ENH: Stop calling generate_metadata() on export when config is not valid

docs/src/dataio_3_migration.rst

@@ -197,6 +197,14 @@ Additionally
  - The ``return_symlink`` argument to ``export()`` is deprecated. It is redundant and can be removed.
 
 
+Getting partial metadata from generate_metadata() when config is invalid
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+It was previously possible to get partial metadata from ``generate_metadata()``
+when the global config file was invalid. This partial metadata was not valid according
+to the datamodel and could not be uploaded to Sumo. Creating invalid metadata is no
+longer supported, if the config is invalid an empty dictionary is returned instead.
+
+
 Providing settings through environment
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 It was previously possible to have a yml-file specifying global input arguments to 

src/fmu/dataio/_utils.py

@@ -90,6 +90,9 @@ def export_file(
 ) -> str:
     """Export a valid object to file"""
 
+    # create output folder if not existing
+    filename.parent.mkdir(parents=True, exist_ok=True)
+
     if filename.suffix == ".gri" and isinstance(obj, xtgeo.RegularSurface):
         obj.to_file(filename, fformat="irap_binary")
     elif filename.suffix == ".csv" and isinstance(obj, (xtgeo.Polygons, xtgeo.Points)):

src/fmu/dataio/dataio.py

@@ -41,7 +41,9 @@
 from .aggregation import AggregatedData
 from .case import CreateCaseMetadata
 from .preprocessed import ExportPreprocessedData
+from .providers._filedata import FileDataProvider
 from .providers._fmu import FmuProvider, get_fmu_context_from_environment
+from .providers.objectdata._provider import objectdata_provider_factory
 
 # DATAIO_EXAMPLES: Final = dataio_examples()
 INSIDE_RMS: Final = detect_inside_rms()
@@ -728,6 +730,16 @@ def _update_fmt_flag(self) -> None:
 
     def _update_check_settings(self, newsettings: dict) -> None:
         """Update instance settings (properties) from other routines."""
+        # if no newsettings (kwargs) this rutine is not needed
+        if not newsettings:
+            return
+
+        warnings.warn(
+            "In the future it will not be possible to enter following arguments "
+            f"inside the export() / generate_metadata() methods: {list(newsettings)}. "
+            "Please move them up to initialization of the ExportData instance.",
+            FutureWarning,
+        )
         logger.info("Try new settings %s", newsettings)
 
         # derive legal input from dataclass signature
@@ -801,6 +813,25 @@ def _get_fmu_provider(self) -> FmuProvider:
             workflow=self.workflow,
         )
 
+    def _export_without_metadata(self, obj: types.Inferrable) -> str:
+        """
+        Export the object without a metadata file. The absolute export path
+        is found using the FileDataProvider directly.
+        A string with full path to the exported item is returned.
+        """
+        self._update_fmt_flag()
+
+        fmudata = self._get_fmu_provider() if self._fmurun else None
+
+        filemeta = FileDataProvider(
+            dataio=self,
+            objdata=objectdata_provider_factory(obj, self),
+            runpath=fmudata.get_runpath() if fmudata else None,
+        ).get_metadata()
+
+        assert filemeta.absolute_path is not None  # for mypy
+        return export_file(obj, filename=filemeta.absolute_path, flag=self._usefmtflag)
+
     # ==================================================================================
     # Public methods:
     # ==================================================================================
@@ -837,14 +868,14 @@ def generate_metadata(
         logger.info("Generate metadata...")
         logger.info("KW args %s", kwargs)
 
-        if kwargs:
+        if not isinstance(self.config, GlobalConfiguration):
             warnings.warn(
-                "In the future it will not be possible to enter following arguments "
-                f"inside the export() / generate_metadata() methods: {list(kwargs)}. "
-                "Please move them up to initialization of the ExportData instance.",
+                "From fmu.dataio version 3.0 it will not be possible to produce "
+                "metadata when the global config is invalid.",
                 FutureWarning,
             )
-            self._update_check_settings(kwargs)
+
+        self._update_check_settings(kwargs)
 
         if isinstance(obj, (str, Path)):
             if self.casepath is None:
@@ -899,21 +930,21 @@ def export(
                 is_observation=self.is_observation,
             ).export(obj)
 
