pyogrio 0.7.2__cp39-cp39-manylinux_2_28_aarch64.whl → 0.9.0__cp39-cp39-manylinux_2_28_aarch64.whl

pyogrio/tests/test_geopandas_io.py

@@ -1,10 +1,12 @@
 import contextlib
 from datetime import datetime
-import os
+from io import BytesIO
+import locale
+
 import numpy as np
 import pytest
 
-from pyogrio import list_layers, read_info, __gdal_version__
+from pyogrio import list_layers, list_drivers, read_info, __gdal_version__
 from pyogrio.errors import DataLayerError, DataSourceError, FeatureError, GeometryError
 from pyogrio.geopandas import read_dataframe, write_dataframe, PANDAS_GE_20
 from pyogrio.raw import (
@@ -14,10 +16,11 @@ from pyogrio.raw import (
 from pyogrio.tests.conftest import (
     ALL_EXTS,
     DRIVERS,
-    requires_arrow_api,
+    requires_pyarrow_api,
+    requires_arrow_write_api,
     requires_gdal_geos,
 )
-from pyogrio._compat import PANDAS_GE_15
+from pyogrio._compat import PANDAS_GE_15, HAS_ARROW_WRITE_API
 
 try:
     import pandas as pd
@@ -45,13 +48,30 @@ pytest.importorskip("geopandas")
     scope="session",
     params=[
         False,
-        pytest.param(True, marks=requires_arrow_api),
+        pytest.param(True, marks=requires_pyarrow_api),
     ],
 )
 def use_arrow(request):
     return request.param
 
 
+@pytest.fixture(autouse=True)
+def skip_if_no_arrow_write_api(request):
+    # automatically skip tests with use_arrow=True and that require Arrow write
+    # API (marked with `@pytest.mark.requires_arrow_write_api`) if it is not available
+    use_arrow = (
+        request.getfixturevalue("use_arrow")
+        if "use_arrow" in request.fixturenames
+        else False
+    )
+    if (
+        use_arrow
+        and not HAS_ARROW_WRITE_API
+        and request.node.get_closest_marker("requires_arrow_write_api")
+    ):
+        pytest.skip("GDAL>=3.8 required for Arrow write API")
+
+
 def spatialite_available(path):
     try:
         _ = read_dataframe(
@@ -62,6 +82,45 @@ def spatialite_available(path):
         return False
 
 
+@pytest.mark.parametrize("encoding", ["utf-8", "cp1252", None])
+def test_read_csv_encoding(tmp_path, encoding):
+    # Write csv test file. Depending on the os this will be written in a different
+    # encoding: for linux and macos this is utf-8, for windows it is cp1252.
+    csv_path = tmp_path / "test.csv"
+    with open(csv_path, "w", encoding=encoding) as csv:
+        csv.write("näme,city\n")
+        csv.write("Wilhelm Röntgen,Zürich\n")
+
+    # Read csv. The data should be read with the same default encoding as the csv file
+    # was written in, but should have been converted to utf-8 in the dataframe returned.
+    # Hence, the asserts below, with strings in utf-8, be OK.
+    df = read_dataframe(csv_path, encoding=encoding)
+
+    assert len(df) == 1
+    assert df.columns.tolist() == ["näme", "city"]
+    assert df.city.tolist() == ["Zürich"]
+    assert df.näme.tolist() == ["Wilhelm Röntgen"]
+
+
+@pytest.mark.skipif(
+    locale.getpreferredencoding().upper() == "UTF-8",
+    reason="test requires non-UTF-8 default platform",
+)
+def test_read_csv_platform_encoding(tmp_path):
+    """verify that read defaults to platform encoding; only works on Windows (CP1252)"""
+    csv_path = tmp_path / "test.csv"
+    with open(csv_path, "w", encoding=locale.getpreferredencoding()) as csv:
+        csv.write("näme,city\n")
+        csv.write("Wilhelm Röntgen,Zürich\n")
+
+    df = read_dataframe(csv_path)
+
+    assert len(df) == 1
+    assert df.columns.tolist() == ["näme", "city"]
+    assert df.city.tolist() == ["Zürich"]
+    assert df.näme.tolist() == ["Wilhelm Röntgen"]
+
+
 def test_read_dataframe(naturalearth_lowres_all_ext):
     df = read_dataframe(naturalearth_lowres_all_ext)
 
@@ -77,8 +136,8 @@ def test_read_dataframe(naturalearth_lowres_all_ext):
     ]
 
 
-def test_read_dataframe_vsi(naturalearth_lowres_vsi):
-    df = read_dataframe(naturalearth_lowres_vsi[1])
+def test_read_dataframe_vsi(naturalearth_lowres_vsi, use_arrow):
+    df = read_dataframe(naturalearth_lowres_vsi[1], use_arrow=use_arrow)
     assert len(df) == 177
 
 
@@ -154,6 +213,7 @@ def test_read_force_2d(test_fgdb_vsi, use_arrow):
 
 
 @pytest.mark.filterwarnings("ignore: Measured")
+@pytest.mark.filterwarnings("ignore: More than one layer found in")
 def test_read_layer(test_fgdb_vsi, use_arrow):
     layers = list_layers(test_fgdb_vsi)
     kwargs = {"use_arrow": use_arrow, "read_geometry": False, "max_features": 1}
@@ -186,8 +246,13 @@ def test_read_datetime(test_fgdb_vsi, use_arrow):
         assert df.SURVEY_DAT.dtype.name == "datetime64[ns]"
 
 
-def test_read_datetime_tz(test_datetime_tz, tmp_path):
+@pytest.mark.filterwarnings("ignore: Non-conformant content for record 1 in column ")
+@pytest.mark.requires_arrow_write_api
+def test_read_datetime_tz(test_datetime_tz, tmp_path, use_arrow):
     df = read_dataframe(test_datetime_tz)
+    # Make the index non-consecutive to test this case as well. Added for issue
+    # https://github.com/geopandas/pyogrio/issues/324
+    df = df.set_index(np.array([0, 2]))
     raw_expected = ["2020-01-01T09:00:00.123-05:00", "2020-01-01T10:00:00-05:00"]
 
     if PANDAS_GE_20:
@@ -195,15 +260,22 @@ def test_read_datetime_tz(test_datetime_tz, tmp_path):
     else:
         expected = pd.to_datetime(raw_expected)
     expected = pd.Series(expected, name="datetime_col")
-    assert_series_equal(df.datetime_col, expected)
+    assert_series_equal(df.datetime_col, expected, check_index=False)
     # test write and read round trips
     fpath = tmp_path / "test.gpkg"
-    write_dataframe(df, fpath)
-    df_read = read_dataframe(fpath)
+    write_dataframe(df, fpath, use_arrow=use_arrow)
+    df_read = read_dataframe(fpath, use_arrow=use_arrow)
+    if use_arrow:
+        # with Arrow, the datetimes are always read as UTC
+        expected = expected.dt.tz_convert("UTC")
     assert_series_equal(df_read.datetime_col, expected)
 
 
-def test_write_datetime_mixed_offset(tmp_path):
+@pytest.mark.filterwarnings(
+    "ignore: Non-conformant content for record 1 in column dates"
+)
+@pytest.mark.requires_arrow_write_api
+def test_write_datetime_mixed_offset(tmp_path, use_arrow):
     # Australian Summer Time AEDT (GMT+11), Standard Time AEST (GMT+10)
     dates = ["2023-01-01 11:00:01.111", "2023-06-01 10:00:01.111"]
     naive_col = pd.Series(pd.to_datetime(dates), name="dates")
@@ -217,14 +289,18 @@ def test_write_datetime_mixed_offset(tmp_path):
         crs="EPSG:4326",
     )
     fpath = tmp_path / "test.gpkg"
-    write_dataframe(df, fpath)
-    result = read_dataframe(fpath)
+    write_dataframe(df, fpath, use_arrow=use_arrow)
+    result = read_dataframe(fpath, use_arrow=use_arrow)
     # GDAL tz only encodes offsets, not timezones
     # check multiple offsets are read as utc datetime instead of string values
     assert_series_equal(result["dates"], utc_col)
 
 
-def test_read_write_datetime_tz_with_nulls(tmp_path):
+@pytest.mark.filterwarnings(
+    "ignore: Non-conformant content for record 1 in column dates"
+)
+@pytest.mark.requires_arrow_write_api
+def test_read_write_datetime_tz_with_nulls(tmp_path, use_arrow):
     dates_raw = ["2020-01-01T09:00:00.123-05:00", "2020-01-01T10:00:00-05:00", pd.NaT]
     if PANDAS_GE_20:
         dates = pd.to_datetime(dates_raw, format="ISO8601").as_unit("ms")
@@ -235,13 +311,18 @@ def test_read_write_datetime_tz_with_nulls(tmp_path):
         crs="EPSG:4326",
     )
     fpath = tmp_path / "test.gpkg"
-    write_dataframe(df, fpath)
-    result = read_dataframe(fpath)
+    write_dataframe(df, fpath, use_arrow=use_arrow)
+    result = read_dataframe(fpath, use_arrow=use_arrow)
+    if use_arrow:
+        # with Arrow, the datetimes are always read as UTC
+        df["dates"] = df["dates"].dt.tz_convert("UTC")
     assert_geodataframe_equal(df, result)
 
 
 def test_read_null_values(test_fgdb_vsi, use_arrow):
-    df = read_dataframe(test_fgdb_vsi, use_arrow=use_arrow, read_geometry=False)
+    df = read_dataframe(
+        test_fgdb_vsi, layer="basetable_2", use_arrow=use_arrow, read_geometry=False
+    )
 
