[backport 2.3.x] String dtype: propagate NaNs as False in predicate methods (eg .str.startswith) (#59616) (#60014)

* String dtype: propagate NaNs as False in predicate methods (eg .str.startswith) (#59616)

(cherry picked from commit 88554d0)

* ignore object dtype inference warnings

Author: jorisvandenbossche

pandas/core/arrays/_arrow_string_mixins.py

@@ -10,15 +10,14 @@
 
 import numpy as np
 
+from pandas._libs import lib
 from pandas.compat import (
     pa_version_under10p1,
     pa_version_under11p0,
     pa_version_under13p0,
     pa_version_under17p0,
 )
 
-from pandas.core.dtypes.missing import isna
-
 if not pa_version_under10p1:
     import pyarrow as pa
     import pyarrow.compute as pc
@@ -38,7 +37,7 @@ class ArrowStringArrayMixin:
     def __init__(self, *args, **kwargs) -> None:
         raise NotImplementedError
 
-    def _convert_bool_result(self, result):
+    def _convert_bool_result(self, result, na=lib.no_default, method_name=None):
         # Convert a bool-dtype result to the appropriate result type
         raise NotImplementedError
 
@@ -212,7 +211,9 @@ def _str_removesuffix(self, suffix: str):
         result = pc.if_else(ends_with, removed, self._pa_array)
         return type(self)(result)
 
-    def _str_startswith(self, pat: str | tuple[str, ...], na: Scalar | None = None):
+    def _str_startswith(
+        self, pat: str | tuple[str, ...], na: Scalar | lib.NoDefault = lib.no_default
+    ):
         if isinstance(pat, str):
             result = pc.starts_with(self._pa_array, pattern=pat)
         else:
@@ -225,11 +226,11 @@ def _str_startswith(self, pat: str | tuple[str, ...], na: Scalar | None = None):
 
                 for p in pat[1:]:
                     result = pc.or_(result, pc.starts_with(self._pa_array, pattern=p))
-        if not isna(na):  # pyright: ignore [reportGeneralTypeIssues]
-            result = result.fill_null(na)
-        return self._convert_bool_result(result)
+        return self._convert_bool_result(result, na=na, method_name="startswith")
 
-    def _str_endswith(self, pat: str | tuple[str, ...], na: Scalar | None = None):
+    def _str_endswith(
+        self, pat: str | tuple[str, ...], na: Scalar | lib.NoDefault = lib.no_default
+    ):
         if isinstance(pat, str):
             result = pc.ends_with(self._pa_array, pattern=pat)
         else:
@@ -242,9 +243,7 @@ def _str_endswith(self, pat: str | tuple[str, ...], na: Scalar | None = None):
 
                 for p in pat[1:]:
                     result = pc.or_(result, pc.ends_with(self._pa_array, pattern=p))
-        if not isna(na):  # pyright: ignore [reportGeneralTypeIssues]
-            result = result.fill_null(na)
-        return self._convert_bool_result(result)
+        return self._convert_bool_result(result, na=na, method_name="endswith")
 
     def _str_isalnum(self):
         result = pc.utf8_is_alnum(self._pa_array)
@@ -283,7 +282,12 @@ def _str_isupper(self):
         return self._convert_bool_result(result)
 
     def _str_contains(
-        self, pat, case: bool = True, flags: int = 0, na=None, regex: bool = True
+        self,
+        pat,
+        case: bool = True,
+        flags: int = 0,
+        na: Scalar | lib.NoDefault = lib.no_default,
+        regex: bool = True,
     ):
         if flags:
             raise NotImplementedError(f"contains not implemented with {flags=}")
@@ -293,19 +297,25 @@ def _str_contains(
         else:
             pa_contains = pc.match_substring
         result = pa_contains(self._pa_array, pat, ignore_case=not case)
-        if not isna(na):  # pyright: ignore [reportGeneralTypeIssues]
-            result = result.fill_null(na)
-        return self._convert_bool_result(result)
+        return self._convert_bool_result(result, na=na, method_name="contains")
 
