Add a test for float precision issues

Author: snitish

pandas/_libs/window/aggregations.pyx

@@ -1478,7 +1478,7 @@ def roll_nunique(const float64_t[:] values, ndarray[int64_t] start,
     """
     cdef:
         Py_ssize_t i, j, s, e, N = len(start)
-        int64_t nobs = 0, num_unique = 0
+        int64_t nobs = 0
         float64_t val
         float64_t[::1] output
         unordered_map[float64_t, int64_t] value_counts
@@ -1506,8 +1506,6 @@ def roll_nunique(const float64_t[:] values, ndarray[int64_t] start,
                         nobs += 1
                         value_counts[val] += 1
 
-                num_unique = value_counts.size()
-
             else:
                 # calculate deletes
                 for j in range(start[i - 1], s):
@@ -1525,9 +1523,8 @@ def roll_nunique(const float64_t[:] values, ndarray[int64_t] start,
                         nobs += 1
                         value_counts[val] += 1
 
-                num_unique = value_counts.size()
             if nobs >= minp:
-                output[i] = num_unique
+                output[i] = value_counts.size()
             else:
                 output[i] = NaN
 

pandas/tests/window/test_expanding.py

@@ -256,7 +256,7 @@ def test_rank(window, method, pct, ascending, test_data):
 
 
 @pytest.mark.parametrize("window", [1, 3, 10, 20])
-@pytest.mark.parametrize("test_data", ["default", "duplicates", "nans"])
+@pytest.mark.parametrize("test_data", ["default", "duplicates", "nans", "precision"])
 def test_nunique(window, test_data):
     length = 20
     if test_data == "default":
@@ -269,6 +269,22 @@ def test_nunique(window, test_data):
                 [1.0, 0.25, 0.75, np.nan, np.inf, -np.inf], length
             )
         )
+    elif test_data == "precision":
+        ser = Series(
+            data=[
+                0.3,
+                0.1 * 3,  # Not necessarily exactly 0.3
+                0.6,
+                0.2 * 3,  # Not necessarily exactly 0.6
+                0.9,
+                0.3 * 3,  # Not necessarily exactly 0.9
+                0.5,
+                0.1 * 5,  # Not necessarily exactly 0.5
+                0.8,
+                0.2 * 4,  # Not necessarily exactly 0.8
+            ],
+            dtype=np.float64,
+        )
 
     expected = ser.expanding(window).apply(lambda x: x.nunique())
     result = ser.expanding(window).nunique()

pandas/tests/window/test_rolling.py

@@ -1587,7 +1587,7 @@ def test_rank(window, method, pct, ascending, test_data):
 
 
 @pytest.mark.parametrize("window", [1, 3, 10, 20])
-@pytest.mark.parametrize("test_data", ["default", "duplicates", "nans"])
+@pytest.mark.parametrize("test_data", ["default", "duplicates", "nans", "precision"])
 def test_nunique(window, test_data):
     length = 20
     if test_data == "default":
@@ -1600,6 +1600,22 @@ def test_nunique(window, test_data):
                 [1.0, 0.25, 0.75, np.nan, np.inf, -np.inf], length
             )
         )
+    elif test_data == "precision":
+        ser = Series(
+            data=[
+                0.3,
+                0.1 * 3,  # Not necessarily exactly 0.3
+                0.6,
+                0.2 * 3,  # Not necessarily exactly 0.6
+                0.9,
+                0.3 * 3,  # Not necessarily exactly 0.9
+                0.5,
+                0.1 * 5,  # Not necessarily exactly 0.5
+                0.8,
+                0.2 * 4,  # Not necessarily exactly 0.8
+            ],
+            dtype=np.float64,
+        )
 
     expected = ser.rolling(window).apply(lambda x: x.nunique())
     result = ser.rolling(window).nunique()