     # make sure that Null values are preserved
     assert df.SEGMENT_NAME.isnull().max()
@@ -331,6 +412,21 @@ def test_read_where_invalid(request, naturalearth_lowres_all_ext, use_arrow):
         )
 
 
+def test_read_where_ignored_field(naturalearth_lowres, use_arrow):
+    # column included in where is not also included in list of columns, which means
+    # GDAL will return no features
+    # NOTE: this behavior is inconsistent across drivers so only shapefiles are
+    # tested for this
+    df = read_dataframe(
+        naturalearth_lowres,
+        where=""" "iso_a3" = 'CAN' """,
+        columns=["name"],
+        use_arrow=use_arrow,
+    )
+
+    assert len(df) == 0
+
+
 @pytest.mark.parametrize("bbox", [(1,), (1, 2), (1, 2, 3)])
 def test_read_bbox_invalid(naturalearth_lowres_all_ext, bbox, use_arrow):
     with pytest.raises(ValueError, match="Invalid bbox"):
@@ -349,7 +445,7 @@ def test_read_bbox(naturalearth_lowres_all_ext, use_arrow, bbox, expected):
     if (
         use_arrow
         and __gdal_version__ < (3, 8, 0)
-        and os.path.splitext(naturalearth_lowres_all_ext)[1] == ".gpkg"
+        and naturalearth_lowres_all_ext.suffix == ".gpkg"
     ):
         pytest.xfail(reason="GDAL bug: https://github.com/OSGeo/gdal/issues/8347")
 
@@ -438,7 +534,7 @@ def test_read_mask(
     if (
         use_arrow
         and __gdal_version__ < (3, 8, 0)
-        and os.path.splitext(naturalearth_lowres_all_ext)[1] == ".gpkg"
+        and naturalearth_lowres_all_ext.suffix == ".gpkg"
     ):
         pytest.xfail(reason="GDAL bug: https://github.com/OSGeo/gdal/issues/8347")
 
@@ -470,14 +566,45 @@ def test_read_mask_where(naturalearth_lowres_all_ext, use_arrow):
     assert np.array_equal(df.iso_a3, ["CAN"])
 
 
-def test_read_fids(naturalearth_lowres_all_ext):
+@pytest.mark.parametrize("fids", [[1, 5, 10], np.array([1, 5, 10], dtype=np.int64)])
+def test_read_fids(naturalearth_lowres_all_ext, fids, use_arrow):
     # ensure keyword is properly passed through
-    fids = np.array([1, 10, 5], dtype=np.int64)
-    df = read_dataframe(naturalearth_lowres_all_ext, fids=fids, fid_as_index=True)
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, fids=fids, fid_as_index=True, use_arrow=use_arrow
+    )
     assert len(df) == 3
     assert np.array_equal(fids, df.index.values)
 
 
+@requires_pyarrow_api
+def test_read_fids_arrow_max_exception(naturalearth_lowres):
+    # Maximum number at time of writing is 4997 for "OGRSQL". For e.g. for SQLite based
+    # formats like Geopackage, there is no limit.
+    nb_fids = 4998
+    fids = range(nb_fids)
+    with pytest.raises(ValueError, match=f"error applying filter for {nb_fids} fids"):
+        _ = read_dataframe(naturalearth_lowres, fids=fids, use_arrow=True)
+
+
+@requires_pyarrow_api
+@pytest.mark.skipif(
+    __gdal_version__ >= (3, 8, 0), reason="GDAL >= 3.8.0 does not need to warn"
+)
+def test_read_fids_arrow_warning_old_gdal(naturalearth_lowres_all_ext):
+    # A warning should be given for old GDAL versions, except for some file formats.
+    if naturalearth_lowres_all_ext.suffix not in [".gpkg", ".geojson"]:
+        handler = pytest.warns(
+            UserWarning,
+            match="Using 'fids' and 'use_arrow=True' with GDAL < 3.8 can be slow",
+        )
+    else:
+        handler = contextlib.nullcontext()
+
+    with handler:
+        df = read_dataframe(naturalearth_lowres_all_ext, fids=[22], use_arrow=True)
+        assert len(df) == 1
+
+
 def test_read_fids_force_2d(test_fgdb_vsi):
     with pytest.warns(
         UserWarning, match=r"Measured \(M\) geometry types are not supported"
@@ -573,13 +700,17 @@ def test_read_sql(naturalearth_lowres_all_ext, use_arrow):
     # The geometry column cannot be specified when using the
     # default OGRSQL dialect but is returned nonetheless, so 4 columns.
     sql = "SELECT iso_a3 AS iso_a3_renamed, name, pop_est FROM naturalearth_lowres"
-    df = read_dataframe(naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL")
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL", use_arrow=use_arrow
+    )
     assert len(df.columns) == 4
     assert len(df) == 177
 
     # Should return single row
     sql = "SELECT * FROM naturalearth_lowres WHERE iso_a3 = 'CAN'"
-    df = read_dataframe(naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL")
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL", use_arrow=use_arrow
+    )
     assert len(df) == 1
     assert len(df.columns) == 6
     assert df.iloc[0].iso_a3 == "CAN"
@@ -587,7 +718,9 @@ def test_read_sql(naturalearth_lowres_all_ext, use_arrow):
     sql = """SELECT *
                FROM naturalearth_lowres
               WHERE iso_a3 IN ('CAN', 'USA', 'MEX')"""
-    df = read_dataframe(naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL")
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL", use_arrow=use_arrow
+    )
     assert len(df.columns) == 6
     assert len(df) == 3
     assert df.iso_a3.tolist() == ["CAN", "USA", "MEX"]
@@ -596,7 +729,9 @@ def test_read_sql(naturalearth_lowres_all_ext, use_arrow):
                FROM naturalearth_lowres
               WHERE iso_a3 IN ('CAN', 'USA', 'MEX')
               ORDER BY name"""
-    df = read_dataframe(naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL")
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL", use_arrow=use_arrow
+    )
     assert len(df.columns) == 6
     assert len(df) == 3
     assert df.iso_a3.tolist() == ["CAN", "MEX", "USA"]
@@ -605,7 +740,9 @@ def test_read_sql(naturalearth_lowres_all_ext, use_arrow):
     sql = """SELECT *
                FROM naturalearth_lowres
               WHERE POP_EST >= 10000000 AND POP_EST < 100000000"""
-    df = read_dataframe(naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL")
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL", use_arrow=use_arrow
+    )
     assert len(df) == 75
     assert len(df.columns) == 6
     assert df.pop_est.min() >= 10000000
@@ -613,25 +750,36 @@ def test_read_sql(naturalearth_lowres_all_ext, use_arrow):
 
     # Should match no items.
     sql = "SELECT * FROM naturalearth_lowres WHERE ISO_A3 = 'INVALID'"
-    df = read_dataframe(naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL")
+    df = read_dataframe(
+        naturalearth_lowres_all_ext, sql=sql, sql_dialect="OGRSQL", use_arrow=use_arrow
+    )
     assert len(df) == 0
 
 
-def test_read_sql_invalid(naturalearth_lowres_all_ext):
+def test_read_sql_invalid(naturalearth_lowres_all_ext, use_arrow):
     if naturalearth_lowres_all_ext.suffix == ".gpkg":
         with pytest.raises(Exception, match="In ExecuteSQL().*"):
-            read_dataframe(naturalearth_lowres_all_ext, sql="invalid")
+            read_dataframe(
+                naturalearth_lowres_all_ext, sql="invalid", use_arrow=use_arrow
+            )
     else:
         with pytest.raises(Exception, match="SQL Expression Parsing Error"):
-            read_dataframe(naturalearth_lowres_all_ext, sql="invalid")
+            read_dataframe(
+                naturalearth_lowres_all_ext, sql="invalid", use_arrow=use_arrow
+            )
 
     with pytest.raises(
         ValueError, match="'sql' paramater cannot be combined with 'layer'"
     ):
-        read_dataframe(naturalearth_lowres_all_ext, sql="whatever", layer="invalid")
+        read_dataframe(
+            naturalearth_lowres_all_ext,
+            sql="whatever",
+            layer="invalid",
+            use_arrow=use_arrow,
+        )
 
