TCK - Value Domain issue

[NicoLaval]

Hello,

Externalization of examples enables to build and run TCK.

When value domains are involved, information is missing.

For instance, for this script, the value domain definition is textual, and therefore difficult to support in the TCK creation script.

What do you think @antonio-olleros? How are you dealing with?

[antonio-olleros]

I agree this is a problem... We lack a standard way to provide value domains as input, the same with data structures... And we need to define one. This is related, I think, to the work on fomalising a JSON format for providing the data structures (@vpinna80 )

[NicoLaval]

@vpinna80 @antonio-olleros @hadrienk what do you think about adding a value_domains.json files in examples folders (only when needed) like:

[
  {
    "name": "myGeoValueDomain",
    "type": "String",
    "values": ["AF", "BS", "FJ", "GA", "KH", "MO", "PK", "QA", "UG"]
  },
  ...
]

?

[vpinna80]

My opinion, that I pointed out several times now, is that the JSON schema already has a way to declare valuedomains as codelists.
We just need to switch from "one json for dataset" to "one json for example", and from them generate "single structure files" that can be integrated in the HTML/PDF documentation.
The generation of single files from the complete "example json" will be much easier than the other way around as it is now (no information would have to be inferred or manually added).

For example. in the folder in E&E repo for the examples of the check_hierarchy operator, which requires the VD_1 valuedomain, you can look at ex_1.json which is the json file for the example, compliant with the current json schema.

The json files for all the examples the E&E uses were generated with a bash script from the files @antonio-olleros provided, but the valuedomains have been added manually from the documentation (I inferred from the data some valuedomains which were missing a definition).

[antonio-olleros]

Well, I would not mix the two discussions! I think that whether the json admits value domains is completely separate to having one json for example, which in my view is much more complex than what we have now!

[NicoLaval]

Agree @antonio-olleros, I think we have a simple and efficient basis, instead of making an "earthquake", it would be easier to go step by step, by simply improving what already exists, and when this already allows us to run all the tests, we can look into building a new version if necessary.

[vpinna80]

Changing the current way of generating json files was an idea, but having only one json in TCK for each example (on one json for multiple examples) is the only way to guarantee that the generated metadata is compliant with the VTL current information model.

You know where I'm going... With multiple files, you can't exclude redefintion of variables. In fact, that was the case when I was consuming the current json files and I had to manually change the type.

[antonio-olleros]

Yes, but the discussion on the IM is an open issue, as far as I remember... I still need to provide the exmaples... I've been under the water, but I hope I will be able to provide them by next week