PERF: PeriodArray._format_native_types (#51793)

Author: jbrockmendel

pandas/_libs/tslibs/period.pyi

@@ -42,6 +42,12 @@ def extract_ordinals(
 def extract_freq(
     values: npt.NDArray[np.object_],
 ) -> BaseOffset: ...
+def period_array_strftime(
+    values: npt.NDArray[np.int64],
+    dtype_code: int,
+    na_rep,
+    date_format: str | None,
+) -> npt.NDArray[np.object_]: ...
 
 # exposed for tests
 def period_asfreq(ordinal: int, freq1: int, freq2: int, end: bool) -> int: ...

pandas/_libs/tslibs/period.pyx

@@ -1268,6 +1268,54 @@ cdef str _period_strftime(int64_t value, int freq, bytes fmt):
     return result
 
 
+def period_array_strftime(
+    ndarray values, int dtype_code, object na_rep, str date_format
+):
+    """
+    Vectorized Period.strftime used for PeriodArray._format_native_types.
+
+    Parameters
+    ----------
+    values : ndarray[int64_t], ndim unrestricted
+    dtype_code : int
+        Corresponds to PeriodDtype._dtype_code
+    na_rep : any
+    date_format : str or None
+    """
+    cdef:
+        Py_ssize_t i, n = values.size
+        int64_t ordinal
+        object item_repr
+        ndarray out = cnp.PyArray_EMPTY(
+            values.ndim, values.shape, cnp.NPY_OBJECT, 0
+        )
+        object[::1] out_flat = out.ravel()
+        cnp.broadcast mi = cnp.PyArray_MultiIterNew2(out, values)
+
+    for i in range(n):
+        # Analogous to: ordinal = values[i]
+        ordinal = (<int64_t*>cnp.PyArray_MultiIter_DATA(mi, 1))[0]
+
+        if ordinal == NPY_NAT:
+            item_repr = na_rep
+        else:
+            # This is equivalent to
+            # freq = frequency_corresponding_to_dtype_code(dtype_code)
+            # per = Period(ordinal, freq=freq)
+            # if date_format:
+            #     item_repr = per.strftime(date_format)
+            # else:
+            #     item_repr = str(per)
+            item_repr = period_format(ordinal, dtype_code, date_format)
+
+        # Analogous to: ordinals[i] = ordinal
+        out_flat[i] = item_repr
+
+        cnp.PyArray_MultiIter_NEXT(mi)
+
+    return out
+
+
 # ----------------------------------------------------------------------
 # period accessors
 

pandas/core/arrays/period.py

@@ -616,31 +616,15 @@ def _formatter(self, boxed: bool = False):
             return str
         return "'{}'".format
 
-    @dtl.ravel_compat
     def _format_native_types(
         self, *, na_rep: str | float = "NaT", date_format=None, **kwargs
     ) -> npt.NDArray[np.object_]:
         """
         actually format my specific types
         """
-        values = self.astype(object)
-
-        # Create the formatter function
-        if date_format:
-            formatter = lambda per: per.strftime(date_format)
-        else:
-            # Uses `_Period.str` which in turn uses `format_period`
-            formatter = lambda per: str(per)
-
-        # Apply the formatter to all values in the array, possibly with a mask
-        if self._hasna:
-            mask = self._isnan
-            values[mask] = na_rep
-            imask = ~mask
-            values[imask] = np.array([formatter(per) for per in values[imask]])
-        else:
-            values = np.array([formatter(per) for per in values])
-        return values
+        return libperiod.period_array_strftime(
+            self.asi8, self.dtype._dtype_code, na_rep, date_format
+        )
 
     # ------------------------------------------------------------------
 

pandas/tests/indexes/period/test_formats.py

@@ -13,7 +13,7 @@ def test_to_native_types():
     index = PeriodIndex(["2017-01-01", "2017-01-02", "2017-01-03"], freq="D")
 
     # First, with no arguments.
-    expected = np.array(["2017-01-01", "2017-01-02", "2017-01-03"], dtype="=U10")
+    expected = np.array(["2017-01-01", "2017-01-02", "2017-01-03"], dtype=object)
 
     result = index._format_native_types()
     tm.assert_numpy_array_equal(result, expected)
@@ -23,7 +23,7 @@ def test_to_native_types():
     tm.assert_numpy_array_equal(result, expected)
 
     # Make sure date formatting works
-    expected = np.array(["01-2017-01", "01-2017-02", "01-2017-03"], dtype="=U10")
+    expected = np.array(["01-2017-01", "01-2017-02", "01-2017-03"], dtype=object)
 
     result = index._format_native_types(date_format="%m-%Y-%d")
     tm.assert_numpy_array_equal(result, expected)