 
-def test_read_sql_columns_where(naturalearth_lowres_all_ext):
+def test_read_sql_columns_where(naturalearth_lowres_all_ext, use_arrow):
     sql = "SELECT iso_a3 AS iso_a3_renamed, name, pop_est FROM naturalearth_lowres"
     df = read_dataframe(
         naturalearth_lowres_all_ext,
@@ -639,13 +787,14 @@ def test_read_sql_columns_where(naturalearth_lowres_all_ext):
         sql_dialect="OGRSQL",
         columns=["iso_a3_renamed", "name"],
         where="iso_a3_renamed IN ('CAN', 'USA', 'MEX')",
+        use_arrow=use_arrow,
     )
     assert len(df.columns) == 3
     assert len(df) == 3
     assert df.iso_a3_renamed.tolist() == ["CAN", "USA", "MEX"]
 
 
-def test_read_sql_columns_where_bbox(naturalearth_lowres_all_ext):
+def test_read_sql_columns_where_bbox(naturalearth_lowres_all_ext, use_arrow):
     sql = "SELECT iso_a3 AS iso_a3_renamed, name, pop_est FROM naturalearth_lowres"
     df = read_dataframe(
         naturalearth_lowres_all_ext,
@@ -654,13 +803,14 @@ def test_read_sql_columns_where_bbox(naturalearth_lowres_all_ext):
         columns=["iso_a3_renamed", "name"],
         where="iso_a3_renamed IN ('CRI', 'PAN')",
         bbox=(-85, 8, -80, 10),
+        use_arrow=use_arrow,
     )
     assert len(df.columns) == 3
     assert len(df) == 2
     assert df.iso_a3_renamed.tolist() == ["PAN", "CRI"]
 
 
-def test_read_sql_skip_max(naturalearth_lowres_all_ext):
+def test_read_sql_skip_max(naturalearth_lowres_all_ext, use_arrow):
     sql = """SELECT *
                FROM naturalearth_lowres
               WHERE iso_a3 IN ('CAN', 'MEX', 'USA')
@@ -671,6 +821,7 @@ def test_read_sql_skip_max(naturalearth_lowres_all_ext):
         skip_features=1,
         max_features=1,
         sql_dialect="OGRSQL",
+        use_arrow=use_arrow,
     )
     assert len(df.columns) == 6
     assert len(df) == 1
@@ -678,13 +829,21 @@ def test_read_sql_skip_max(naturalearth_lowres_all_ext):
 
     sql = "SELECT * FROM naturalearth_lowres LIMIT 1"
     df = read_dataframe(
-        naturalearth_lowres_all_ext, sql=sql, max_features=3, sql_dialect="OGRSQL"
+        naturalearth_lowres_all_ext,
+        sql=sql,
+        max_features=3,
+        sql_dialect="OGRSQL",
+        use_arrow=use_arrow,
     )
     assert len(df) == 1
 
     sql = "SELECT * FROM naturalearth_lowres LIMIT 1"
     df = read_dataframe(
-        naturalearth_lowres_all_ext, sql=sql, skip_features=1, sql_dialect="OGRSQL"
+        naturalearth_lowres_all_ext,
+        sql=sql,
+        sql_dialect="OGRSQL",
+        skip_features=1,
+        use_arrow=use_arrow,
     )
     assert len(df) == 0
 
@@ -695,10 +854,12 @@ def test_read_sql_skip_max(naturalearth_lowres_all_ext):
     [ext for ext in ALL_EXTS if ext != ".gpkg"],
     indirect=["naturalearth_lowres"],
 )
-def test_read_sql_dialect_sqlite_nogpkg(naturalearth_lowres):
+def test_read_sql_dialect_sqlite_nogpkg(naturalearth_lowres, use_arrow):
     # Should return singular item
     sql = "SELECT * FROM naturalearth_lowres WHERE iso_a3 = 'CAN'"
-    df = read_dataframe(naturalearth_lowres, sql=sql, sql_dialect="SQLITE")
+    df = read_dataframe(
+        naturalearth_lowres, sql=sql, sql_dialect="SQLITE", use_arrow=use_arrow
+    )
     assert len(df) == 1
     assert len(df.columns) == 6
     assert df.iloc[0].iso_a3 == "CAN"
@@ -708,7 +869,9 @@ def test_read_sql_dialect_sqlite_nogpkg(naturalearth_lowres):
     sql = """SELECT ST_Buffer(geometry, 5) AS geometry, name, pop_est, iso_a3
                FROM naturalearth_lowres
               WHERE ISO_A3 = 'CAN'"""
-    df = read_dataframe(naturalearth_lowres, sql=sql, sql_dialect="SQLITE")
+    df = read_dataframe(
+        naturalearth_lowres, sql=sql, sql_dialect="SQLITE", use_arrow=use_arrow
+    )
     assert len(df) == 1
     assert len(df.columns) == 4
     assert df.iloc[0].geometry.area > area_canada
@@ -718,12 +881,14 @@ def test_read_sql_dialect_sqlite_nogpkg(naturalearth_lowres):
 @pytest.mark.parametrize(
     "naturalearth_lowres", [".gpkg"], indirect=["naturalearth_lowres"]
 )
-def test_read_sql_dialect_sqlite_gpkg(naturalearth_lowres):
+def test_read_sql_dialect_sqlite_gpkg(naturalearth_lowres, use_arrow):
     # "INDIRECT_SQL" prohibits GDAL from passing the SQL statement to sqlite.
     # Because the statement is processed within GDAL it is possible to use
     # spatialite functions even if sqlite isn't built with spatialite support.
     sql = "SELECT * FROM naturalearth_lowres WHERE iso_a3 = 'CAN'"
-    df = read_dataframe(naturalearth_lowres, sql=sql, sql_dialect="INDIRECT_SQLITE")
+    df = read_dataframe(
+        naturalearth_lowres, sql=sql, sql_dialect="INDIRECT_SQLITE", use_arrow=use_arrow
+    )
     assert len(df) == 1
     assert len(df.columns) == 6
     assert df.iloc[0].iso_a3 == "CAN"
@@ -733,29 +898,67 @@ def test_read_sql_dialect_sqlite_gpkg(naturalearth_lowres):
     sql = """SELECT ST_Buffer(geom, 5) AS geometry, name, pop_est, iso_a3
                FROM naturalearth_lowres
               WHERE ISO_A3 = 'CAN'"""
-    df = read_dataframe(naturalearth_lowres, sql=sql, sql_dialect="INDIRECT_SQLITE")
+    df = read_dataframe(
+        naturalearth_lowres, sql=sql, sql_dialect="INDIRECT_SQLITE", use_arrow=use_arrow
+    )
     assert len(df) == 1
     assert len(df.columns) == 4
     assert df.iloc[0].geometry.area > area_canada
 
 
+@pytest.mark.parametrize("encoding", ["utf-8", "cp1252", None])
+def test_write_csv_encoding(tmp_path, encoding):
+    """Test if write_dataframe uses the default encoding correctly."""
+    # Write csv test file. Depending on the os this will be written in a different
+    # encoding: for linux and macos this is utf-8, for windows it is cp1252.
+    csv_path = tmp_path / "test.csv"
+
+    with open(csv_path, "w", encoding=encoding) as csv:
+        csv.write("näme,city\n")
+        csv.write("Wilhelm Röntgen,Zürich\n")
+
+    # Write csv test file with the same data using write_dataframe. It should use the
+    # same encoding as above.
+    df = pd.DataFrame({"näme": ["Wilhelm Röntgen"], "city": ["Zürich"]})
+    csv_pyogrio_path = tmp_path / "test_pyogrio.csv"
+    write_dataframe(df, csv_pyogrio_path, encoding=encoding)
+
+    # Check if the text files written both ways can be read again and give same result.
+    with open(csv_path, "r", encoding=encoding) as csv:
+        csv_str = csv.read()
+    with open(csv_pyogrio_path, "r", encoding=encoding) as csv_pyogrio:
+        csv_pyogrio_str = csv_pyogrio.read()
+    assert csv_str == csv_pyogrio_str
+
+    # Check if they files are binary identical, to be 100% sure they were written with
+    # the same encoding.
+    with open(csv_path, "rb") as csv:
+        csv_bytes = csv.read()
+    with open(csv_pyogrio_path, "rb") as csv_pyogrio:
+        csv_pyogrio_bytes = csv_pyogrio.read()
+    assert csv_bytes == csv_pyogrio_bytes
+
+
 @pytest.mark.parametrize("ext", ALL_EXTS)
-def test_write_dataframe(tmp_path, naturalearth_lowres, ext):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe(tmp_path, naturalearth_lowres, ext, use_arrow):
     input_gdf = read_dataframe(naturalearth_lowres)
     output_path = tmp_path / f"test{ext}"
 
     if ext == ".fgb":
         # For .fgb, spatial_index=False to avoid the rows being reordered
-        write_dataframe(input_gdf, output_path, spatial_index=False)
+        write_dataframe(
+            input_gdf, output_path, use_arrow=use_arrow, spatial_index=False
+        )
     else:
-        write_dataframe(input_gdf, output_path)
+        write_dataframe(input_gdf, output_path, use_arrow=use_arrow)
 
     assert output_path.exists()
     result_gdf = read_dataframe(output_path)
 
     geometry_types = result_gdf.geometry.type.unique()
     if DRIVERS[ext] in DRIVERS_NO_MIXED_SINGLE_MULTI:
-        assert geometry_types == ["MultiPolygon"]
+        assert list(geometry_types) == ["MultiPolygon"]
     else:
         assert set(geometry_types) == set(["MultiPolygon", "Polygon"])
 
