BUG: support min/max functions for rolling windows with custom BaseIndexer

doc/source/whatsnew/v1.1.0.rst

@@ -101,8 +101,8 @@ Other API changes
 - Added :meth:`DataFrame.value_counts` (:issue:`5377`)
 - :meth:`Groupby.groups` now returns an abbreviated representation when called on large dataframes (:issue:`1135`)
 - ``loc`` lookups with an object-dtype :class:`Index` and an integer key will now raise ``KeyError`` instead of ``TypeError`` when key is missing (:issue:`31905`)
-- Using a :func:`pandas.api.indexers.BaseIndexer` with ``min``, ``max``, ``std``, ``var``, ``count``, ``skew``, ``cov``, ``corr`` will now raise a ``NotImplementedError`` (:issue:`32865`)
--
+- Using a :func:`pandas.api.indexers.BaseIndexer` with ``std``, ``var``, ``count``, ``skew``, ``cov``, ``corr`` will now raise a ``NotImplementedError`` (:issue:`32865`)
+- Using a :func:`pandas.api.indexers.BaseIndexer` with ``min``, ``max`` will now return correct results for any monotonic :func:`pandas.api.indexers.BaseIndexer` descendant (:issue:`32865`)
 
 Backwards incompatible API changes
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

pandas/_libs/window/aggregations.pyx

@@ -1051,7 +1051,7 @@ cdef _roll_min_max_variable(ndarray[numeric] values,
                             bint is_max):
     cdef:
         numeric ai
-        int64_t i, close_offset, curr_win_size
+        int64_t i, k, curr_win_size, start
         Py_ssize_t nobs = 0, N = len(values)
         deque Q[int64_t]  # min/max always the front
         deque W[int64_t]  # track the whole window for nobs compute
@@ -1068,60 +1068,45 @@ cdef _roll_min_max_variable(ndarray[numeric] values,
         # The original impl didn't deal with variable window sizes
         # So the code was optimized for that
 
-        for i in range(starti[0], endi[0]):
-            ai = init_mm(values[i], &nobs, is_max)
-
-            # Discard previous entries if we find new min or max
-            if is_max:
-                while not Q.empty() and ((ai >= values[Q.back()]) or
-                                         values[Q.back()] != values[Q.back()]):
-                    Q.pop_back()
-            else:
-                while not Q.empty() and ((ai <= values[Q.back()]) or
-                                         values[Q.back()] != values[Q.back()]):
-                    Q.pop_back()
-            Q.push_back(i)
-            W.push_back(i)
-
-        # if right is open then the first window is empty
-        close_offset = 0 if endi[0] > starti[0] else 1
         # first window's size
         curr_win_size = endi[0] - starti[0]
+        # GH 32865
+        # Anchor output index to values index to provide custom
+        # BaseIndexer support
+        for i in range(N):
 
-        for i in range(endi[0], endi[N-1]):
-            if not Q.empty() and curr_win_size > 0:
-                output[i-1+close_offset] = calc_mm(
-                    minp, nobs, values[Q.front()])
-            else:
-                output[i-1+close_offset] = NaN
-
-            ai = init_mm(values[i], &nobs, is_max)
-
-            # Discard previous entries if we find new min or max
-            if is_max:
-                while not Q.empty() and ((ai >= values[Q.back()]) or
-                                         values[Q.back()] != values[Q.back()]):
-                    Q.pop_back()
+            curr_win_size = endi[i] - starti[i]
+            if i == 0:
+                start = starti[i]
             else:
-                while not Q.empty() and ((ai <= values[Q.back()]) or
-                                         values[Q.back()] != values[Q.back()]):
-                    Q.pop_back()
+                start = endi[i - 1]
 
