How to replace a Dataset object with m-ld?

[danielbeeke]

I wonder how to swap a Dataset object with m-ld. I have created: https://shacl-form.mediaworks.global/0
I want to make the form collaborative.

It works with a Dataset that conforms to the RDF/JS spec.
I now wonder could I just replace the dataset and would it automatically work? Or do I also need to manually do read and writes on meld?

[gsvarovsky]

Hi Daniel, welcome!

A Javascript m-ld 'clone' implements RDF/JS Source, with additional methods for SPARQL queries:
https://js.m-ld.org/interfaces/queryablerdfsource.html. Of course, first you have to create the clone (a local dataset), to get this interface.

At the moment writes have to be done using the native JSON-LD-based format (json-rql). It's JSON-LD with delete/insert keywords. If you'd like to write triples, we could add that pretty easily – in fact it's possible we might want that ourselves, soon.

We're working on some much improved getting-started materials right now. You can preview here: https://m-ld-js-git-getting-started-overhaul-gsvarovsky.vercel.app/. Doesn't talk about RDF yet, but I would be happy to help join the dots.

[gsvarovsky]

OK nice, I'll experiment with the Store-callback API 👍

[gsvarovsky]

We will also need to think about what to do with some other features, for which the data representation in the json-rql API is abstracted away from RDF, and would need more thought. These are Lists, Binary Data and Collaborative Text. I think your form had examples of all three! For now perhaps let's just find out which ones you want to use.

Lists in m-ld, using the @list JSON-LD keyword, do not transform to RDF Collections. They are more like RDF Containers (but still, not quite 🙂). Plain RDF Collections don't converge nicely with concurrent edits – more about this in our Semantics poster. If you want to use lists we'll have to consider what to do here.

Binary data is not a big deal because it's only an efficiency gain, you can always use xs:base64Binary, and we can probably easily create a custom data factory to even take away the need for the API client to encode anything. Or you may want to handle attachments completely separately anyway.

Collaborative text will probably be useful for description fields, so multiple users can edit the field at the same time. If we want to use this feature it's not going to be pure RDF any more.

[gsvarovsky]

• RDF/JS Sink & Store interface support m-ld-js#81

[danielbeeke]

Lists in m-ld, using the @list JSON-LD keyword, do not transform to RDF Collections. They are more like RDF Containers (but still, not quite 🙂). Plain RDF Collections don't converge nicely with concurrent edits – more about this in our Semantics poster. If you want to use lists we'll have to consider what to do here.

I read the semantics poster.
I can adapt the system to also support the structure m-ld is using. Another idea could be to abstract lists away. Do I see it correct that I could propagate list changes in m-ld specific format, but keep them in my own dataset as a rdf list?

Yes attachments should be IRIs and then on the storage end a cron task should run to clean up unused attachments.

Super cool stuff :)

[gsvarovsky]

• List representation in RDF/JS m-ld-js#172