@@ -776,14 +979,21 @@ def test_write_dataframe(tmp_path, naturalearth_lowres, ext):
 
 
 @pytest.mark.filterwarnings("ignore:.*No SRS set on layer.*")
+@pytest.mark.parametrize("write_geodf", [True, False])
 @pytest.mark.parametrize("ext", [ext for ext in ALL_EXTS + [".xlsx"] if ext != ".fgb"])
-def test_write_dataframe_no_geom(tmp_path, naturalearth_lowres, ext):
-    """Test writing a dataframe without a geometry column.
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_no_geom(
+    request, tmp_path, naturalearth_lowres, write_geodf, ext, use_arrow
+):
+    """Test writing a (geo)dataframe without a geometry column.
 
     FlatGeobuf (.fgb) doesn't seem to support this, and just writes an empty file.
     """
     # Prepare test data
     input_df = read_dataframe(naturalearth_lowres, read_geometry=False)
+    if write_geodf:
+        input_df = gp.GeoDataFrame(input_df)
+
     output_path = tmp_path / f"test{ext}"
 
     # A shapefile without geometry column results in only a .dbf file.
@@ -793,7 +1003,7 @@ def test_write_dataframe_no_geom(tmp_path, naturalearth_lowres, ext):
     # Determine driver
     driver = DRIVERS[ext] if ext != ".xlsx" else "XLSX"
 
-    write_dataframe(input_df, output_path, driver=driver)
+    write_dataframe(input_df, output_path, use_arrow=use_arrow, driver=driver)
 
     assert output_path.exists()
     result_df = read_dataframe(output_path)
@@ -806,6 +1016,9 @@ def test_write_dataframe_no_geom(tmp_path, naturalearth_lowres, ext):
     if ext in [".gpkg", ".shp", ".xlsx"]:
         # These file types return a DataFrame when read.
         assert not isinstance(result_df, gp.GeoDataFrame)
+        if isinstance(input_df, gp.GeoDataFrame):
+            input_df = pd.DataFrame(input_df)
+
         pd.testing.assert_frame_equal(
             result_df, input_df, check_index_type=False, check_dtype=check_dtype
         )
@@ -822,12 +1035,27 @@ def test_write_dataframe_no_geom(tmp_path, naturalearth_lowres, ext):
         )
 
 
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_index(tmp_path, naturalearth_lowres, use_arrow):
+    # dataframe writing ignores the index
+    input_gdf = read_dataframe(naturalearth_lowres)
+    input_gdf = input_gdf.set_index("iso_a3")
+
+    output_path = tmp_path / "test.shp"
+    write_dataframe(input_gdf, output_path, use_arrow=use_arrow)
+
+    result_gdf = read_dataframe(output_path)
+    assert isinstance(result_gdf.index, pd.RangeIndex)
+    assert_geodataframe_equal(result_gdf, input_gdf.reset_index(drop=True))
+
+
 @pytest.mark.parametrize("ext", [ext for ext in ALL_EXTS if ext not in ".geojsonl"])
-def test_write_empty_dataframe(tmp_path, ext):
+@pytest.mark.requires_arrow_write_api
+def test_write_empty_dataframe(tmp_path, ext, use_arrow):
     expected = gp.GeoDataFrame(geometry=[], crs=4326)
 
     filename = tmp_path / f"test{ext}"
-    write_dataframe(expected, filename)
+    write_dataframe(expected, filename, use_arrow=use_arrow)
 
     assert filename.exists()
     df = read_dataframe(filename)
@@ -835,83 +1063,119 @@ def test_write_empty_dataframe(tmp_path, ext):
 
 
 @pytest.mark.parametrize("ext", [".geojsonl", ".geojsons"])
-def test_write_read_empty_dataframe_unsupported(tmp_path, ext):
+@pytest.mark.requires_arrow_write_api
+def test_write_read_empty_dataframe_unsupported(tmp_path, ext, use_arrow):
     # Writing empty dataframe to .geojsons or .geojsonl results logically in a 0 byte
     # file, but gdal isn't able to read those again at the time of writing.
     # Issue logged here: https://github.com/geopandas/pyogrio/issues/94
     expected = gp.GeoDataFrame(geometry=[], crs=4326)
 
     filename = tmp_path / f"test{ext}"
-    write_dataframe(expected, filename)
+    write_dataframe(expected, filename, use_arrow=use_arrow)
 
     assert filename.exists()
     with pytest.raises(
-        Exception, match=".* not recognized as a supported file format."
+        Exception, match=".* not recognized as( being in)? a supported file format."
     ):
-        _ = read_dataframe(filename)
+        _ = read_dataframe(filename, use_arrow=use_arrow)
 
 
-def test_write_dataframe_gpkg_multiple_layers(tmp_path, naturalearth_lowres):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_gpkg_multiple_layers(tmp_path, naturalearth_lowres, use_arrow):
     input_gdf = read_dataframe(naturalearth_lowres)
-    output_path = tmp_path / "test.gpkg"
+    filename = tmp_path / "test.gpkg"
 
-    write_dataframe(input_gdf, output_path, layer="first", promote_to_multi=True)
+    write_dataframe(
+        input_gdf,
+        filename,
+        layer="first",
+        promote_to_multi=True,
+        use_arrow=use_arrow,
+    )
 
-    assert os.path.exists(output_path)
-    assert np.array_equal(list_layers(output_path), [["first", "MultiPolygon"]])
+    assert filename.exists()
+    assert np.array_equal(list_layers(filename), [["first", "MultiPolygon"]])
 
-    write_dataframe(input_gdf, output_path, layer="second", promote_to_multi=True)
+    write_dataframe(
+        input_gdf,
+        filename,
+        layer="second",
+        promote_to_multi=True,
+        use_arrow=use_arrow,
+    )
     assert np.array_equal(
-        list_layers(output_path),
+        list_layers(filename),
         [["first", "MultiPolygon"], ["second", "MultiPolygon"]],
     )
 
 
 @pytest.mark.parametrize("ext", ALL_EXTS)
-def test_write_dataframe_append(tmp_path, naturalearth_lowres, ext):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_append(request, tmp_path, naturalearth_lowres, ext, use_arrow):
     if ext == ".fgb" and __gdal_version__ <= (3, 5, 0):
         pytest.skip("Append to FlatGeobuf fails for GDAL <= 3.5.0")
 
     if ext in (".geojsonl", ".geojsons") and __gdal_version__ <= (3, 6, 0):
         pytest.skip("Append to GeoJSONSeq only available for GDAL >= 3.6.0")
 
+    if use_arrow and ext.startswith(".geojson"):
+        # Bug in GDAL when appending int64 to GeoJSON
+        # (https://github.com/OSGeo/gdal/issues/9792)
+        request.node.add_marker(
+            pytest.mark.xfail(reason="Bugs with append when writing Arrow to GeoJSON")
+        )
+
     input_gdf = read_dataframe(naturalearth_lowres)
-    output_path = tmp_path / f"test{ext}"
+    filename = tmp_path / f"test{ext}"
 
-    write_dataframe(input_gdf, output_path)
+    write_dataframe(input_gdf, filename, use_arrow=use_arrow)
 
-    assert os.path.exists(output_path)
-    assert len(read_dataframe(output_path)) == 177
+    filename.exists()
+    assert len(read_dataframe(filename)) == 177
 
-    write_dataframe(input_gdf, output_path, append=True)
-    assert len(read_dataframe(output_path)) == 354
+    write_dataframe(input_gdf, filename, use_arrow=use_arrow, append=True)
+    assert len(read_dataframe(filename)) == 354
 
 
 @pytest.mark.parametrize("spatial_index", [False, True])
-def test_write_dataframe_gdal_options(tmp_path, naturalearth_lowres, spatial_index):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_gdal_options(
+    tmp_path, naturalearth_lowres, spatial_index, use_arrow
+):
     df = read_dataframe(naturalearth_lowres)
 
     outfilename1 = tmp_path / "test1.shp"
-    write_dataframe(df, outfilename1, SPATIAL_INDEX="YES" if spatial_index else "NO")
+    write_dataframe(
+        df,
+        outfilename1,
+        use_arrow=use_arrow,
+        SPATIAL_INDEX="YES" if spatial_index else "NO",
+    )
     assert outfilename1.exists() is True
     index_filename1 = tmp_path / "test1.qix"
     assert index_filename1.exists() is spatial_index
 
