], name="B")
+ tm.assert_series_equal(result, expected)
+
+ result = df.groupby(column).rank(pct=True)
+ expected = DataFrame({"B": [0.5, np.nan, 1.0]})
+ tm.assert_frame_equal(result, expected)