Links under top-level 'links' may be objects

[merkys]

I am trying to validate COD implementation (in development) using optimade-python-tools v0.9.8. The validator seems to require the values under the top-level links being strings:

$ optimade_validator https://www.crystallography.net/cod-test/optimade -v 5
Testing entire implementation at https://www.crystallography.net/cod-test/optimade...

Mandatory tests:
[snip]
2020-07-10 14:51:48,137 - optimade.validator.validator |    DEBUG: Testing multiple entry endpoint of references
✔: https://www.crystallography.net/cod-test/optimade/references?page_limit=5 - request successful.
✖: https://www.crystallography.net/cod-test/optimade/references?page_limit=5 - deserialize_response - failed with error
[snip]
	links -> first
	  str type expected (type=type_error.str)
	links -> next
	  str type expected (type=type_error.str)
Traceback (most recent call last):
[snip]
  File "/home/andrius/.local/lib/python3.6/site-packages/optimade/validator/validator.py", line 619, in get_endpoint
    request_str = request_str.replace("\n", "")
AttributeError: 'dict' object has no attribute 'replace'

In the COD implementation these values are "link objects", as per OPTIMADE specification:

   "links" : {
      "first" : {
         "href" : "http://www.crystallography.net/cod-test/optimade/v1.0.0/references?page_limit=5"
      },
      "next" : {
         "href" : "http://www.crystallography.net/cod-test/optimade/v1.0.0/references?page_limit=5&page_offset=5"
      },
      "base_url" : "http://www.crystallography.net/cod-test/optimade/v1.0.0/"
   }

[ml-evs]

Hi @merkys, something is indeed going wrong here. It looks like we expect lists of links objects (not just in the validator) rather than a dictionary, which looks like the intention in the spec. I'll look into this more shortly.

This one might require a bit more work (as this error really shouldn't crash the validator, no matter what model is expected), but hopefully we can get a hotfix release out tonight (provided someone else is around to review it!)

In the meantime, if you want to validate other endpoints that would be checked after this crash, you can try using --as_type e.g.

optimade_validator -v 4 'https://www.crystallography.net/cod-test/optimade/structures?filter=elements HAS "Na"' --as_type structures

[merkys]

Thanks a lot @ml-evs for the prompt response! No need to hurry, I will get back to validation later.

In the meantime, if you want to validate other endpoints that would be checked after this crash, you can try using --as_type e.g.

optimade_validator -v 4 'https://www.crystallography.net/cod-test/optimade/structures?filter=elements HAS "Na"' --as_type structures

Sadly, this one crashes the validator with a looong traceback. Will open a separate issue with that, as it seems unrelated to me.

Edit: I have opened #397 to describe the aforementioned problem.