retrieving non-STAC entities via Links throws unintended error

[philvarner]

This code attempts to retrieve the root service-desc link as required by OAFeat. This file is OpenAPI rather than STAC.

catalog: Client = Client.open("https://earth-search.aws.element84.com/v0")
print(list(catalog.get_stac_objects('service-desc')))

Expected:

I would expect to get from this some type of error response that it's not a STAC object -- empty genexp, None, or a documented exception.

Actual:

The attempted read of the non-existant 'links' key in the object dict returns a KeyError.

  File "/Users/philvarner/.local/share/virtualenvs/stac-api-validation-suite-tzI1nfla/lib/python3.9/site-packages/pystac/stac_object.py", line 343, in get_stac_objects
    link.resolve_stac_object(root=self.get_root())
  File "/Users/philvarner/.local/share/virtualenvs/stac-api-validation-suite-tzI1nfla/lib/python3.9/site-packages/pystac/link.py", line 146, in resolve_stac_object
    obj = STAC_IO.read_stac_object(target_href, root=root)
  File "/Users/philvarner/.local/share/virtualenvs/stac-api-validation-suite-tzI1nfla/lib/python3.9/site-packages/pystac/stac_io.py", line 131, in read_stac_object
    return cls.stac_object_from_dict(d, href=uri, root=root)
  File "/Users/philvarner/.local/share/virtualenvs/stac-api-validation-suite-tzI1nfla/lib/python3.9/site-packages/pystac/serialization/__init__.py", line 37, in stac_object_from_dict
    return Catalog.from_dict(d, href=href, root=root)
  File "/Users/philvarner/.local/share/virtualenvs/stac-api-validation-suite-tzI1nfla/lib/python3.9/site-packages/pystac/catalog.py", line 786, in from_dict
    catalog_type = CatalogType.determine_type(d)
  File "/Users/philvarner/.local/share/virtualenvs/stac-api-validation-suite-tzI1nfla/lib/python3.9/site-packages/pystac/catalog.py", line 60, in determine_type
    for link in stac_json['links']:
KeyError: 'links'

[duckontheweb]

Thanks for reporting this @philvarner. It looks like this is coming via pystac-client, which is pinned to 0.5.6, but I've verified that it's also an issue in the latest version (v1.0.0-beta.2).

I would expect to get from this some type of error response that it's not a STAC object -- empty genexp, None, or a documented exception.

Given that this method is called get_stac_objects my vote would be for raising an exception in this case.

It looks like the root cause is that in pystac.serialization.identify_stac_object any JSON objects that don't have a type field and don't match conditions for any other STAC object are assumed to be Catalogs here. We should probably implement something closer to this suggestion by checking for stac_version to see if it's a Catalog and otherwise raising an exception (since we don't claim to support STAC <v0.6 this should work fine).

It seems like having pystac.serialization.identify_stac_object raise a STACTypeException in this case would be appropriate.

We should fix this for the latest version, but also patch the other supported release lines.

[duckontheweb]

Closed via #402