     def _str_match(
-        self, pat: str, case: bool = True, flags: int = 0, na: Scalar | None = None
+        self,
+        pat: str,
+        case: bool = True,
+        flags: int = 0,
+        na: Scalar | lib.NoDefault = lib.no_default,
     ):
         if not pat.startswith("^"):
             pat = f"^{pat}"
         return self._str_contains(pat, case, flags, na, regex=True)
 
     def _str_fullmatch(
-        self, pat, case: bool = True, flags: int = 0, na: Scalar | None = None
+        self,
+        pat,
+        case: bool = True,
+        flags: int = 0,
+        na: Scalar | lib.NoDefault = lib.no_default,
     ):
         if not pat.endswith("$") or pat.endswith("\\$"):
             pat = f"{pat}$"

pandas/core/arrays/arrow/array.py

@@ -2285,7 +2285,11 @@ def _apply_elementwise(self, func: Callable) -> list[list[Any]]:
             for chunk in self._pa_array.iterchunks()
         ]
 
-    def _convert_bool_result(self, result):
+    def _convert_bool_result(self, result, na=lib.no_default, method_name=None):
+        if na is not lib.no_default and not isna(
+            na
+        ):  # pyright: ignore [reportGeneralTypeIssues]
+            result = result.fill_null(na)
         return type(self)(result)
 
     def _convert_int_result(self, result):

pandas/core/arrays/categorical.py

@@ -2675,16 +2675,28 @@ def _replace(self, *, to_replace, value, inplace: bool = False):
     # ------------------------------------------------------------------------
     # String methods interface
     def _str_map(
-        self, f, na_value=np.nan, dtype=np.dtype("object"), convert: bool = True
+        self, f, na_value=lib.no_default, dtype=np.dtype("object"), convert: bool = True
     ):
         # Optimization to apply the callable `f` to the categories once
         # and rebuild the result by `take`ing from the result with the codes.
         # Returns the same type as the object-dtype implementation though.
-        from pandas.core.arrays import NumpyExtensionArray
-
         categories = self.categories
         codes = self.codes
-        result = NumpyExtensionArray(categories.to_numpy())._str_map(f, na_value, dtype)
+        if categories.dtype == "string":
+            result = categories.array._str_map(f, na_value, dtype)  # type: ignore[attr-defined]
+            if (
+                categories.dtype.na_value is np.nan  # type: ignore[union-attr]
+                and is_bool_dtype(dtype)
+                and (na_value is lib.no_default or isna(na_value))
+            ):
+                # NaN propagates as False for functions with boolean return type
+                na_value = False
+        else:
+            from pandas.core.arrays import NumpyExtensionArray
+
+            result = NumpyExtensionArray(categories.to_numpy())._str_map(
+                f, na_value, dtype
+            )
         return take_nd(result, codes, fill_value=na_value)
 
     def _str_get_dummies(self, sep: str = "|"):

pandas/core/arrays/string_.py

@@ -377,7 +377,11 @@ def _from_scalars(cls, scalars, dtype: DtypeObj) -> Self:
         return cls._from_sequence(scalars, dtype=dtype)
 
     def _str_map(
-        self, f, na_value=None, dtype: Dtype | None = None, convert: bool = True
+        self,
+        f,
+        na_value=lib.no_default,
+        dtype: Dtype | None = None,
+        convert: bool = True,
     ):
         if self.dtype.na_value is np.nan:
             return self._str_map_nan_semantics(
@@ -388,7 +392,7 @@ def _str_map(
 
         if dtype is None:
             dtype = self.dtype
-        if na_value is None:
+        if na_value is lib.no_default:
             na_value = self.dtype.na_value
 
         mask = isna(self)
@@ -458,12 +462,20 @@ def _str_map_str_or_object(
             return lib.map_infer_mask(arr, f, mask.view("uint8"))
 