-        metadata = self.generate_metadata(obj, compute_md5=True, **kwargs)
         logger.info("Object type is: %s", type(obj))
 
+        # should only export object if config is not valid
+        if not isinstance(self.config, GlobalConfiguration):
+            warnings.warn("Data will be exported, but without metadata.", UserWarning)
+            self._update_check_settings(kwargs)
+            return self._export_without_metadata(obj)
+
+        metadata = self.generate_metadata(obj, compute_md5=True, **kwargs)
         outfile = Path(metadata["file"]["absolute_path"])
-        # create output folders if they don't exist
-        outfile.parent.mkdir(parents=True, exist_ok=True)
         metafile = outfile.parent / f".{outfile.name}.yml"
 
         export_file(obj, outfile, flag=self._usefmtflag)
         logger.info("Actual file is:   %s", outfile)
 
-        if isinstance(self.config, GlobalConfiguration):
-            export_metadata_file(metafile, metadata, savefmt=self.meta_format)
-            logger.info("Metadata file is: %s", metafile)
-        else:
-            warnings.warn("Data will be exported, but without metadata.", UserWarning)
-
+        export_metadata_file(metafile, metadata, savefmt=self.meta_format)
+        logger.info("Metadata file is: %s", metafile)
         return str(outfile)

tests/test_units/test_dataio.py

@@ -48,7 +48,8 @@ def test_missing_or_wrong_config_exports_with_warning(monkeypatch, tmp_path, reg
     with pytest.warns(UserWarning, match="The global config"):
         edata = ExportData(config={}, content="depth", name="mysurface")
 
-    meta = edata.generate_metadata(regsurf)
+    with pytest.warns(FutureWarning):
+        meta = edata.generate_metadata(regsurf)
     assert "masterdata" not in meta
 
     # check that obj is created but no metadata is found
@@ -60,6 +61,93 @@ def test_missing_or_wrong_config_exports_with_warning(monkeypatch, tmp_path, reg
         read_metadata(out)
 
 
+def test_wrong_config_exports_correctly_ouside_fmu(
+    monkeypatch, tmp_path, globalconfig1, regsurf
+):
+    """
+    In case a config is invalid, objects are exported without metadata.
+    Test that the export path is correct and equal one with valid config,
+    outside an fmu run.
+    """
+
+    monkeypatch.chdir(tmp_path)
+    name = "mysurface"
+
+    with pytest.warns(UserWarning, match="The global config"), pytest.warns(
+        UserWarning, match="without metadata"
+    ):
+        objpath_cfg_invalid = ExportData(
+            config={},
+            content="depth",
+            name=name,
+        ).export(regsurf)
+
+    objpath_cfg_valid = ExportData(
+        config=globalconfig1,
+        content="depth",
+        name=name,
+    ).export(regsurf)
+
+    assert Path(objpath_cfg_invalid) == tmp_path / f"share/results/maps/{name}.gri"
+    assert Path(objpath_cfg_invalid).exists()
+    assert Path(objpath_cfg_valid).exists()
+    assert objpath_cfg_invalid == objpath_cfg_valid
+
+    # test that it works with deprecated pattern also
+    with pytest.warns(FutureWarning):
+        objpath_cfg_valid = ExportData(config=globalconfig1).export(
+            regsurf,
+            content="depth",
+            name=name,
+        )
+    assert objpath_cfg_invalid == objpath_cfg_valid
+
+
+def test_wrong_config_exports_correctly_in_fmu(
+    monkeypatch, fmurun_w_casemetadata, globalconfig1, regsurf
+):
+    """
+    In case a config is invalid, objects are exported without metadata.
+    Test that the export path is correct and equal to exports with valid config,
+    inside an fmu run.
+    """
+
+    monkeypatch.chdir(fmurun_w_casemetadata)
+    name = "mysurface"
+
+    with pytest.warns(UserWarning, match="The global config"), pytest.warns(
+        UserWarning, match="without metadata"
+    ):
+        objpath_cfg_invalid = ExportData(
+            config={},
+            content="depth",
+            name=name,
+        ).export(regsurf)
+
+    objpath_cfg_valid = ExportData(
+        config=globalconfig1,
+        content="depth",
+        name=name,
+    ).export(regsurf)
+
+    assert (
+        Path(objpath_cfg_invalid)
+        == fmurun_w_casemetadata / f"share/results/maps/{name}.gri"
+    )
+    assert Path(objpath_cfg_invalid).exists()
+    assert Path(objpath_cfg_valid).exists()
+    assert objpath_cfg_invalid == objpath_cfg_valid
+
+    # test that it works with deprecated pattern also
+    with pytest.warns(FutureWarning):
+        objpath_cfg_valid = ExportData(config=globalconfig1).export(
+            regsurf,
+            content="depth",
+            name=name,
+        )
+    assert objpath_cfg_invalid == objpath_cfg_valid
+
+
 def test_config_miss_required_fields(monkeypatch, tmp_path, globalconfig1, regsurf):
     """Global config exists but missing critical data; export file but skip metadata."""
 
