made load_stac nicer to mismatch in band names

CHANGELOG.md

@@ -15,6 +15,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 
+- When the bands provided to `Connection.load_stac(..., bands=[...])` do not fully match the bands the client extracted from the STAC metadata, a warning will be triggered, but the provided band names will still be used during the client-side preparation of the process graph. This is a pragmatic approach to bridge the gap between differing interpretations of band detection in STAC. Note that this might produce process graphs that are technically invalid and might not work on other backends or future versions of the backend you currently use. It is recommended to consult with the provider of the STAC metadata and openEO backend on the correct and future-proof band names. ([#752](https://github.com/Open-EO/openeo-python-client/issues/752))
+
 ### Removed
 
 ### Fixed

openeo/metadata.py

@@ -218,6 +218,16 @@ def rename_labels(self, target, source) -> Dimension:
     def rename(self, name) -> Dimension:
         return BandDimension(name=name, bands=self.bands)
 
+    def contains_band(self, band: Union[int, str]) -> bool:
+        """
+        Check if the given band name or index is present in the dimension.
+        """
+        try:
+            self.band_index(band)
+            return True
+        except ValueError:
+            return False
+
 
 class GeometryDimension(Dimension):
     # TODO: how to model/store labels of geometry dimension?

openeo/rest/datacube.py

@@ -441,10 +441,24 @@ def load_stac(
         graph = PGNode("load_stac", arguments=arguments)
         try:
             metadata = metadata_from_stac(url)
+            # TODO: also apply spatial/temporal filters to metadata?
+
             if isinstance(bands, list):
-                # TODO: also apply spatial/temporal filters to metadata?
-                metadata = metadata.filter_bands(band_names=bands)
-        except Exception:
+                if not metadata.has_band_dimension():
+                    metadata = metadata.add_dimension(
+                        name="bands",
+                        type="bands",
+                        label=None,
+                    )
+                unknown_bands = [b for b in bands if not metadata.band_dimension.contains_band(b)]
+                if len(unknown_bands) == 0:
+                    metadata = metadata.filter_bands(band_names=bands)
+                else:
+                    logging.warning(
+                        f"The specified bands {bands} are not a subset of the bands {metadata.band_dimension.band_names} found in the STAC metadata (unknown bands: {unknown_bands}). Using specified bands as is."
+                    )
+                    metadata = metadata.rename_labels(dimension="bands", target=bands)
+        except Exception as e:
             log.warning(f"Failed to extract cube metadata from STAC URL {url}", exc_info=True)
             metadata = None
         return cls(graph=graph, connection=connection, metadata=metadata)

tests/rest/test_connection.py

@@ -2957,7 +2957,40 @@ def test_load_stac_no_cube_extension_temporal_dimension(self, con120, tmp_path,
         cube = con120.load_stac(str(stac_path))
         assert cube.metadata.temporal_dimension == TemporalDimension(name="t", extent=dim_extent)
 
-    def test_load_stac_band_filtering(self, con120, tmp_path):
+    @pytest.mark.parametrize(
+        "bands, expected_warning",
+        [
+            (
+                ["B04"],
+                "The specified bands ['B04'] are not a subset of the bands ['B01', 'B02', 'B03'] found in the STAC metadata (unknown bands: ['B04']). Using specified bands as is.",
+            ),
+            (
+                ["B03", "B04", "B05"],
+                "The specified bands ['B03', 'B04', 'B05'] are not a subset of the bands ['B01', 'B02', 'B03'] found in the STAC metadata (unknown bands: ['B04', 'B05']). Using specified bands as is.",
+            ),
+            (["B03", "B02"], None),
+            (["B01", "B02", "B03"], None),
+        ],
+    )
+    def test_load_stac_band_filtering(self, con120, tmp_path, caplog, bands, expected_warning):
+        stac_path = tmp_path / "stac.json"
+        stac_data = StacDummyBuilder.collection(
+            summaries={"eo:bands": [{"name": "B01"}, {"name": "B02"}, {"name": "B03"}]}
+        )
+        # TODO #738 real request mocking of STAC resources compatible with pystac?
+        stac_path.write_text(json.dumps(stac_data))
+
+        caplog.set_level(logging.WARNING)
+        # Test with non-existing bands in the collection metadata
+        cube = con120.load_stac(str(stac_path), bands=bands)
+        assert cube.metadata.band_names == bands
+        if expected_warning is None:
+            assert caplog.text == ""
+        else:
+            assert expected_warning in caplog.text
+        caplog.clear()
+
+    def test_load_stac_band_filtering_no_requested_bands(self, con120, tmp_path):
         stac_path = tmp_path / "stac.json"
         stac_data = StacDummyBuilder.collection(
             summaries={"eo:bands": [{"name": "B01"}, {"name": "B02"}, {"name": "B03"}]}
@@ -2968,8 +3001,24 @@ def test_load_stac_band_filtering(self, con120, tmp_path):
         cube = con120.load_stac(str(stac_path))
         assert cube.metadata.band_names == ["B01", "B02", "B03"]
 
-        cube = con120.load_stac(str(stac_path), bands=["B03", "B02"])
-        assert cube.metadata.band_names == ["B03", "B02"]
+    def test_load_stac_band_filtering_no_metadata(self, con120, tmp_path, caplog):
+        stac_path = tmp_path / "stac.json"
+        stac_data = StacDummyBuilder.collection()
+        # TODO #738 real request mocking of STAC resources compatible with pystac?
+        stac_path.write_text(json.dumps(stac_data))
+
+        cube = con120.load_stac(str(stac_path))
+        assert cube.metadata.band_names == []
+
+        caplog.set_level(logging.WARNING)
+        cube = con120.load_stac(str(stac_path), bands=["B01", "B02"])
+        assert cube.metadata.band_names == ["B01", "B02"]
+        assert (
+            "The specified bands ['B01', 'B02'] are not a subset of the bands [] found in the STAC metadata (unknown bands: ['B01', 'B02']). Using specified bands as is."
+            in caplog.text
+        )
+        caplog.clear()
+
 
     @pytest.mark.parametrize(
         "bands",

tests/test_metadata.py

@@ -100,6 +100,28 @@ def test_band_dimension_band_index():
         bdim.band_index("yellow")
 
 
+def test_band_dimension_contains_band():
+    bdim = BandDimension(
+        name="spectral",
+        bands=[
+            Band("B02", "blue", 0.490),
+            Band("B03", "green", 0.560),
+            Band("B04", "red", 0.665),
+        ],
+    )
+
+    # Test band names
+    assert bdim.contains_band("B02")
+    assert not bdim.contains_band("B05")
+
+    # Test indexes
+    assert bdim.contains_band(0)
+    assert not bdim.contains_band(4)
+
+    # Test common names
+    assert bdim.contains_band("blue")
+    assert not bdim.contains_band("yellow")
+
 def test_band_dimension_band_name():
     bdim = BandDimension(
         name="spectral",