Merge pull request #20 from nasa/DAS-1414-netcdf-assets

DAS-1414 - Check input granule is NetCDF-4 via media type or extension.

harmony_netcdf_to_zarr/stac_utilities.py

@@ -13,6 +13,7 @@
 from pystac import Asset, Catalog, Item
 
 
+VALID_EXTENSIONS = ('.nc4', '.nc', '.h5', '.hdf5', '.hdf')
 VALID_MEDIA_TYPES = ['application/x-hdf5', 'application/x-netcdf',
                      'application/x-netcdf4']
 
@@ -38,11 +39,21 @@ def get_item_url(item: Item) -> Optional[str]:
 
     """
     return next((asset.href for asset in item.assets.values()
-                 if 'data' in (asset.roles or [])
-                 and asset.media_type in VALID_MEDIA_TYPES),
+                 if 'data' in (asset.roles or []) and is_netcdf_asset(asset)),
                 None)
 
 
+def is_netcdf_asset(asset: Asset) -> bool:
+    """ Check that a `pystac.Asset` is a valid NetCDF-4 granule. This can be
+        ascertained via either the media type or by checking the extension of
+        granule itself if that media type is absent.
+
+    """
+    return (asset.media_type in VALID_MEDIA_TYPES
+            or (asset.media_type is None
+                and asset.href.lower().endswith(VALID_EXTENSIONS)))
+
+
 def get_output_catalog(input_catalog: Catalog, zarr_root: str) -> Catalog:
     """ Clone the input STAC catalog and add an item for the Zarr store output.
         This item will need to have the correct spatial and temporal

tests/unit/test_stac_utilities.py

@@ -11,7 +11,8 @@
                                                    get_output_bounding_box,
                                                    get_output_catalog,
                                                    get_output_date_range,
-                                                   get_output_item)
+                                                   get_output_item,
+                                                   is_netcdf_asset)
 
 
 class TestStacUtilities(TestCase):
@@ -271,3 +272,32 @@ def test_get_item_date_range(self):
         with self.subTest('Start and end datetimes'):
             self.assertTupleEqual(get_item_date_range(start_end_item),
                                   (self.datetime_one, self.datetime_two))
+
+    def test_is_netcdf_asset(self):
+        """ Ensure that a NetCDF-4 asset can be correctly identified via either
+            the asset media type or the file extension. The check on the file
+            extension should handle both uppercase and lowercase.
+
+        """
+        test_args = [['NetCDF-4 media type', 'application/x-netcdf4', '.h5'],
+                     ['NetCDF media type', 'application/x-netcdf', '.nc4'],
+                     ['HDF-5 media type', 'application/x-hdf5', '.h5'],
+                     ['.nc4 extension', None, '.nc4'],
+                     ['.nc extension', None, '.nc'],
+                     ['.h5 extension', None, '.h5'],
+                     ['.hdf5 extension', None, '.hdf5'],
+                     ['.hdf extension', None, '.hdf'],
+                     ['.HDF5 extension', None, '.HDF5']]
+
+        for description, media_type, extension in test_args:
+            with self.subTest(description):
+                test_asset = Asset(f'test{extension}', media_type=media_type)
+                self.assertTrue(is_netcdf_asset(test_asset))
+
+        bad_args = [['Bad media-type', 'application/tiff', '.tiff'],
+                    ['Missing media-type, bad extension', None, '.tiff']]
+
+        for description, media_type, extension in bad_args:
+            with self.subTest(description):
+                test_asset = Asset(f'test{extension}', media_type=media_type)
+                self.assertFalse(is_netcdf_asset(test_asset))