     # using explicit layer_options instead
     outfilename2 = tmp_path / "test2.shp"
-    write_dataframe(df, outfilename2, layer_options=dict(spatial_index=spatial_index))
+    write_dataframe(
+        df,
+        outfilename2,
+        use_arrow=use_arrow,
+        layer_options=dict(spatial_index=spatial_index),
+    )
     assert outfilename2.exists() is True
     index_filename2 = tmp_path / "test2.qix"
     assert index_filename2.exists() is spatial_index
 
 
-def test_write_dataframe_gdal_options_unknown(tmp_path, naturalearth_lowres):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_gdal_options_unknown(tmp_path, naturalearth_lowres, use_arrow):
     df = read_dataframe(naturalearth_lowres)
 
     # geojson has no spatial index, so passing keyword should raise
     outfilename = tmp_path / "test.geojson"
     with pytest.raises(ValueError, match="unrecognized option 'SPATIAL_INDEX'"):
-        write_dataframe(df, outfilename, spatial_index=True)
+        write_dataframe(df, outfilename, use_arrow=use_arrow, spatial_index=True)
 
 
 def _get_gpkg_table_names(path):
@@ -924,21 +1188,25 @@ def _get_gpkg_table_names(path):
     return [res[0] for res in result]
 
 
-def test_write_dataframe_gdal_options_dataset(tmp_path, naturalearth_lowres):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_gdal_options_dataset(tmp_path, naturalearth_lowres, use_arrow):
     df = read_dataframe(naturalearth_lowres)
 
     test_default_filename = tmp_path / "test_default.gpkg"
-    write_dataframe(df, test_default_filename)
+    write_dataframe(df, test_default_filename, use_arrow=use_arrow)
     assert "gpkg_ogr_contents" in _get_gpkg_table_names(test_default_filename)
 
     test_no_contents_filename = tmp_path / "test_no_contents.gpkg"
-    write_dataframe(df, test_default_filename, ADD_GPKG_OGR_CONTENTS="NO")
+    write_dataframe(
+        df, test_default_filename, use_arrow=use_arrow, ADD_GPKG_OGR_CONTENTS="NO"
+    )
     assert "gpkg_ogr_contents" not in _get_gpkg_table_names(test_no_contents_filename)
 
     test_no_contents_filename2 = tmp_path / "test_no_contents2.gpkg"
     write_dataframe(
         df,
         test_no_contents_filename2,
+        use_arrow=use_arrow,
         dataset_options=dict(add_gpkg_ogr_contents=False),
     )
     assert "gpkg_ogr_contents" not in _get_gpkg_table_names(test_no_contents_filename2)
@@ -955,6 +1223,7 @@ def test_write_dataframe_gdal_options_dataset(tmp_path, naturalearth_lowres):
         (".geojson", False, ["MultiPolygon", "Polygon"], "Unknown"),
     ],
 )
+@pytest.mark.requires_arrow_write_api
 def test_write_dataframe_promote_to_multi(
     tmp_path,
     naturalearth_lowres,
@@ -962,11 +1231,14 @@ def test_write_dataframe_promote_to_multi(
     promote_to_multi,
     expected_geometry_types,
     expected_geometry_type,
+    use_arrow,
 ):
     input_gdf = read_dataframe(naturalearth_lowres)
 
     output_path = tmp_path / f"test_promote{ext}"
-    write_dataframe(input_gdf, output_path, promote_to_multi=promote_to_multi)
+    write_dataframe(
+        input_gdf, output_path, use_arrow=use_arrow, promote_to_multi=promote_to_multi
+    )
 
     assert output_path.exists()
     output_gdf = read_dataframe(output_path)
@@ -999,6 +1271,7 @@ def test_write_dataframe_promote_to_multi(
         (".shp", True, "Unknown", ["MultiPolygon", "Polygon"], "Polygon"),
     ],
 )
+@pytest.mark.requires_arrow_write_api
 def test_write_dataframe_promote_to_multi_layer_geom_type(
     tmp_path,
     naturalearth_lowres,
@@ -1007,6 +1280,7 @@ def test_write_dataframe_promote_to_multi_layer_geom_type(
     geometry_type,
     expected_geometry_types,
     expected_geometry_type,
+    use_arrow,
 ):
     input_gdf = read_dataframe(naturalearth_lowres)
 
@@ -1023,6 +1297,7 @@ def test_write_dataframe_promote_to_multi_layer_geom_type(
         write_dataframe(
             input_gdf,
             output_path,
+            use_arrow=use_arrow,
             promote_to_multi=promote_to_multi,
             geometry_type=geometry_type,
         )
@@ -1041,9 +1316,15 @@ def test_write_dataframe_promote_to_multi_layer_geom_type(
         (".fgb", False, "Polygon", "Mismatched geometry type"),
         (".fgb", None, "Point", "Mismatched geometry type"),
         (".fgb", None, "Polygon", "Mismatched geometry type"),
-        (".shp", None, "Point", "Could not add feature to layer at index"),
+        (
+            ".shp",
+            None,
+            "Point",
+            "Could not add feature to layer at index|Error while writing batch to OGR layer",
+        ),
     ],
 )
+@pytest.mark.requires_arrow_write_api
 def test_write_dataframe_promote_to_multi_layer_geom_type_invalid(
     tmp_path,
     naturalearth_lowres,
@@ -1051,31 +1332,37 @@ def test_write_dataframe_promote_to_multi_layer_geom_type_invalid(
     promote_to_multi,
     geometry_type,
     expected_raises_match,
+    use_arrow,
 ):
     input_gdf = read_dataframe(naturalearth_lowres)
 
     output_path = tmp_path / f"test{ext}"
-    with pytest.raises(FeatureError, match=expected_raises_match):
+    with pytest.raises((FeatureError, DataLayerError), match=expected_raises_match):
         write_dataframe(
             input_gdf,
             output_path,
+            use_arrow=use_arrow,
             promote_to_multi=promote_to_multi,
             geometry_type=geometry_type,
         )
 
 
-def test_write_dataframe_layer_geom_type_invalid(tmp_path, naturalearth_lowres):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_layer_geom_type_invalid(
+    tmp_path, naturalearth_lowres, use_arrow
+):
     df = read_dataframe(naturalearth_lowres)
 
     filename = tmp_path / "test.geojson"
     with pytest.raises(
         GeometryError, match="Geometry type is not supported: NotSupported"
     ):
-        write_dataframe(df, filename, geometry_type="NotSupported")
+        write_dataframe(df, filename, use_arrow=use_arrow, geometry_type="NotSupported")
 
 
 @pytest.mark.parametrize("ext", [ext for ext in ALL_EXTS if ext not in ".shp"])
-def test_write_dataframe_truly_mixed(tmp_path, ext):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_truly_mixed(tmp_path, ext, use_arrow):
     geometry = [
         shapely.Point(0, 0),
         shapely.LineString([(0, 0), (1, 1)]),
@@ -1095,9 +1382,9 @@ def test_write_dataframe_truly_mixed(tmp_path, ext):
 
     if ext == ".fgb":
         # For .fgb, spatial_index=False to avoid the rows being reordered
-        write_dataframe(df, filename, spatial_index=False)
+        write_dataframe(df, filename, use_arrow=use_arrow, spatial_index=False)
     else:
-        write_dataframe(df, filename)
+        write_dataframe(df, filename, use_arrow=use_arrow)
 
     # Drivers that support mixed geometries will default to "Unknown" geometry type
     assert read_info(filename)["geometry_type"] == "Unknown"
@@ -1105,7 +1392,8 @@ def test_write_dataframe_truly_mixed(tmp_path, ext):
     assert_geodataframe_equal(result, df, check_geom_type=True)
 
 
-def test_write_dataframe_truly_mixed_invalid(tmp_path):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_truly_mixed_invalid(tmp_path, use_arrow):
     # Shapefile doesn't support generic "Geometry" / "Unknown" type
     # for mixed geometries
 
@@ -1123,9 +1411,12 @@ def test_write_dataframe_truly_mixed_invalid(tmp_path):
     msg = (
         "Could not add feature to layer at index 1: Attempt to "
         r"write non-point \(LINESTRING\) geometry to point shapefile."
+        # DataLayerError when using Arrow
+        "|Error while writing batch to OGR layer: Attempt to "
+        r"write non-point \(LINESTRING\) geometry to point shapefile."
     )
-    with pytest.raises(FeatureError, match=msg):
-        write_dataframe(df, tmp_path / "test.shp")
+    with pytest.raises((FeatureError, DataLayerError), match=msg):
+        write_dataframe(df, tmp_path / "test.shp", use_arrow=use_arrow)
 