-            # Maintain window/nobs retention
-            curr_win_size = endi[i + close_offset] - starti[i + close_offset]
-            while not Q.empty() and Q.front() <= i - curr_win_size:
+            for k in range(start, endi[i]):
+                ai = init_mm(values[k], &nobs, is_max)
+                # Discard previous entries if we find new min or max
+                if is_max:
+                    while not Q.empty() and ((ai >= values[Q.back()]) or
+                                             values[Q.back()] != values[Q.back()]):
+                        Q.pop_back()
+                else:
+                    while not Q.empty() and ((ai <= values[Q.back()]) or
+                                             values[Q.back()] != values[Q.back()]):
+                        Q.pop_back()
+                Q.push_back(k)
+                W.push_back(k)
+
+            # Discard entries outside and left of current window
+            while not Q.empty() and Q.front() <= starti[i] - 1:
                 Q.pop_front()
-            while not W.empty() and W.front() <= i - curr_win_size:
+            while not W.empty() and W.front() <= starti[i] - 1:
                 remove_mm(values[W.front()], &nobs)
                 W.pop_front()
 
-            Q.push_back(i)
-            W.push_back(i)
-
-        if not Q.empty() and curr_win_size > 0:
-            output[N-1] = calc_mm(minp, nobs, values[Q.front()])
-        else:
-            output[N-1] = NaN
+            # Save output based on index in input value array
+            if not Q.empty() and curr_win_size > 0:
+                output[i] = calc_mm(minp, nobs, values[Q.front()])
+            else:
+                output[i] = NaN
 
     return output
 

pandas/core/window/common.py

@@ -327,7 +327,7 @@ def func(arg, window, min_periods=None):
 
 def validate_baseindexer_support(func_name: Optional[str]) -> None:
     # GH 32865: These functions work correctly with a BaseIndexer subclass
-    BASEINDEXER_WHITELIST = {"mean", "sum", "median", "kurt", "quantile"}
+    BASEINDEXER_WHITELIST = {"min", "max", "mean", "sum", "median", "kurt", "quantile"}
     if isinstance(func_name, str) and func_name not in BASEINDEXER_WHITELIST:
         raise NotImplementedError(
             f"{func_name} is not supported with using a BaseIndexer "

pandas/tests/window/test_base_indexer.py

@@ -82,9 +82,7 @@ def get_window_bounds(self, num_values, min_periods, center, closed):
         df.rolling(indexer, win_type="boxcar")
 
 
-@pytest.mark.parametrize(
-    "func", ["min", "max", "std", "var", "count", "skew", "cov", "corr"]
-)
+@pytest.mark.parametrize("func", ["std", "var", "count", "skew", "cov", "corr"])
 def test_notimplemented_functions(func):
     # GH 32865
     class CustomIndexer(BaseIndexer):

pandas/tests/window/test_rolling.py

@@ -9,6 +9,7 @@
 import pandas as pd
 from pandas import DataFrame, Index, Series
 import pandas._testing as tm
+from pandas.api.indexers import BaseIndexer
 from pandas.core.window import Rolling
 from pandas.tests.window.common import Base
 
@@ -465,3 +466,35 @@ def test_rolling_count_default_min_periods_with_null_values(constructor):
     result = constructor(values).rolling(3).count()
     expected = constructor(expected_counts)
     tm.assert_equal(result, expected)
+
+
+@pytest.mark.parametrize("constructor", [Series, DataFrame])
+@pytest.mark.parametrize(
+    "func,expected",
+    [
+        ("min", [0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, np.nan]),
+        ("max", [2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 9.0, np.nan]),
+    ],
+)
+def test_rolling_forward_window_min(constructor, func, expected):
+    # GH 32865
+    class ForwardIndexer(BaseIndexer):
+        def get_window_bounds(self, num_values, min_periods, center, closed):
+            start = np.empty(num_values, dtype=np.int64)
+            end = np.empty(num_values, dtype=np.int64)
+            for i in range(num_values):
+                if i + min_periods <= num_values:
+                    start[i] = i
+                    end[i] = min(i + self.window_size, num_values)
+                else:
+                    start[i] = i
+                    end[i] = i + 1
+            return start, end
+
+    values = np.arange(10)
+
+    indexer = ForwardIndexer(window_size=3)
+    rolling = constructor(values).rolling(window=indexer, min_periods=2)
+    result = getattr(rolling, func)()
+    expected = constructor(expected)
+    tm.assert_equal(result, expected)