Bugfix for error describing CSV with semi-colon delimeter

HISTORY.rst

@@ -3,6 +3,13 @@ Release History
 
 Unreleased Changes
 ------------------
+* Fixed a bug where extra attributes returned from frictionless
+  would cause a ValidationError when instantiating a ``Resource``.
+  https://github.com/natcap/geometamaker/issues/128
+* Removed the ``encoding`` value when describing raster or vector
+  datasets as they are generally binary files and the value that was
+  given did not reflect a true encoding even for the attribute table
+  of a GDAL vector. https://github.com/natcap/geometamaker/issues/121
 * Added an optional ``spatial`` attribute for tables, archives, and
   collections. The ``spatial`` attribute for rasters and vectors remains
   required. Spatial information for Collections represents the union of the

src/geometamaker/geometamaker.py

@@ -287,9 +287,16 @@ def describe_file(source_dataset_path, scheme):
         {description["path"]}'.encode('utf-8'))
     description['uid'] = f'sizetimestamp:{hash_func.hexdigest()}'
 
-    # We don't have a use for including these attributes in our metadata:
+    # These are other attributes sometimes returned by frictionless.
+    # We don't have a use for them in our metadata and we do not permit
+    # arbitrary extra attributes in our models.
     description.pop('mediatype', None)
     description.pop('name', None)
+    description.pop('profile', None)
+    description.pop('dialect', None)
+    description.pop('hash', None)
+    description.pop('sources', None)
+    description.pop('licenses', None)
     return description
 
 
@@ -357,6 +364,7 @@ def describe_vector(source_dataset_path, scheme, **kwargs):
 
     """
     description = describe_file(source_dataset_path, scheme)
+    description.pop('encoding', None)  # does not make sense for binary data
 
     if 'http' in scheme:
         source_dataset_path = f'/vsicurl/{source_dataset_path}'
@@ -404,6 +412,7 @@ def describe_raster(source_dataset_path, scheme, **kwargs):
     """
     compute_stats = kwargs.get('compute_stats', False)
     description = describe_file(source_dataset_path, scheme)
+    description.pop('encoding', None)  # does not make sense for binary data
     if 'http' in scheme:
         source_dataset_path = f'/vsicurl/{source_dataset_path}'
     info = pygeoprocessing.get_raster_info(source_dataset_path)

src/geometamaker/models.py

@@ -38,7 +38,12 @@ def _deep_update_dict(self_dict, other_dict):
 
 
 class Parent(BaseModel):
-    """Parent class on which to configure validation."""
+    """Parent class on which to configure validation.
+
+    Extra attributes are forbidden because we anticipate
+    users editing YML docs manually and this can help catch
+    accidental edits like a typo in an attribute name.
+    """
 
     model_config = ConfigDict(validate_assignment=True,
                               extra='forbid',

tests/test_geometamaker.py

@@ -172,6 +172,19 @@ def test_describe_csv(self):
         self.assertEqual(field.units, units)
         self.assertEqual(resource.spatial, spatial)
 
+    def test_describe_csv_semicolon_dialect(self):
+        """Test a CSV that uses semicolon delimiter."""
+        import geometamaker
+
+        datasource_path = os.path.join(self.workspace_dir, 'data.csv')
+        with open(datasource_path, 'w') as file:
+            file.write('a;b;c\n')
+            file.write('1;2;3\n')
+
+        resource = geometamaker.describe(datasource_path)
+        field = resource.get_field_description('a')
+        self.assertEqual(field.type, 'integer')
+
     def test_describe_bad_csv(self):
         """MetadataControl: CSV with extra item in row does not fail."""
         import geometamaker