ENH: Export inplace volumes table as parquet

src/fmu/dataio/_utils.py

@@ -18,6 +18,7 @@
 import pandas as pd
 import xtgeo
 import yaml
+from pyarrow import Table, output_stream, parquet
 
 from fmu.config import utilities as ut
 
@@ -117,11 +118,7 @@ def export_file(
         )
         obj.to_csv(file, index=False)
     elif file_suffix == ".parquet":
-        from pyarrow import Table
-
         if isinstance(obj, Table):
-            from pyarrow import output_stream, parquet
-
             parquet.write_table(obj, where=output_stream(file))
 
     elif file_suffix == ".json":

src/fmu/dataio/export/rms/inplace_volumes.py

@@ -7,6 +7,7 @@
 
 import numpy as np
 import pandas as pd
+import pyarrow as pa
 
 import fmu.dataio as dio
 from fmu.dataio._logging import null_logger
@@ -248,7 +249,10 @@ def _export_volume_table(self) -> ExportResult:
             rep_include=False,
             table_index=_enums.InplaceVolumes.index_columns(),
         )
-        absolute_export_path = edata.export(self._dataframe)
+
+        volume_table = pa.Table.from_pandas(self._dataframe)
+        absolute_export_path = edata.export(volume_table)
+
         _logger.debug("Volume result to: %s", absolute_export_path)
         return ExportResult(
             items=[

src/fmu/dataio/providers/objectdata/_provider.py

@@ -90,6 +90,7 @@
 from typing import TYPE_CHECKING, Final
 
 import pandas as pd
+import pyarrow as pa
 import xtgeo
 
 from fmu.dataio._definitions import ExportFolder, ValidFormats
@@ -150,10 +151,7 @@ def objectdata_provider_factory(
         return FaultRoomSurfaceProvider(obj=obj, dataio=dataio)
     if isinstance(obj, dict):
         return DictionaryDataProvider(obj=obj, dataio=dataio)
-
-    from pyarrow import Table
-
-    if isinstance(obj, Table):
+    if isinstance(obj, pa.Table):
         return ArrowTableDataProvider(obj=obj, dataio=dataio)
 
     raise NotImplementedError(f"This data type is not currently supported: {type(obj)}")

tests/test_export_rms/test_export_rms_volumetrics.py

@@ -6,6 +6,7 @@
 import jsonschema
 import numpy as np
 import pandas as pd
+import pyarrow.parquet as pq
 import pytest
 
 import fmu.dataio as dataio
@@ -139,7 +140,12 @@ def test_convert_table_from_legacy_to_standard_format(
 
     # check that the exported table is equal to the expected
     out = instance._export_volume_table()
-    exported_table = pd.read_csv(out.items[0].absolute_path)
+    # Note that using `read_parquet()` more than once in the same pytest module causes
+    # errors due to an issue in pandas registering a type extension globally on every
+    # invocation. This is probably a pandas bug.
+    # https://github.com/apache/arrow/issues/41857
+    exported_table = pd.read_parquet(out.items[0].absolute_path)
+
     pd.testing.assert_frame_equal(voltable_standard, exported_table)
 
     # check that the fluid column exists and contains oil and gas
@@ -310,12 +316,12 @@ def test_rms_volumetrics_export_function(
         )
     vol_table_file = result.items[0].absolute_path
 
-    absoulte_path = (
+    absolute_path = (
         rmssetup_with_fmuconfig.parent.parent
-        / "share/results/tables/volumes/geogrid.csv"
+        / "share/results/tables/volumes/geogrid.parquet"
     )
 
-    assert vol_table_file == absoulte_path
+    assert vol_table_file == absolute_path
 
     assert Path(vol_table_file).is_file()
     metadata = dataio.read_metadata(vol_table_file)
@@ -335,8 +341,12 @@ def test_inplace_volumes_payload_validates_against_model(
     model."""
 
     out = exportvolumetrics._export_volume_table()
-    with open(out.items[0].absolute_path) as f:
-        df = pd.read_csv(f).replace(np.nan, None).to_dict(orient="records")
+    df = (
+        pq.read_table(out.items[0].absolute_path)
+        .to_pandas()
+        .replace(np.nan, None)
+        .to_dict(orient="records")
+    )
     InplaceVolumesResult.model_validate(df)  # Throws if invalid
 
 
@@ -349,8 +359,12 @@ def test_inplace_volumes_payload_validates_against_schema(
     schema."""
 
     out = exportvolumetrics._export_volume_table()
-    with open(out.items[0].absolute_path) as f:
-        df = pd.read_csv(f).replace(np.nan, None).to_dict(orient="records")
+    df = (
+        pq.read_table(out.items[0].absolute_path)
+        .to_pandas()
+        .replace(np.nan, None)
+        .to_dict(orient="records")
+    )
     jsonschema.validate(instance=df, schema=dump())  # Throws if invalid