     def _str_map_nan_semantics(
-        self, f, na_value=None, dtype: Dtype | None = None, convert: bool = True
+        self,
+        f,
+        na_value=lib.no_default,
+        dtype: Dtype | None = None,
+        convert: bool = True,
     ):
         if dtype is None:
             dtype = self.dtype
-        if na_value is None:
-            na_value = self.dtype.na_value
+        if na_value is lib.no_default:
+            if is_bool_dtype(dtype):
+                # NaN propagates as False
+                na_value = False
+            else:
+                na_value = self.dtype.na_value
 
         mask = isna(self)
         arr = np.asarray(self)
@@ -474,7 +486,8 @@ def _str_map_nan_semantics(
                 if is_integer_dtype(dtype):
                     na_value = 0
                 else:
-                    na_value = True
+                    # NaN propagates as False
+                    na_value = False
 
             result = lib.map_infer_mask(
                 arr,
@@ -484,15 +497,13 @@ def _str_map_nan_semantics(
                 na_value=na_value,
                 dtype=np.dtype(cast(type, dtype)),
             )
-            if na_value_is_na and mask.any():
+            if na_value_is_na and is_integer_dtype(dtype) and mask.any():
                 # TODO: we could alternatively do this check before map_infer_mask
                 #  and adjust the dtype/na_value we pass there. Which is more
                 #  performant?
-                if is_integer_dtype(dtype):
-                    result = result.astype("float64")
-                else:
-                    result = result.astype("object")
+                result = result.astype("float64")
                 result[mask] = np.nan
+
             return result
 
         else:

pandas/core/arrays/string_arrow.py

@@ -211,9 +211,29 @@ def insert(self, loc: int, item) -> ArrowStringArray:
             raise TypeError("Scalar must be NA or str")
         return super().insert(loc, item)
 
-    def _convert_bool_result(self, values):
+    def _convert_bool_result(self, values, na=lib.no_default, method_name=None):
+        if na is not lib.no_default and not isna(na) and not isinstance(na, bool):
+            # GH#59561
+            warnings.warn(
+                f"Allowing a non-bool 'na' in obj.str.{method_name} is deprecated "
+                "and will raise in a future version.",
+                FutureWarning,
+                stacklevel=find_stack_level(),
+            )
+            na = bool(na)
+
         if self.dtype.na_value is np.nan:
-            return ArrowExtensionArray(values).to_numpy(na_value=np.nan)
+            if na is lib.no_default or isna(na):
+                # NaN propagates as False
+                values = values.fill_null(False)
+            else:
+                values = values.fill_null(na)
+            return values.to_numpy()
+        else:
+            if na is not lib.no_default and not isna(
+                na
+            ):  # pyright: ignore [reportGeneralTypeIssues]
+                values = values.fill_null(na)
         return BooleanDtype().__from_arrow__(values)
 
     def _maybe_convert_setitem_value(self, value):
@@ -309,22 +329,16 @@ def _data(self):
     _str_slice = ArrowStringArrayMixin._str_slice
 
     def _str_contains(
-        self, pat, case: bool = True, flags: int = 0, na=np.nan, regex: bool = True
+        self,
+        pat,
+        case: bool = True,
+        flags: int = 0,
+        na=lib.no_default,
+        regex: bool = True,
     ):
         if flags:
             return super()._str_contains(pat, case, flags, na, regex)
 
-        if not isna(na):
-            if not isinstance(na, bool):
-                # GH#59561
-                warnings.warn(
-                    "Allowing a non-bool 'na' in obj.str.contains is deprecated "
-                    "and will raise in a future version.",
-                    FutureWarning,
-                    stacklevel=find_stack_level(),
-                )
-                na = bool(na)
-
         return ArrowStringArrayMixin._str_contains(self, pat, case, flags, na, regex)
 
     def _str_replace(