Add support for fids filter with use_arrow=True

CHANGES.md

@@ -23,6 +23,7 @@
     intersect the bounding box of the mask when using the Arrow interface for
     some drivers; this has been fixed in GDAL 3.8.0.
 -   Removed warning when no features are read from the data source (#299).
+-   Add support for `fids` filter with `use_arrow=True` (#304)
 
 ### Other changes
 

pyogrio/_io.pyx

@@ -1253,7 +1253,11 @@ def ogr_open_arrow(
         raise ValueError("forcing 2D is not supported for Arrow")
 
     if fids is not None:
-        raise ValueError("reading by FID is not supported for Arrow")
+        if where is not None or bbox is not None or mask is not None or sql is not None or skip_features or max_features:
+            raise ValueError(
+                "cannot set both 'fids' and any of 'where', 'bbox', 'mask', "
+                "'sql', 'skip_features' or 'max_features'"
+            )
 
     IF CTE_GDAL_VERSION < (3, 8, 0):
         if skip_features:
@@ -1315,10 +1319,19 @@ def ogr_open_arrow(
         geometry_name = get_string(OGR_L_GetGeometryColumn(ogr_layer))
 
         fid_column = get_string(OGR_L_GetFIDColumn(ogr_layer))
+        fid_column_where = fid_column
         # OGR_L_GetFIDColumn returns the column name if it is a custom column,
-        # or "" if not. For arrow, the default column name is "OGC_FID".
+        # or "" if not. For arrow, the default column name used to return the FID data
+        # read is "OGC_FID". When accessing the underlying datasource like when using a
+        # where clause, the default column name is "FID".
         if fid_column == "":
             fid_column = "OGC_FID"
+            fid_column_where = "FID"
+
+        # Use fids list to create a where clause, as arrow doesn't support direct fid
+        # filtering.
+        if fids is not None:
+            where = f"{fid_column_where} IN ({','.join(fids.astype(str))})"
 
         # Apply the attribute filter
         if where is not None and where != "":

pyogrio/geopandas.py

@@ -115,7 +115,9 @@ def read_dataframe(
         the starting index is driver and file specific (e.g. typically 0 for
         Shapefile and 1 for GeoPackage, but can still depend on the specific
         file). The performance of reading a large number of features usings FIDs
-        is also driver specific.
+        is also driver specific. The order of the rows returned is undefined. If you
+        would like to sort based on FID, use ``fid_as_index=True`` to have the index of
+        the GeoDataFrame returned set to the FIDs of the features read.
     sql : str, optional (default: None)
         The SQL statement to execute. Look at the sql_dialect parameter for more
         information on the syntax to use for the query. When combined with other

pyogrio/tests/test_geopandas_io.py

@@ -408,10 +408,12 @@ 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):
+def test_read_fids(naturalearth_lowres_all_ext, 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)
+    fids = np.array([1, 5, 10], dtype=np.int64)
+    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)