 
 @pytest.mark.parametrize("ext", [ext for ext in ALL_EXTS if ext not in ".fgb"])
@@ -1138,11 +1429,12 @@ def test_write_dataframe_truly_mixed_invalid(tmp_path):
         [None, None],
     ],
 )
-def test_write_dataframe_infer_geometry_with_nulls(tmp_path, geoms, ext):
+@pytest.mark.requires_arrow_write_api
+def test_write_dataframe_infer_geometry_with_nulls(tmp_path, geoms, ext, use_arrow):
     filename = tmp_path / f"test{ext}"
 
     df = gp.GeoDataFrame({"col": [1.0, 2.0]}, geometry=geoms, crs="EPSG:4326")
-    write_dataframe(df, filename)
+    write_dataframe(df, filename, use_arrow=use_arrow)
     result = read_dataframe(filename)
     assert_geodataframe_equal(result, df)
 
@@ -1150,16 +1442,19 @@ def test_write_dataframe_infer_geometry_with_nulls(tmp_path, geoms, ext):
 @pytest.mark.filterwarnings(
     "ignore: You will likely lose important projection information"
 )
-def test_custom_crs_io(tmpdir, naturalearth_lowres_all_ext):
+@pytest.mark.requires_arrow_write_api
+def test_custom_crs_io(tmp_path, naturalearth_lowres_all_ext, use_arrow):
     df = read_dataframe(naturalearth_lowres_all_ext)
     # project Belgium to a custom Albers Equal Area projection
-    expected = df.loc[df.name == "Belgium"].to_crs(
-        "+proj=aea +lat_1=49.5 +lat_2=51.5 +lon_0=4.3"
+    expected = (
+        df.loc[df.name == "Belgium"]
+        .reset_index(drop=True)
+        .to_crs("+proj=aea +lat_1=49.5 +lat_2=51.5 +lon_0=4.3")
     )
-    filename = os.path.join(str(tmpdir), "test.shp")
-    write_dataframe(expected, filename)
+    filename = tmp_path / "test.shp"
+    write_dataframe(expected, filename, use_arrow=use_arrow)
 
-    assert os.path.exists(filename)
+    assert filename.exists()
 
     df = read_dataframe(filename)
 
@@ -1171,6 +1466,7 @@ def test_custom_crs_io(tmpdir, naturalearth_lowres_all_ext):
 
 
 def test_write_read_mixed_column_values(tmp_path):
+    # use_arrow=True is tested separately below
     mixed_values = ["test", 1.0, 1, datetime.now(), None, np.nan]
     geoms = [shapely.Point(0, 0) for _ in mixed_values]
     test_gdf = gp.GeoDataFrame(
@@ -1187,7 +1483,21 @@ def test_write_read_mixed_column_values(tmp_path):
             assert output_gdf["mixed"][idx] == str(value)
 
 
-def test_write_read_null(tmp_path):
+@requires_arrow_write_api
+def test_write_read_mixed_column_values_arrow(tmp_path):
+    # Arrow cannot represent a column of mixed types
+    mixed_values = ["test", 1.0, 1, datetime.now(), None, np.nan]
+    geoms = [shapely.Point(0, 0) for _ in mixed_values]
+    test_gdf = gp.GeoDataFrame(
+        {"geometry": geoms, "mixed": mixed_values}, crs="epsg:31370"
+    )
+    output_path = tmp_path / "test_write_mixed_column.gpkg"
+    with pytest.raises(TypeError, match=".*Conversion failed for column"):
+        write_dataframe(test_gdf, output_path, use_arrow=True)
+
+
+@pytest.mark.requires_arrow_write_api
+def test_write_read_null(tmp_path, use_arrow):
     output_path = tmp_path / "test_write_nan.gpkg"
     geom = shapely.Point(0, 0)
     test_data = {
@@ -1196,7 +1506,7 @@ def test_write_read_null(tmp_path):
         "object_str": ["test", None, np.nan],
     }
     test_gdf = gp.GeoDataFrame(test_data, crs="epsg:31370")
-    write_dataframe(test_gdf, output_path)
+    write_dataframe(test_gdf, output_path, use_arrow=use_arrow)
     result_gdf = read_dataframe(output_path)
     assert len(test_gdf) == len(result_gdf)
     assert result_gdf["float64"][0] == 1.0
@@ -1219,7 +1529,7 @@ def test_write_read_null(tmp_path):
             ["2.5D MultiLineString", "MultiLineString Z"],
         ),
         (
-            "MultiPolygon Z (((0 0 0, 0 1 0, 1 1 0, 0 0 0)), ((1 1 1, 1 2 1, 2 2 1, 1 1 1)))",  # NOQA
+            "MultiPolygon Z (((0 0 0, 0 1 0, 1 1 0, 0 0 0)), ((1 1 1, 1 2 1, 2 2 1, 1 1 1)))",
             ["2.5D MultiPolygon", "MultiPolygon Z"],
         ),
         (
@@ -1228,11 +1538,12 @@ def test_write_read_null(tmp_path):
         ),
     ],
 )
-def test_write_geometry_z_types(tmp_path, wkt, geom_types):
+@pytest.mark.requires_arrow_write_api
+def test_write_geometry_z_types(tmp_path, wkt, geom_types, use_arrow):
     filename = tmp_path / "test.fgb"
     gdf = gp.GeoDataFrame(geometry=from_wkt([wkt]), crs="EPSG:4326")
     for geom_type in geom_types:
-        write_dataframe(gdf, filename, geometry_type=geom_type)
+        write_dataframe(gdf, filename, use_arrow=use_arrow, geometry_type=geom_type)
         df = read_dataframe(filename)
         assert_geodataframe_equal(df, gdf)
 
@@ -1261,7 +1572,7 @@ def test_write_geometry_z_types(tmp_path, wkt, geom_types):
             "MultiPolygon Z",
             False,
             [
-                "MultiPolygon Z (((0 0 0, 0 1 0, 1 1 0, 0 0 0)), ((1 1 1, 1 2 1, 2 2 1, 1 1 1)))"  # noqa: E501
+                "MultiPolygon Z (((0 0 0, 0 1 0, 1 1 0, 0 0 0)), ((1 1 1, 1 2 1, 2 2 1, 1 1 1)))"
             ],
         ),
         (
@@ -1286,8 +1597,9 @@ def test_write_geometry_z_types(tmp_path, wkt, geom_types):
         ),
     ],
 )
+@pytest.mark.requires_arrow_write_api
 def test_write_geometry_z_types_auto(
-    tmp_path, ext, test_descr, exp_geometry_type, mixed_dimensions, wkt
+    tmp_path, ext, test_descr, exp_geometry_type, mixed_dimensions, wkt, use_arrow
 ):
     # Shapefile has some different behaviour that other file types
     if ext == ".shp":
@@ -1314,10 +1626,10 @@ def test_write_geometry_z_types_auto(
             DataSourceError,
             match=("Mixed 2D and 3D coordinates are not supported by"),
         ):
-            write_dataframe(gdf, filename)
+            write_dataframe(gdf, filename, use_arrow=use_arrow)
         return
     else:
-        write_dataframe(gdf, filename)
+        write_dataframe(gdf, filename, use_arrow=use_arrow)
 
     info = read_info(filename)
     assert info["geometry_type"] == exp_geometry_type
@@ -1329,11 +1641,48 @@ def test_write_geometry_z_types_auto(
     assert_geodataframe_equal(gdf, result_gdf)
 
 
-def test_read_multisurface(data_dir):
-    df = read_dataframe(data_dir / "test_multisurface.gpkg")
+@pytest.mark.parametrize(
+    "on_invalid, message",
+    [
+        (
+            "warn",
+            "Invalid WKB: geometry is returned as None. IllegalArgumentException: "
+            "Invalid number of points in LinearRing found 2 - must be 0 or >=",
+        ),
+        ("raise", "Invalid number of points in LinearRing found 2 - must be 0 or >="),
+        ("ignore", None),
+    ],
+)
+def test_read_invalid_shp(data_dir, use_arrow, on_invalid, message):
+    if on_invalid == "raise":
+        handler = pytest.raises(shapely.errors.GEOSException, match=message)
+    elif on_invalid == "warn":
+        handler = pytest.warns(match=message)
+    elif on_invalid == "ignore":
+        handler = contextlib.nullcontext()
+    else:
+        raise ValueError(f"unknown value for on_invalid: {on_invalid}")
+
+    with handler:
+        df = read_dataframe(
+            data_dir / "poly_not_enough_points.shp.zip",
+            use_arrow=use_arrow,
+            on_invalid=on_invalid,
+        )
+        df.geometry.isnull().all()
+
+
+def test_read_multisurface(data_dir, use_arrow):
+    if use_arrow:
+        with pytest.raises(shapely.errors.GEOSException):
+            # TODO(Arrow)
+            # shapely fails parsing the WKB
+            read_dataframe(data_dir / "test_multisurface.gpkg", use_arrow=True)
+    else:
+        df = read_dataframe(data_dir / "test_multisurface.gpkg")
 
