New exported format: JSON-LD metadata to explain the CSVs, using W3C Tabular Data (Basic implementation only)

[fititnt]

• CSVW Namespace Vocabulary Terms
  • https://www.w3.org/ns/csvw
    • https://www.w3.org/ns/csvw.jsonld
• Metadata Vocabulary for Tabular Data
  • https://www.w3.org/TR/tabular-metadata/
• Model for Tabular Data and Metadata on the Web
  • http://www.w3.org/TR/tabular-data-model/#standard-file-metadata
• CLDF - Cross-Linguistic Data Formats
  • https://cldf.clld.org/
  • https://github.com/clld/clld
  • https://clld.org/docs/reflex/clld.pdf
  • Examples
    • https://dictionaria.clld.org/download
      • Direct link, example with several tables
        • https://github.com/dictionaria/daakaka/blob/master/cldf/cldf-metadata.json
• Related issues
  • New exported format "tabular data" on JSON, use case Wikimedia Commons New exported format "tabular data" on JSON, use case Wikimedia Commons #26
  • New exported format: frictionlessdata Tabular Data Package + Data Package Catalogs New exported format: frictionlessdata Tabular Data Package + Data Package Catalogs #35

---

We already export CSVs which can be explained by more than one convention. The more popular one is frictionlessdata mentioned on #35, but W3C, while started based on frictionlessdata, diverged to be compatible with RDF in a way similar to what happened to JSON-LD (which also allow to be mapped directly to anything supported by RDFs).

Basic implementation only

The W3C Tabular Data is obviously more complex to implement than frictionlessdata, but in case things get complicated, this will focus on basic implementation only. For example, ff we share in similar way the CLLD do (https://github.com/dictionaria/daakaka/blob/master/cldf/cldf-metadata.json) the "package" would require each group of dictionaries also export the subset of other tables it uses (in a way that someone could import the full think in a database).

Note that maybe already is possible to convert frictionlessdata to W3C Tabular Data, so we may implement at least one way to do it

About catalog of all datasets: use different approach (out of this issue)

• https://github.com/ckan/ckanext-dcat
    - https://www.w3.org/TR/vocab-dcat/

This would be better move to a different issue, but while frictionlessdata have some way to organize more than one dataset (a catalog, https://specs.frictionlessdata.io/patterns/#describing-data-package-catalogs-using-the-data-package-format), which I'm not sure if this is a new feature) there is no direct equivalent with same strategy of this issue.

The close would more likely emulate what CKAN does. At https://github.com/ckan/ckanext-dcat there are some drafts about this. However, CKAN implies there is some way to at least paginate the global list (which is obviously not viable using static generation). In the best case, I think either pushing data to a CKAN (or making a fetch for some static version of all datasets) could be one way to synchronize dictionaries.