Allow scalar data sources in json metadata descriptor

[vpinna80]

This patch allows the json metadata file to also represent scalar values. It's a terminology change to make the change more evident.

[hadrienk]

Should we take the opportunity to version the schema so we can make these changes without breaking implementations?

[vpinna80]

I'll prepare another pull request to start a versioning of the guidelines.

[antonio-olleros]

For me the change in terminology, but I'm not sure with this change how would scalars be represented, can you please clarify, @vpinna80 ?

Regarding versioning, I would be careful, because versioning is not included in the standard...

[vpinna80]

For me the change in terminology, but I'm not sure with this change how would scalars be represented, can you please clarify, @vpinna80 ?

the scalars would be represented by directly inputing a value domain instead of a structure name:

{
  "data": [
    {
      "name": "DS_1",
      "metadata": "DS_1_str"
    },
    {
      "name": "an_integer_scalar"
      "metadata": "integer"
    },
    {
      "name": "a_country_scalar"
      "metadata": "CL_COUNTRY"
    }
  ]
}

[antonio-olleros]

I see, thanks for the clarification.
Wouldn't it make sense to add a property type, with possible values {'structure', 'scalar', 'valuedomain'}? Other wise we need to implement logic to find out what it is, and we can have conflicts (e.g., someone decides to have a valuedomain and a structure with the same name...)

[vpinna80]

@antonio-olleros you have good eyes... I'll make a different proposal.

Regarding the versioning, please move that discussion to #531

[vpinna80]

Here's the example with the new proposal, i used two alternative properties to maintain compatibility with the previous json:

{
  "data": [
    {
      "name": "DS_1",
      "structure": "DS_1_str"
    },
    {
      "name": "an_integer_scalar"
      "valuedomain": "integer"
    },
    {
      "name": "a_country_scalar"
      "valuedomain": "CL_COUNTRY"
    }
  ]
}

[antonio-olleros]

Thanks for the suggestion, @vpinna80! Still not fully convinced, for me the data types are not the same as valuedomains, but the valuedomains are defined on basic scalar types (see: https://sdmx-twg.github.io/vtl/2.1/html/user_manual/model.html#the-variables-and-value-domains-artefacts).

I think this is relevant enough to be worth having a meeting...

[vpinna80]

I hope there will be one session about it in Frankfurt, right @linardian ?

[linardian]

Yes, of course! I am inserting this topic in the agenda.