-    # MultiSurface should be converted to MultiPolygon
-    assert df.geometry.type.tolist() == ["MultiPolygon"]
+        # MultiSurface should be converted to MultiPolygon
+        assert df.geometry.type.tolist() == ["MultiPolygon"]
 
 
 def test_read_dataset_kwargs(data_dir, use_arrow):
@@ -1372,7 +1721,8 @@ def test_read_invalid_dataset_kwargs(naturalearth_lowres, use_arrow):
         read_dataframe(naturalearth_lowres, use_arrow=use_arrow, INVALID="YES")
 
 
-def test_write_nullable_dtypes(tmp_path):
+@pytest.mark.requires_arrow_write_api
+def test_write_nullable_dtypes(tmp_path, use_arrow):
     path = tmp_path / "test_nullable_dtypes.gpkg"
     test_data = {
         "col1": pd.Series([1, 2, 3], dtype="int64"),
@@ -1384,7 +1734,7 @@ def test_write_nullable_dtypes(tmp_path):
     input_gdf = gp.GeoDataFrame(
         test_data, geometry=[shapely.Point(0, 0)] * 3, crs="epsg:31370"
     )
-    write_dataframe(input_gdf, path)
+    write_dataframe(input_gdf, path, use_arrow=use_arrow)
     output_gdf = read_dataframe(path)
     # We read it back as default (non-nullable) numpy dtypes, so we cast
     # to those for the expected result
@@ -1393,19 +1743,21 @@ def test_write_nullable_dtypes(tmp_path):
     expected["col3"] = expected["col3"].astype("float32")
     expected["col4"] = expected["col4"].astype("float64")
     expected["col5"] = expected["col5"].astype(object)
+    expected.loc[1, "col5"] = None  # pandas converts to pd.NA on line above
     assert_geodataframe_equal(output_gdf, expected)
 
 
 @pytest.mark.parametrize(
     "metadata_type", ["dataset_metadata", "layer_metadata", "metadata"]
 )
-def test_metadata_io(tmpdir, naturalearth_lowres, metadata_type):
+@pytest.mark.requires_arrow_write_api
+def test_metadata_io(tmp_path, naturalearth_lowres, metadata_type, use_arrow):
     metadata = {"level": metadata_type}
 
     df = read_dataframe(naturalearth_lowres)
 
-    filename = os.path.join(str(tmpdir), "test.gpkg")
-    write_dataframe(df, filename, **{metadata_type: metadata})
+    filename = tmp_path / "test.gpkg"
+    write_dataframe(df, filename, use_arrow=use_arrow, **{metadata_type: metadata})
 
     metadata_key = "layer_metadata" if metadata_type == "metadata" else metadata_type
 
@@ -1421,22 +1773,27 @@ def test_metadata_io(tmpdir, naturalearth_lowres, metadata_type):
         {"key": 1},
     ],
 )
-def test_invalid_metadata(tmpdir, naturalearth_lowres, metadata_type, metadata):
+@pytest.mark.requires_arrow_write_api
+def test_invalid_metadata(
+    tmp_path, naturalearth_lowres, metadata_type, metadata, use_arrow
+):
+    df = read_dataframe(naturalearth_lowres)
     with pytest.raises(ValueError, match="must be a string"):
-        filename = os.path.join(str(tmpdir), "test.gpkg")
         write_dataframe(
-            read_dataframe(naturalearth_lowres), filename, **{metadata_type: metadata}
+            df, tmp_path / "test.gpkg", use_arrow=use_arrow, **{metadata_type: metadata}
         )
 
 
 @pytest.mark.parametrize("metadata_type", ["dataset_metadata", "layer_metadata"])
-def test_metadata_unsupported(tmpdir, naturalearth_lowres, metadata_type):
+@pytest.mark.requires_arrow_write_api
+def test_metadata_unsupported(tmp_path, naturalearth_lowres, metadata_type, use_arrow):
     """metadata is silently ignored"""
 
-    filename = os.path.join(str(tmpdir), "test.geojson")
+    filename = tmp_path / "test.geojson"
     write_dataframe(
         read_dataframe(naturalearth_lowres),
         filename,
+        use_arrow=use_arrow,
         **{metadata_type: {"key": "value"}},
     )
 