@@ -156,7 +244,7 @@ def test_update_check_settings_shall_fail(globalconfig1):
 
     newsettings = {"invalidkey": "some"}
     some = ExportData(config=globalconfig1, content="depth")
-    with pytest.raises(KeyError):
+    with pytest.warns(FutureWarning), pytest.raises(KeyError):
         some._update_check_settings(newsettings)
 
 
@@ -285,7 +373,8 @@ def test_classification(globalconfig1, regsurf):
     # verify that classification is defaulted to internal
     with pytest.warns(UserWarning):
         exp = ExportData(config={}, content="depth")
-    mymeta = exp.generate_metadata(regsurf)
+    with pytest.warns(FutureWarning):
+        mymeta = exp.generate_metadata(regsurf)
     assert mymeta["access"]["classification"] == "internal"
 
 
@@ -308,16 +397,19 @@ def test_rep_include(globalconfig1, regsurf):
     mymeta = exp.generate_metadata(regsurf)
     assert mymeta["access"]["ssdl"]["rep_include"] is True
 
-    # test that rep_include is taken from config if not provided
+    # test that rep_include is defaulted to false if not provided
     exp = ExportData(config=globalconfig1, content="depth")
     mymeta = exp.generate_metadata(regsurf)
     assert mymeta["access"]["ssdl"]["rep_include"] is False
 
-    # test that rep_include is defaulted False
-    with pytest.warns(UserWarning):
-        exp = ExportData(config={}, content="depth")
-    mymeta = exp.generate_metadata(regsurf)
-    assert mymeta["access"]["ssdl"]["rep_include"] is False
+    # add ssdl.rep_include to the config
+    config = deepcopy(globalconfig1)
+    config["access"]["ssdl"] = {"rep_include": True}
+
+    # test that rep_include can be read from config
+    with pytest.warns(FutureWarning, match="is deprecated"):
+        mymeta = ExportData(config=config, content="depth").generate_metadata(regsurf)
+    assert mymeta["access"]["ssdl"]["rep_include"] is True
 
 
 def test_unit_is_none(globalconfig1, regsurf):

tests/test_units/test_preprocessed.py

@@ -317,10 +317,7 @@ def test_export_preprocessed_file_exportdata_casepath_on_export(
         edata = dataio.ExportData(config=rmsglobalconfig, is_observation=True)
 
     # test that error is thrown when missing casepath
-    # (UserWarning initially in ExportData)
-    with pytest.warns(UserWarning, match="case metadata"), pytest.raises(
-        TypeError, match="No 'casepath' argument provided"
-    ):
+    with pytest.raises(TypeError, match="No 'casepath' argument provided"):
         edata.export(surfacepath)
 
     # test that export() works if casepath is provided