Some IRIs not read correctly

[tilgovi]

The code uses a regex to recognize IRIs in the object position during a get. The problem is that this regex does not recognize many IRIs that it should. In particular, an IRI may have any scheme and need not have slashes right after the scheme.

For example, try using "mailto:jane@example.com" in an object and it will interpreted as the @value of an object rather than as an @id (IRI).

[mcollina]

Good catch! Could you do a pull-request that fix this? If not, can you
please point me to the right regexp to use?

[tilgovi]

I'm wondering if something else shouldn't be done. The whole approach of regex seems brittle, to me. It seems like doPut should maybe not lose the type information to begin with. That would break compatibility, though. Although we could have an upgrade path. Since the triple's values are all strings now, we could make them objects and upgrade on the fly. Or something. Not sure how it'd interact with levelgraph, though. I'm very new to these projects.

[tilgovi]

For example, if we "fix" the regex, a literal string value that starts with an IRI would be misinterpreted on the way out.

[mcollina]

First a note about semver: we can 'break' things if we have a good reason too (version < 1.0.0). This project is early in its life and have not seen a production scale usage, yet. I expect API changes and breakages as more people use it. So, let's discuss the best solution for this and then we can discuss an upgrade path.

Second, I think I was too quick implementing put: the toRDF method of jsonld.js outputs all the information we need. Check out http://json-ld.org/playground/index.html and try the 'expanded' or 'nquads' formats. Here it is were we are losing that information: https://github.com/mcollina/levelgraph-jsonld/blob/master/index.js#L39-L47.
So, we can store the additional type information in LevelGraph. All the other keys a part from S-P-O are not indexed and stored as-is: it is also restored when it came back from the database.

Do you fancy wrapping up a pull-request that fixes this?

[tilgovi]

Sure.

[elf-pavlik]

@tilgovi i start working on toRDF() to fix #5 did you already put some work into this issue here?

[elf-pavlik]

i guess we just need to use: http://json-ld.org/spec/latest/json-ld-api/#serialize-rdf-as-json-ld-algorithm
directly from: https://github.com/digitalbazaar/jsonld.js/blob/master/js/jsonld.js#L2882

[elf-pavlik]

since we want to make sure that everyone can access the same graph with LevelGraph-N3 we could maybe also use https://github.com/RubenVerborgh/N3.js#utility

[mcollina]

I'm 👍 for reusing N3.js.

@elf-pavlik if you want the commit bit also for LevelGraph-N3, just ask.

Weird ideas: might it be better if we reuse part of LevelGraph-N3?
How about merging LevelGraph-N3 and LevelGraph-jsonld into levelgraph-rdf?

[elf-pavlik]

@mcollina you read my mind with merging stuff into single levelgraph-rdf 😄

let's take little more time to chew on it and think if we can see any cons of such approach...

[mcollina]

The only con I see is library size if we want to use it in the browser.
I would love to put LevelGraph on a diet, but I am not sure where to start.

[elf-pavlik]

#10 should fix it... i still need to add relevant tests!