ENH: expose to_pandas_kwargs in read_parquet with pyarrow backend

doc/source/whatsnew/v3.0.0.rst

@@ -54,6 +54,7 @@ Other enhancements
 - :meth:`Series.cummin` and :meth:`Series.cummax` now supports :class:`CategoricalDtype` (:issue:`52335`)
 - :meth:`Series.plot` now correctly handle the ``ylabel`` parameter for pie charts, allowing for explicit control over the y-axis label (:issue:`58239`)
 - :meth:`DataFrame.plot.scatter` argument ``c`` now accepts a column of strings, where rows with the same string are colored identically (:issue:`16827` and :issue:`16485`)
+- :func:`read_parquet` accepts ``to_pandas_kwargs`` which are forwarded to :meth:`pyarrow.Table.to_pandas` which enables passing additional keywords to customize the conversion to pandas, such as ``maps_as_pydicts`` to read the Parquet map data type as python dictionaries (:issue:`56842`)
 - :meth:`DataFrameGroupBy.transform`, :meth:`SeriesGroupBy.transform`, :meth:`DataFrameGroupBy.agg`, :meth:`SeriesGroupBy.agg`, :meth:`RollingGroupby.apply`, :meth:`ExpandingGroupby.apply`, :meth:`Rolling.apply`, :meth:`Expanding.apply`, :meth:`DataFrame.apply` with ``engine="numba"`` now supports positional arguments passed as kwargs (:issue:`58995`)
 - :meth:`Series.map` can now accept kwargs to pass on to func (:issue:`59814`)
 - :meth:`pandas.concat` will raise a ``ValueError`` when ``ignore_index=True`` and ``keys`` is not ``None`` (:issue:`59274`)

pandas/io/_util.py

@@ -60,9 +60,12 @@ def arrow_table_to_pandas(
     table: pyarrow.Table,
     dtype_backend: DtypeBackend | Literal["numpy"] | lib.NoDefault = lib.no_default,
     null_to_int64: bool = False,
+    to_pandas_kwargs: dict | None = None,
 ) -> pd.DataFrame:
     pa = import_optional_dependency("pyarrow")
 
+    to_pandas_kwargs = {} if to_pandas_kwargs is None else to_pandas_kwargs
+
     types_mapper: type[pd.ArrowDtype] | None | Callable
     if dtype_backend == "numpy_nullable":
         mapping = _arrow_dtype_mapping()
@@ -80,5 +83,5 @@ def arrow_table_to_pandas(
     else:
         raise NotImplementedError
 
-    df = table.to_pandas(types_mapper=types_mapper)
+    df = table.to_pandas(types_mapper=types_mapper, **to_pandas_kwargs)
     return df

pandas/io/parquet.py

@@ -242,6 +242,7 @@ def read(
         dtype_backend: DtypeBackend | lib.NoDefault = lib.no_default,
         storage_options: StorageOptions | None = None,
         filesystem=None,
+        to_pandas_kwargs: dict[str, Any] | None = None,
         **kwargs,
     ) -> DataFrame:
         kwargs["use_pandas_metadata"] = True
@@ -266,7 +267,11 @@ def read(
                     "make_block is deprecated",
                     DeprecationWarning,
                 )
-                result = arrow_table_to_pandas(pa_table, dtype_backend=dtype_backend)
+                result = arrow_table_to_pandas(
+                    pa_table,
+                    dtype_backend=dtype_backend,
+                    to_pandas_kwargs=to_pandas_kwargs,
+                )
 
             if pa_table.schema.metadata:
                 if b"PANDAS_ATTRS" in pa_table.schema.metadata:
@@ -347,6 +352,7 @@ def read(
         filters=None,
         storage_options: StorageOptions | None = None,
         filesystem=None,
+        to_pandas_kwargs: dict | None = None,
         **kwargs,
     ) -> DataFrame:
         parquet_kwargs: dict[str, Any] = {}
@@ -362,6 +368,10 @@ def read(
             raise NotImplementedError(
                 "filesystem is not implemented for the fastparquet engine."
             )
+        if to_pandas_kwargs is not None:
+            raise NotImplementedError(
+                "to_pandas_kwargs is not implemented for the fastparquet engine."
+            )
         path = stringify_path(path)
         handles = None
         if is_fsspec_url(path):
@@ -452,7 +462,7 @@ def to_parquet(
         .. versionadded:: 2.1.0
 
     kwargs
-        Additional keyword arguments passed to the engine
+        Additional keyword arguments passed to the engine.
 
     Returns
     -------
@@ -491,6 +501,7 @@ def read_parquet(
     dtype_backend: DtypeBackend | lib.NoDefault = lib.no_default,
     filesystem: Any = None,
     filters: list[tuple] | list[list[tuple]] | None = None,
+    to_pandas_kwargs: dict | None = None,
     **kwargs,
 ) -> DataFrame:
     """
@@ -564,6 +575,12 @@ def read_parquet(
 
         .. versionadded:: 2.1.0
 
+    to_pandas_kwargs : dict | None, default None
+        Keyword arguments to pass through to :func:`pyarrow.Table.to_pandas`
+        when ``engine="pyarrow"``.
+
+        .. versionadded:: 3.0.0
+
     **kwargs
         Any additional kwargs are passed to the engine.
 
@@ -636,5 +653,6 @@ def read_parquet(
         storage_options=storage_options,
         dtype_backend=dtype_backend,
         filesystem=filesystem,
+        to_pandas_kwargs=to_pandas_kwargs,
         **kwargs,
     )

pandas/tests/io/test_parquet.py

@@ -1172,6 +1172,20 @@ def test_non_nanosecond_timestamps(self, temp_file):
         )
         tm.assert_frame_equal(result, expected)
 
+    def test_maps_as_pydicts(self, pa):
+        pyarrow = pytest.importorskip("pyarrow", "13.0.0")
+
+        schema = pyarrow.schema(
+            [("foo", pyarrow.map_(pyarrow.string(), pyarrow.int64()))]
+        )
+        df = pd.DataFrame([{"foo": {"A": 1}}, {"foo": {"B": 2}}])
+        check_round_trip(
+            df,
+            pa,
+            write_kwargs={"schema": schema},
+            read_kwargs={"to_pandas_kwargs": {"maps_as_pydicts": "strict"}},
+        )
+
 
 class TestParquetFastParquet(Base):
     def test_basic(self, fp, df_full, request):