Metadata functionality in EnsembleTableProvider (#1135)

.github/workflows/subsurface.yml

@@ -55,6 +55,7 @@ jobs:
         pip install "scipy<1.9.3"   # breaking change in scipy==1.9.3
         pip install "pytest<7.2.0"
         pip install "pytest-xdist<3.0"
+        pip install "xtgeo<2.20.2"
         pip install .
 
         # Testing against our latest release (including pre-releases)

tests/unit_tests/provider_tests/test_ensemble_table_provider.py

@@ -4,6 +4,7 @@
 import pandas as pd
 
 from webviz_subsurface._providers import (
+    ColumnMetadata,
     EnsembleTableProvider,
     EnsembleTableProviderFactory,
 )
@@ -58,6 +59,8 @@ def test_synthetic_get_column_data(testdata_folder: Path) -> None:
     assert df.shape == (4, 2)
     assert df.columns.tolist() == ["REAL", "STR"]
 
+    assert model.column_metadata("REAL") is None
+
 
 def test_create_from_aggregated_csv_file_smry_csv(
     testdata_folder: Path, tmp_path: Path
@@ -85,6 +88,10 @@ def test_create_from_aggregated_csv_file_smry_csv(
     assert valdf.columns[1] == "YEARS"
     assert valdf["REAL"].nunique() == 3
 
+    # No metadata in csv files
+    meta: Optional[ColumnMetadata] = provider.column_metadata("FOPR")
+    assert meta is None
+
 
 def test_create_from_per_realization_csv_file(
     testdata_folder: Path, tmp_path: Path
@@ -110,6 +117,10 @@ def test_create_from_per_realization_csv_file(
     assert valdf["CONIDX"].nunique() == 24
     assert sorted(valdf["CONIDX"].unique()) == list(range(1, 25))
 
+    # No metadata in csv files
+    meta: Optional[ColumnMetadata] = provider.column_metadata("CONIDX")
+    assert meta is None
+
 
 def test_create_from_per_realization_arrow_file(
     testdata_folder: Path, tmp_path: Path
@@ -126,6 +137,11 @@ def test_create_from_per_realization_arrow_file(
     assert "FOPT" in valdf.columns
     assert valdf["REAL"].nunique() == 100
 
+    # Test metadata
+    meta: Optional[ColumnMetadata] = provider.column_metadata("FOPR")
+    assert meta is not None
+    assert meta.unit == "SM3/DAY"
+
 
 def test_create_from_per_realization_parameter_file(
     testdata_folder: Path, tmp_path: Path
@@ -140,6 +156,12 @@ def test_create_from_per_realization_parameter_file(
     assert "GLOBVAR:FAULT_SEAL_SCALING" in valdf.columns
     assert valdf["REAL"].nunique() == 100
 
+    # No metadata in parameter files
+    meta: Optional[ColumnMetadata] = provider.column_metadata(
+        "GLOBVAR:FAULT_SEAL_SCALING"
+    )
+    assert meta is None
+
 
 def test_create_provider_set_from_aggregated_csv_file(tmp_path: Path) -> None:
     """This tests importing a csv file with an ensemble column with multiple
@@ -165,3 +187,7 @@ def test_create_provider_set_from_aggregated_csv_file(tmp_path: Path) -> None:
             "STOIIP_OIL",
             "SOURCE",
         }.issubset(set(provider.column_names()))
+
+        # No metadata in csv files
+        meta: Optional[ColumnMetadata] = provider.column_metadata("ZONE")
+        assert meta is None

webviz_subsurface/_providers/__init__.py

@@ -27,6 +27,7 @@
     SurfaceServer,
 )
 from .ensemble_table_provider import (
+    ColumnMetadata,
     EnsembleTableProvider,
     EnsembleTableProviderFactory,
     EnsembleTableProviderImplArrow,

webviz_subsurface/_providers/ensemble_table_provider/__init__.py

@@ -1,3 +1,3 @@
-from .ensemble_table_provider import EnsembleTableProvider
+from .ensemble_table_provider import ColumnMetadata, EnsembleTableProvider
 from .ensemble_table_provider_factory import EnsembleTableProviderFactory
 from .ensemble_table_provider_impl_arrow import EnsembleTableProviderImplArrow

webviz_subsurface/_providers/ensemble_table_provider/_field_metadata.py

@@ -0,0 +1,24 @@
+from typing import Optional
+
+import pyarrow as pa
+
+from .ensemble_table_provider import ColumnMetadata
+
+
+def create_column_metadata_from_field_meta(
+    field: pa.Field,
+) -> Optional[ColumnMetadata]:
+    """Create VectorMetadata from keywords stored in the field's metadata"""
+
+    meta_dict = field.metadata
+    if not meta_dict:
+        return None
+
+    try:
+        unit_bytestr = meta_dict[b"unit"]
+    except KeyError:
+        return ColumnMetadata(unit=None)
+
+    return ColumnMetadata(
+        unit=unit_bytestr.decode("ascii"),
+    )

webviz_subsurface/_providers/ensemble_table_provider/ensemble_table_provider.py

@@ -1,9 +1,15 @@
 import abc
+from dataclasses import dataclass
 from typing import List, Optional, Sequence
 
 import pandas as pd
 
 
+@dataclass(frozen=True)
+class ColumnMetadata:
+    unit: Optional[str]
+
+
 class EnsembleTableProvider(abc.ABC):
     @abc.abstractmethod
     def column_names(self) -> List[str]:
@@ -18,3 +24,12 @@ def get_column_data(
         self, column_names: Sequence[str], realizations: Optional[Sequence[int]] = None
     ) -> pd.DataFrame:
         ...
+
+    @abc.abstractmethod
+    def column_metadata(self, column_name: str) -> Optional[ColumnMetadata]:
+        """Returns metadata for the specified column.
+
+        Returns None if no metadata is found for the column.
+        Returns a empty ColumnMetadata object if there is metadata, but it's
+        not the columns specified in ColumnMetadata.
+        """

webviz_subsurface/_providers/ensemble_table_provider/ensemble_table_provider_impl_arrow.py

@@ -13,7 +13,8 @@
     add_per_vector_min_max_to_table_schema_metadata,
     find_min_max_for_numeric_table_columns,
 )
-from .ensemble_table_provider import EnsembleTableProvider
+from ._field_metadata import create_column_metadata_from_field_meta
+from .ensemble_table_provider import ColumnMetadata, EnsembleTableProvider
 
 # Since PyArrow's actual compute functions are not seen by pylint
 # pylint: disable=no-member
@@ -170,6 +171,13 @@ def from_backing_store(
 
         return None
 
+    def _get_or_read_schema(self) -> pa.Schema:
+        if self._cached_reader:
+            return self._cached_reader.schema
+
+        source = pa.memory_map(self._arrow_file_name, "r")
+        return pa.ipc.RecordBatchFileReader(source).schema
+
     def column_names(self) -> List[str]:
         return self._column_names
 
@@ -215,3 +223,8 @@ def get_column_data(
         )
 
         return df
+
+    def column_metadata(self, column_name: str) -> Optional[ColumnMetadata]:
+        schema = self._get_or_read_schema()
+        field = schema.field(column_name)
+        return create_column_metadata_from_field_meta(field)