Re: [rdf4j-dev] [GSOC 2017] #743 -Add support for Shapes Constraint Lang

[Håvard Ottestad <hmottestad@xxxxxxxxx>]

I think having the SHACL rules immutable will help with isolation and concurrency. 

The plan (from Heshan) is to do the work bit by bit, with getting some structure up first and then doing minCount, first with insertion and then with update and deletion. I have som ideas for how to handle everything, but it's the development method here that's going to be the key to solving it. 

Håvard

On Mon, Apr 10, 2017 at 1:09 AM, Jeen Broekstra <jeen.broekstra@xxxxxxxxx> wrote:

> On 9 Apr 2017, at 22:26, Håvard Ottestad <hmottestad@xxxxxxxxx> wrote:
>
> I'll be available Monday to Wednesday this coming week. From 5 pm Oslo time. After that it's easter break in Norway, so the next time I can meet you will be on Wednesday.
>
> You said you also had some holidays in Sri Lanka, so I propose we try for a meeting on the 19th, 20th or 21st after 5pm. One of those days I might end up being busy in the evening due to it being the final week for a member of our project team and we're going out for a dinner. Not sure which day it's going to be though.
>
> *** To the RDF4J developers *** if any of you have any input as to the architecture of the SHACL module it would be great if you could write by then, or join the hangouts call (let me know and I'll give you a link). As outlined in the proposition; Heshan is trying for a DB style architecture with an AST and "query" planner with a subsequent optimisation step, all for the purpose of limiting the number of rules that need to be run and the amount of data to be fetched from the underlying store.

I have little to add at this point beyond that I think it looks like a solid plan. I haven’t quite gotten my head around how it will handle incremental updates (and, more challenging, deletions), but I assume you’ve given it thought. The main difficulty will probably be how to correctly deal with isolation and concurrency-related issues. The compliance tests should help pinpoint that part of it. Anyway, I’m probably getting ahead of myself here.

Cheers,

Jeen

_______________________________________________
rdf4j-dev mailing list
rdf4j-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/rdf4j-dev