Feature-completeness of SHACL & SPARQL

[rmfranken]

Hi there,

thanks for your great work, I'm always happy to see more and fasters SHACL or SPARQL engine implementations. Can you confirm which SHACL and SPARQL features are currently covered? Perhaps in a similar manner to how PyShacl did this - or point me to some documentation if I missed it?

Looking forward to the support for sh:rules as I am using those heavily in my ontologies!

Kind regards,
Robin

[magbak]

Hi Robin!
Thank you for reaching out!

Currently we support almost all of https://www.w3.org/TR/shacl/
What is missing is the SPARQL Constraint Components (we have SPARQL constraint support, only the components are missing) and Non-Validating Property Shape Characteristics.
From SHACL Advanced we only have SPARQL Targets.

SHACL Rules are in progress - we have planned to support it this fall. It would be really helpful to us if you would like to be a pilot user for SHACL Rules, that would really help us prioritize features and get some feedback on performance and usability.

If that sounds interesting please reach out at magnus at data-treehouse.com

Best regards,
Magnus

[magbak]

Hi again,
Noted that you asked about SPARQL Support also.

Most of it is there, but the most notable missing bits from SPARQL 1.1 are:

• Named Graph support is partial (full support ETA. end of June)
• The parts of SPARQL Update that use static data
• Service
• Minus support is partial

These can be supported quite easily.

We are missing some operations on durations as well as comparisons between datetimes where only one datetime has a timezone, otherwise operations are implemented.

The following functions are missing - they can mostly be easily implemented, reach out if you need them:

• sameTerm
• BNODE
• STRDT
• ENCODE_FOR_URI
• UCASE
• LCASE
• SUBSTR
• RAND
• now
• timezone
• tz
• hash functions (sha1, md5)
• xpath

Best regards,
Magnus

[rmfranken]

Hi Magnus, Thanks for your reply. This gives me enough information to want to indeed test out your stuff - especially the SHACL engine. I'll let you know what I find! The SHACL rules are currently mostly used in projects with pretty low data volumes, so the performance is less essential.
As far as SPARQL goes, there we do indeed use named graphs quite a bit, and the BNODE function.

I think for now, a CLI would be very useful, most devs I know are using CLI tools for doing SHACL validation, or sending requests to some API. I see you have some API documentation, so I will get started there to try to run your tool.

[rmfranken]

Hello Magnus, sorry for the long break - I only got around to trying this now.

I'm a little bit confused on how to use your tooling. (note, I am primarily a semantic engineer so I don't have that much experience around getting tools to run in an API, but would like to explore that, hence why I'm here :) )

I've tried to create a quick python fast-api to wrap your maplib library, and am running into errors with the fact that maplib seems to be expecting some kind of "stottr" document? (I'm feeding it turtle)

exceptions.MaplibException: Template parsing error: Could not parse entire string as sttotr document,

I've never heard of that rdf serialization, it seems to be https://dev.spec.ottr.xyz/stOTTR/ - but I don't know if I'm willing to dive into the specifics of what seems to be a pretty obscure standard just to use your tooling.

Then, for your validate funciton I would usually expect:

shapes graph , data graph - this should be the bare minimum that any SHACL validation requires. PyShacl of course has a bunch of extra options, but the TopBraid SHACL API pretty much only has those two as mandatory. In your docs I also see

Validate the contained knowledge graph using SHACL Assumes that the contained knowledge graph also contains SHACL Shapes.

I think that's a strong assumption, and not necessarily always the case. I would put the default the other way around:
if there is a data and shapes graph - use those, unless there is only a data graph, then assume that the shapes are in the data graph or the other way around.

Just some points of feedback - I'll plod on a little more trying to understand what I'm doing wrong.