String dtype: deprecate the pyarrow_numpy storage option (pandas-dev#60152)

* String dtype: deprecate the pyarrow_numpy storage option

* add pyarrow skip

(cherry picked from commit 1908f2e)

Author: jorisvandenbossche

doc/source/whatsnew/v2.3.0.rst

@@ -54,6 +54,7 @@ notable_bug_fix1
 Deprecations
 ~~~~~~~~~~~~
 - Deprecated allowing non-``bool`` values for ``na`` in :meth:`.str.contains`, :meth:`.str.startswith`, and :meth:`.str.endswith` for dtypes that do not already disallow these (:issue:`59615`)
+- Deprecated the ``"pyarrow_numpy"`` storage option for :class:`StringDtype` (:issue:`60152`)
 - The deprecation of setting the argument ``include_groups`` to ``True`` in :meth:`DataFrameGroupBy.apply` has been promoted from a ``DeprecationWarning`` to ``FutureWarning``; only ``False`` will be allowed (:issue:`7155`)
 
 .. ---------------------------------------------------------------------------

pandas/core/arrays/string_.py

@@ -7,6 +7,7 @@
     Literal,
     cast,
 )
+import warnings
 
 import numpy as np
 
@@ -27,6 +28,7 @@
 )
 from pandas.compat.numpy import function as nv
 from pandas.util._decorators import doc
+from pandas.util._exceptions import find_stack_level
 
 from pandas.core.dtypes.base import (
     ExtensionDtype,
@@ -150,7 +152,16 @@ def __init__(
                     storage = "python"
 
         if storage == "pyarrow_numpy":
-            # TODO raise a deprecation warning
+            warnings.warn(
+                "The 'pyarrow_numpy' storage option name is deprecated and will be "
+                'removed in pandas 3.0. Use \'pd.StringDtype(storage="pyarrow", '
+                "na_value-np.nan)' to construct the same dtype.\nOr enable the "
+                "'pd.options.future.infer_string = True' option globally and use "
+                'the "str" alias as a shorthand notation to specify a dtype '
+                '(instead of "string[pyarrow_numpy]").',
+                FutureWarning,
+                stacklevel=find_stack_level(),
+            )
             storage = "pyarrow"
             na_value = np.nan
 
@@ -250,7 +261,7 @@ def construct_from_string(cls, string) -> Self:
         elif string == "string[pyarrow]":
             return cls(storage="pyarrow")
         elif string == "string[pyarrow_numpy]":
-            # TODO deprecate
+            # this is deprecated in the dtype __init__, remove this in pandas 3.0
             return cls(storage="pyarrow_numpy")
         else:
             raise TypeError(f"Cannot construct a '{cls.__name__}' from '{string}'")

pandas/tests/arrays/string_/test_string.py

@@ -41,6 +41,14 @@ def cls(dtype):
     return dtype.construct_array_type()
 
 
+def test_dtype_constructor():
+    pytest.importorskip("pyarrow")
+
+    with tm.assert_produces_warning(FutureWarning):
+        dtype = pd.StringDtype("pyarrow_numpy")
+    assert dtype == pd.StringDtype("pyarrow", na_value=np.nan)
+
+
 def test_dtype_equality():
     pytest.importorskip("pyarrow")
 

pandas/tests/extension/test_string.py

@@ -104,8 +104,8 @@ def test_eq_with_str(self, dtype):
             # only the NA-variant supports parametrized string alias
             assert dtype == f"string[{dtype.storage}]"
         elif dtype.storage == "pyarrow":
-            # TODO(infer_string) deprecate this
-            assert dtype == "string[pyarrow_numpy]"
+            with tm.assert_produces_warning(FutureWarning):
+                assert dtype == "string[pyarrow_numpy]"
 
     def test_is_not_string_type(self, dtype):
         # Different from BaseDtypeTests.test_is_not_string_type