Five safes on upstream develop

[EttoreM]

Closes #112

---

Five Safes Crate on Upstream RO-Crate 1.2

Goal

The goal of the issue that this PR addresses is checking whether the Five Safes Crate profile, fixtures, and tests from this repository still work when placed on top of the upstream crs4/rocrate-validator develop branch, where the RO-Crate base profile has moved to RO-Crate 1.2.

Steps Taken

1. Created a branch from upstream develop

A new local branch was created from crs4/rocrate-validator/develop:

git remote add upstream-a https://github.com/crs4/rocrate-validator.git
git fetch upstream-a
git switch -c five-safes-on-upstream-develop upstream-a/develop

This made the upstream project the base of the work. This approach was deemed to be more suitable than backporting upstream RO-Crate 1.2 changes into the child repository.

2. Imported Five Safes Crate material

The Five Safes Crate material was copied from this repository's develop branch into the new upstream-based branch:

• rocrate_validator/profiles/five-safes-crate/
• tests/integration/profiles/five-safes-crate/
• tests/data/crates/valid/five-safes-crate-request/
• tests/data/crates/valid/five-safes-crate-result/
• tests/data/crates/valid/five-safes-crate-multiple-context/
• tests/data/crates/valid/five-safes-profile-crate/
• tests/data/crates/invalid/five_safes_crate/

3. Added Five Safes fixture helpers

tests/ro_crates.py was updated with helpers for the imported Five Safes fixtures.

It also adds Invalid5sROC for the invalid Five Safes context-version fixtures.

4. Restored Five Safes SPARQL prefixes

tests/shared.py was updated so SPARQL-based fixture mutation tests could use the necessary prefixes (shp:, rdf:, rocrate:, and dct:, in addition to schema:).

The rocrate: prefix was set to the upstream RO-Crate 1.2 internal namespace:

https://github.com/crs4/rocrate-validator/profiles/ro-crate-1.2/

5. Updated Five Safes to inherit RO-Crate 1.2

The Five Safes profile was updated from:

prof:isTransitiveProfileOf <https://w3id.org/ro/crate/1.1> ;

to:

prof:isTransitiveProfileOf <https://w3id.org/ro/crate/1.2> ;

This makes the Five Safes profile inherit the upstream RO-Crate 1.2 checks.

6. Updated internal RO-Crate shape namespace references

The imported Five Safes SHACL files referred to the old internal RO-Crate namespace:

https://github.com/crs4/rocrate-validator/profiles/ro-crate/

Those references were updated to the upstream RO-Crate 1.2 namespace:

https://github.com/crs4/rocrate-validator/profiles/ro-crate-1.2/

This was needed because upstream RO-Crate 1.2 marks internal inferred classes, such as RootDataEntity, under the ro-crate-1.2 namespace.

7. Updated one valid fixture for RO-Crate 1.2

The result fixture contains a SoftwareApplication entity:

"@id": "https://tre72.example.com/#crate-validator"

RO-Crate 1.2 requires SoftwareApplication entities to have name, url, and version. The fixture already had name, so url and version were added.

Tests Run

Positive tests:

poetry run pytest tests/integration/profiles/five-safes-crate/test_valid_5src.py -q -n 0

Result:

4 passed, 1 xfailed

Full Five Safes Crate profile suite (including negative tests based on graph modifications:

poetry run pytest tests/integration/profiles/five-safes-crate -q -n 0

Result:

115 passed, 1 xfailed

The test output includes many rdflib and pyparsing deprecation warnings. These warnings were already present during successful runs and did not indicate validation failures.

[elichad]

@EttoreM this PR is so big it makes the GitHub interface really slow!

I have created #127 as an intermediate point (WIP while I merge it with our dev branch properly) - that brings in the updates from upstream, and then your updates to the profiles will sit in this PR and are easier to review in isolation.

[elichad]

@EttoreM your changes look good. I think a couple of checks are now obsolete with this update (see comments)

[elichad]

Now that the profile is based on 1.2, we should be able to remove the 15_metadata_file checks altogether

[elichad]

this referenced issue has now been fixed, so this comment can be removed. I think we may as well keep the code the same though now we have it set up

[elichad]

please keep this comment as it's still relevant for now. FYI I've raised crs4#187 upstream to see if it's time for DEFAULT_PROFILE_IDENTIFIER to be updated to 1.2

[elichad]

I actually don't think we use this prefix in any of our tests, so suggest we remove it

Suggested change

PREFIX rocrate: <https://github.com/crs4/rocrate-validator/profiles/ro-crate-1.2/>

[elichad]

Clarifying the plan here as I'm not sure I was clear enough in the meeting:

• Bring in changes from upstream v0.11.2 #127 (branch upstream-develop) can be merged once the comment Bring in changes from upstream v0.11.2 #127 (comment) is answered. Merging will cause the CI to fail on develop, this is fine.
• this PR currently targets branch upstream-develop as its base, so that the diff only shows the changes @EttoreM made to update our fork to align with upstream. This makes it a lot easier to review
• once Bring in changes from upstream v0.11.2 #127 is merged, the base of this PR should be changed back to develop like normal
• then once the comments here are resolved, this PR can be merged

it's fine to go ahead with all of this while I'm away!