Update links resources

[CasperWA]

Fixes #299

This PR implements the upcoming changes to the LinksResource, updating the OpenAPIs from the pydantic models as well as the related tests.

ChildResource, ParentResource, and ProviderResource has all been removed. They were not used in any case, and the new link_type has gotten the Literal type to "validate" the value.

Note: This PR will fail until providers.optimade.org updates their links resources to the new schema, and changes may still occur, since the update has not yet been merged into the OPTIMADE API specification.

[CasperWA]

This has now been implemented in the specification through Materials-Consortia/OPTIMADE#273. This PR should align itself with the spec PR and as soon as the providers repository has been updated accordingly, this PR is ready.

Edit: Currently waiting for the providers PR Materials-Consortia/providers#34

[giovannipizzi]

Ah, maybe this is why the resources started returning plain dicts instead of python objects with attributes in the tests I am adding in Materials-Consortia/providers#34?

When this is merged, will dictionary key access continue working or I should change back the test code?

[CasperWA]

Ah, maybe this is why the resources started returning plain dicts instead of python objects with attributes in the tests I am adding in Materials-Consortia/providers#34?

Right, yeah! So we have ourselves a circular dependency issue here.

When this is merged, will dictionary key access continue working or I should change back the test code?

I've just tested this. No, unfortunately not.

Maybe I should not have closed Materials-Consortia/providers#37 after all?

[giovannipizzi]

I see. I guess that anyway the ultimate solution is to pin the version of the python tools in the tests. Feel free to reopen Materials-Consortia/providers#37 and we merge that first (note my comment on one missing link_type in there)

[codecov]

Codecov Report

Merging #306 into master will decrease coverage by 0.00%.
The diff coverage is 68.42%.

@@            Coverage Diff             @@
##           master     #306      +/-   ##
==========================================
- Coverage   90.12%   90.11%   -0.01%     
==========================================
  Files          55       55              
  Lines        2289     2287       -2     
==========================================
- Hits         2063     2061       -2     
  Misses        226      226              

Flag	Coverage Δ
#unittests	90.11% <68.42%> (-0.01%)	⬇️

Impacted Files	Coverage Δ
optimade/server/routers/index_info.py	100.00% <ø> (ø)
optimade/validator/validator.py	74.29% <0.00%> (+0.26%)	⬆️
optimade/models/links.py	90.90% <83.33%> (-1.69%)	⬇️
optimade/models/index_metadb.py	94.44% <88.88%> (-5.56%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c0e0b35...bf0e161. Read the comment docs.

[CasperWA]

Can anyone tell me why the OpenAPI CI check fails for these changes?

[shyamd]

Equally perplexed by the OpenAPI JSON. I don't like how much of a circular dependency we have. In the future we should:
1.) Use dependabot to track changes in that repo
2.) Have pulling the "version" of the providers specification as part of the build/test workflow rather than just grabbing the latest version.

[CasperWA]

Can anyone tell me why the OpenAPI CI check fails for these changes?

Doing some local testing it seems the checker fails if using Literal in the pydantic types - for whatever reason. I don't know if this means the openapi.json schema is wrong when generating it using Literal, or if the validator is wrong ...

[CasperWA]

Equally perplexed by the OpenAPI JSON. I don't like how much of a circular dependency we have. In the future we should:
1.) Use dependabot to track changes in that repo
2.) Have pulling the "version" of the providers specification as part of the build/test workflow rather than just grabbing the latest version.

What do you mean exactly with "version" in 2)? We have been testing pulling the list of providers always, albeit implicitly.

[CasperWA]

@ml-evs @shyamd please review, this is all ready to go :)

[ml-evs]

🥇