Parquet column names

openeo_driver/save_result.py

@@ -144,6 +144,12 @@ class WorkspaceExport:
         merge: Optional[str]
 
 
+def clean_parquet_column_name(name: str):
+    # Could not find official SPEC for this, but cleaning just for consistency
+    # https://geoparquet.org/releases/v1.0.0/#column-metadata
+    return re.sub(r"[()/\\]", "_", name)
+
+
 def get_temp_file(suffix="", prefix="openeo-pydrvr-"):
     # TODO: make sure temp files are cleaned up when read
     _, filename = tempfile.mkstemp(suffix=suffix, prefix=prefix)
@@ -631,7 +637,9 @@ def to_geoparquet(self, destination: Optional[str] = None) -> str:
         # TODO: avoid accessing _geometries to combine _regions and CSV
         gdf = self._regions._geometries
 
-        gpd.GeoDataFrame(stats.join(gdf, on='feature_index')).to_parquet(filename)
+        (gpd.GeoDataFrame(stats.join(gdf, on='feature_index'))
+         .rename(columns=clean_parquet_column_name)
+         .to_parquet(filename))
         return filename
 
     def to_driver_vector_cube(self) -> DriverVectorCube:
@@ -837,7 +845,7 @@ def to_geoparquet(self, destination: Optional[str] = None) -> str:
 
         (gdf
          .join(stats.set_index('feature_index'), on='feature_index')
-         .rename(columns=lambda col_name: re.sub(r"\W", "_", col_name))  # adhere to naming restriction [A-Za-z0-9_]
+         .rename(columns=clean_parquet_column_name)
          .to_parquet(filename))
         # TODO: Is naming restriction required for parquet files?
 

tests/data/aggregate_spatial_spatial_cube_name_collision/1.csv

@@ -0,0 +1,3 @@
+feature_index,avg(band_0),avg_band_0_,avg(band_2)
+1,4645.719597475695,4865.467252259935,5177.803342998465
+0,4646.262612301313,4865.926572218383,5178.517363510712

tests/test_save_result_parquet.py

@@ -1,3 +1,5 @@
+import pytest
+
 from openeo_driver.datacube import DriverVectorCube
 from openeo_driver.save_result import AggregatePolygonSpatialResult
 from .data import get_path
@@ -27,6 +29,20 @@ def test_save_aggregate_polygon_spatial_result(tmp_path):
     }
 
 
+def test_save_column_name(tmp_path):
+    csv_dir = get_path("aggregate_spatial_spatial_cube_name_collision")
+
+    vector_cube = DriverVectorCube(gpd.read_file(str(get_path("geojson/FeatureCollection02.json"))))
+
+    output_file = tmp_path / "test.parquet"
+
+    spatial_result = AggregatePolygonSpatialResult(csv_dir, regions=vector_cube, format="Parquet")
+
+    with pytest.raises(ValueError) as exc_info:
+        spatial_result.to_geoparquet(destination=str(output_file))
+    assert "Duplicate column names found" in str(exc_info.value)
+
+
 def test_write_driver_vector_cube_to_parquet(tmp_path):
     vector_cube = DriverVectorCube(gpd.read_file(str(get_path("geojson/FeatureCollection02.json"))))
     vector_cube.write_assets(tmp_path / "dummy", format="Parquet")