@@ -1466,3 +1823,352 @@ def test_read_dataframe_arrow_dtypes(tmp_path):
     assert isinstance(result["col"].dtype, pd.ArrowDtype)
     result["col"] = result["col"].astype("float64")
     assert_geodataframe_equal(result, df)
+
+
+@requires_pyarrow_api
+@pytest.mark.skipif(
+    __gdal_version__ < (3, 8, 3), reason="Arrow bool value bug fixed in GDAL >= 3.8.3"
+)
+@pytest.mark.parametrize("ext", ALL_EXTS)
+def test_arrow_bool_roundtrip(tmp_path, ext):
+    filename = tmp_path / f"test{ext}"
+
+    kwargs = {}
+
+    if ext == ".fgb":
+        # For .fgb, spatial_index=False to avoid the rows being reordered
+        kwargs["spatial_index"] = False
+
+    df = gp.GeoDataFrame(
+        {"bool_col": [True, False, True, False, True], "geometry": [Point(0, 0)] * 5},
+        crs="EPSG:4326",
+    )
+
+    write_dataframe(df, filename, **kwargs)
+    result = read_dataframe(filename, use_arrow=True)
+    # Shapefiles do not support bool columns; these are returned as int32
+    assert_geodataframe_equal(result, df, check_dtype=ext != ".shp")
+
+
+@requires_pyarrow_api
+@pytest.mark.skipif(
+    __gdal_version__ >= (3, 8, 3), reason="Arrow bool value bug fixed in GDAL >= 3.8.3"
+)
+@pytest.mark.parametrize("ext", ALL_EXTS)
+def test_arrow_bool_exception(tmp_path, ext):
+    filename = tmp_path / f"test{ext}"
+
+    df = gp.GeoDataFrame(
+        {"bool_col": [True, False, True, False, True], "geometry": [Point(0, 0)] * 5},
+        crs="EPSG:4326",
+    )
+
+    write_dataframe(df, filename)
+
+    if ext in {".fgb", ".gpkg"}:
+        # only raise exception for GPKG / FGB
+        with pytest.raises(
+            RuntimeError,
+            match="GDAL < 3.8.3 does not correctly read boolean data values using "
+            "the Arrow API",
+        ):
+            read_dataframe(filename, use_arrow=True)
+
+        # do not raise exception if no bool columns are read
+        read_dataframe(filename, use_arrow=True, columns=[])
+
+    else:
+        _ = read_dataframe(filename, use_arrow=True)
+
+
+@pytest.mark.filterwarnings("ignore:File /vsimem:RuntimeWarning")
+@pytest.mark.parametrize("driver", ["GeoJSON", "GPKG"])
+def test_write_memory(naturalearth_lowres, driver):
+    df = read_dataframe(naturalearth_lowres)
+
+    buffer = BytesIO()
+    write_dataframe(df, buffer, driver=driver, layer="test")
+
+    assert len(buffer.getbuffer()) > 0
+
+    actual = read_dataframe(buffer)
+    assert len(actual) == len(df)
+
+    is_json = driver == "GeoJSON"
+
+    assert_geodataframe_equal(
+        actual,
+        df,
+        check_less_precise=is_json,
+        check_index_type=False,
+        check_dtype=not is_json,
+    )
+
+
+def test_write_memory_driver_required(naturalearth_lowres):
+    df = read_dataframe(naturalearth_lowres)
+
+    buffer = BytesIO()
+
+    with pytest.raises(
+        ValueError,
+        match="driver must be provided to write to in-memory file",
+    ):
+        write_dataframe(df.head(1), buffer, driver=None, layer="test")
+
+
+@pytest.mark.parametrize("driver", ["ESRI Shapefile", "OpenFileGDB"])
+def test_write_memory_unsupported_driver(naturalearth_lowres, driver):
+    if driver == "OpenFileGDB" and __gdal_version__ < (3, 6, 0):
+        pytest.skip("OpenFileGDB write support only available for GDAL >= 3.6.0")
+
+    df = read_dataframe(naturalearth_lowres)
+
+    buffer = BytesIO()
+
+    with pytest.raises(
+        ValueError, match=f"writing to in-memory file is not supported for {driver}"
+    ):
+        write_dataframe(df, buffer, driver=driver, layer="test")
+
+
+@pytest.mark.parametrize("driver", ["GeoJSON", "GPKG"])
+def test_write_memory_append_unsupported(naturalearth_lowres, driver):
+    df = read_dataframe(naturalearth_lowres)
+
+    buffer = BytesIO()
+
+    with pytest.raises(
+        NotImplementedError, match="append is not supported for in-memory files"
+    ):
+        write_dataframe(df.head(1), buffer, driver=driver, layer="test", append=True)
+
+
+def test_write_memory_existing_unsupported(naturalearth_lowres):
+    df = read_dataframe(naturalearth_lowres)
+
+    buffer = BytesIO(b"0000")
+    with pytest.raises(
+        NotImplementedError,
+        match="writing to existing in-memory object is not supported",
+    ):
+        write_dataframe(df.head(1), buffer, driver="GeoJSON", layer="test")
+
+
+@pytest.mark.parametrize("ext", ["gpkg", "geojson"])
+def test_non_utf8_encoding_io(tmp_path, ext, encoded_text):
+    """Verify that we write non-UTF data to the data source
+
+    IMPORTANT: this may not be valid for the data source and will likely render
+    them unusable in other tools, but should successfully roundtrip unless we
+    disable writing using other encodings.
+
+    NOTE: FlatGeobuff driver cannot handle non-UTF data in GDAL >= 3.9
+
+    NOTE: pyarrow cannot handle non-UTF-8 characters in this way
+    """
+
+    encoding, text = encoded_text
+    output_path = tmp_path / f"test.{ext}"
+
+    df = gp.GeoDataFrame({text: [text], "geometry": [Point(0, 0)]}, crs="EPSG:4326")
+    write_dataframe(df, output_path, encoding=encoding)
+
+    # cannot open these files without specifying encoding
+    with pytest.raises(UnicodeDecodeError):
+        read_dataframe(output_path)
+
+    # must provide encoding to read these properly
+    actual = read_dataframe(output_path, encoding=encoding)
+    assert actual.columns[0] == text
+    assert actual[text].values[0] == text
+
+
+@requires_pyarrow_api
+@pytest.mark.parametrize("ext", ["gpkg", "geojson"])
+def test_non_utf8_encoding_io_arrow_exception(tmp_path, ext, encoded_text):
+    encoding, text = encoded_text
+    output_path = tmp_path / f"test.{ext}"
+
+    df = gp.GeoDataFrame({text: [text], "geometry": [Point(0, 0)]}, crs="EPSG:4326")
+    write_dataframe(df, output_path, encoding=encoding)
+
+    # cannot open these files without specifying encoding
+    with pytest.raises(UnicodeDecodeError):
+        read_dataframe(output_path)
+
+    with pytest.raises(
+        ValueError, match="non-UTF-8 encoding is not supported for Arrow"
+    ):
+        read_dataframe(output_path, encoding=encoding, use_arrow=True)
+
+
+def test_non_utf8_encoding_io_shapefile(tmp_path, encoded_text, use_arrow):
+    encoding, text = encoded_text
+
+    output_path = tmp_path / "test.shp"
+
+    df = gp.GeoDataFrame({text: [text], "geometry": [Point(0, 0)]}, crs="EPSG:4326")
+    write_dataframe(df, output_path, encoding=encoding)
+
+    # NOTE: GDAL automatically creates a cpg file with the encoding name, which
+    # means that if we read this without specifying the encoding it uses the
+    # correct one
+    actual = read_dataframe(output_path, use_arrow=use_arrow)
+    assert actual.columns[0] == text
+    assert actual[text].values[0] == text
+
+    # verify that if cpg file is not present, that user-provided encoding must be used
+    output_path.with_suffix(".cpg").unlink()
+
+    # We will assume ISO-8859-1, which is wrong
+    miscoded = text.encode(encoding).decode("ISO-8859-1")
+
+    if use_arrow:
+        # pyarrow cannot decode column name with incorrect encoding
+        with pytest.raises(UnicodeDecodeError):
+            read_dataframe(output_path, use_arrow=True)
+    else:
+        bad = read_dataframe(output_path, use_arrow=False)
+        assert bad.columns[0] == miscoded
+        assert bad[miscoded].values[0] == miscoded
+
+    # If encoding is provided, that should yield correct text
+    actual = read_dataframe(output_path, encoding=encoding, use_arrow=use_arrow)
+    assert actual.columns[0] == text
+    assert actual[text].values[0] == text
+
+    # if ENCODING open option, that should yield correct text
+    actual = read_dataframe(output_path, use_arrow=use_arrow, ENCODING=encoding)
+    assert actual.columns[0] == text
+    assert actual[text].values[0] == text
+
+
+def test_encoding_read_option_collision_shapefile(naturalearth_lowres, use_arrow):
+    """Providing both encoding parameter and ENCODING open option (even if blank) is not allowed"""
+
+    with pytest.raises(
+        ValueError, match='cannot provide both encoding parameter and "ENCODING" option'
+    ):
+        read_dataframe(
+            naturalearth_lowres, encoding="CP936", ENCODING="", use_arrow=use_arrow
+        )
+
+
+def test_encoding_write_layer_option_collision_shapefile(tmp_path, encoded_text):
+    """Providing both encoding parameter and ENCODING layer creation option (even if blank) is not allowed"""
+    encoding, text = encoded_text
+
+    output_path = tmp_path / "test.shp"
+    df = gp.GeoDataFrame({text: [text], "geometry": [Point(0, 0)]}, crs="EPSG:4326")
+
+    with pytest.raises(
+        ValueError,
+        match='cannot provide both encoding parameter and "ENCODING" layer creation option',
+    ):
+        write_dataframe(
+            df, output_path, encoding=encoding, layer_options={"ENCODING": ""}
+        )
+
+
+def test_non_utf8_encoding_shapefile_sql(tmp_path, use_arrow):
+    encoding = "CP936"
+
+    output_path = tmp_path / "test.shp"
+
+    mandarin = "中文"
+    df = gp.GeoDataFrame(
+        {mandarin: mandarin, "geometry": [Point(0, 0)]}, crs="EPSG:4326"
+    )
+    write_dataframe(df, output_path, encoding=encoding)
+
+    actual = read_dataframe(
+        output_path,
+        sql=f"select * from test where \"{mandarin}\" = '{mandarin}'",
+        use_arrow=use_arrow,
+    )
+    assert actual.columns[0] == mandarin
+    assert actual[mandarin].values[0] == mandarin
+
+    actual = read_dataframe(
+        output_path,
+        sql=f"select * from test where \"{mandarin}\" = '{mandarin}'",
+        encoding=encoding,
+        use_arrow=use_arrow,
+    )
+    assert actual.columns[0] == mandarin
+    assert actual[mandarin].values[0] == mandarin
+
+
+@pytest.mark.requires_arrow_write_api
+def test_write_kml_file_coordinate_order(tmp_path, use_arrow):
+    # confirm KML coordinates are written in lon, lat order even if CRS axis specifies otherwise
+    points = [Point(10, 20), Point(30, 40), Point(50, 60)]
+    gdf = gp.GeoDataFrame(geometry=points, crs="EPSG:4326")
+    output_path = tmp_path / "test.kml"
+    write_dataframe(
+        gdf, output_path, layer="tmp_layer", driver="KML", use_arrow=use_arrow
+    )
+
+    gdf_in = read_dataframe(output_path, use_arrow=use_arrow)
+
+    assert np.array_equal(gdf_in.geometry.values, points)
+
+    if "LIBKML" in list_drivers():
+        # test appending to the existing file only if LIBKML is available
+        # as it appears to fall back on LIBKML driver when appending.
+        points_append = [Point(70, 80), Point(90, 100), Point(110, 120)]
+        gdf_append = gp.GeoDataFrame(geometry=points_append, crs="EPSG:4326")
+
+        write_dataframe(
+            gdf_append,
+            output_path,
+            layer="tmp_layer",
+            driver="KML",
+            use_arrow=use_arrow,
+            append=True,
+        )
+        # force_2d used to only compare xy geometry as z-dimension is undesirably
+        # introduced when the kml file is over-written.
+        gdf_in_appended = read_dataframe(
+            output_path, use_arrow=use_arrow, force_2d=True
+        )
+
+        assert np.array_equal(gdf_in_appended.geometry.values, points + points_append)
+
+
+@pytest.mark.requires_arrow_write_api
+def test_write_geojson_rfc7946_coordinates(tmp_path, use_arrow):
+    points = [Point(10, 20), Point(30, 40), Point(50, 60)]
+    gdf = gp.GeoDataFrame(geometry=points, crs="EPSG:4326")
+    output_path = tmp_path / "test.geojson"
+    write_dataframe(
+        gdf,
+        output_path,
+        layer="tmp_layer",
+        driver="GeoJSON",
+        RFC7946=True,
+        use_arrow=use_arrow,
+    )
+
+    gdf_in = read_dataframe(output_path, use_arrow=use_arrow)
+
+    assert np.array_equal(gdf_in.geometry.values, points)
+
+    # test appending to the existing file
+
+    points_append = [Point(70, 80), Point(90, 100), Point(110, 120)]
+    gdf_append = gp.GeoDataFrame(geometry=points_append, crs="EPSG:4326")
+
+    write_dataframe(
+        gdf_append,
+        output_path,
+        layer="tmp_layer",
+        driver="GeoJSON",
+        RFC7946=True,
+        use_arrow=use_arrow,
+        append=True,
+    )
+
+    gdf_in_appended = read_dataframe(output_path, use_arrow=use_arrow)
+    assert np.array_equal(gdf_in_appended.geometry.values, points + points_append)