SHACL Targets

44 views
Skip to first unread message

Matt Goldberg

unread,
Nov 2, 2020, 8:07:44 PM11/2/20
to TopBraid Suite Users
Hello-

I've experimented with the Custom Targets and Custom Target Types and I can't seem to find a reasonable way to define targets in a particular way. What I'd like to do is have the focus nodes specified by resources themselves via a property that points to the shape to use, effectively how rdf:type does for implicit shapes, but by using a specified term in the domain ontology instead of rdf:type. In other words, I'd like to specify that the focus nodes for some node shape S are the subjects of triples with predicate P and object S, where P is specified and is not necessarily rdf:type. 

sh:targetSubjectsOf and sh:targetObjectsOf do not provide this functionality, as they only look at the predicate. I don't think $currentShape is pre-bound for custom target types like for constraint components (at least my experiments didn't seem to work), but even if it is, it would likely be an optional feature as is stated in the specification. An ideal solution would be something that would not use any TopBraid specific features, as we are using SHACL in other systems as well.

Thanks for any help!

Matt Goldberg

unread,
Nov 2, 2020, 8:20:06 PM11/2/20
to TopBraid Suite Users
Left out a detail- an ideal solution would also avoid sh:node such that errors in validation reports would be informative instead of sh:node's uninformative "does not conform to shape" message.

Thanks for any help! 

Irene Polikoff

unread,
Nov 2, 2020, 10:44:27 PM11/2/20
to topbrai...@googlegroups.com
Hi Matt,

There are pre-defined target types in SHACL. You can not change their behavior.

There is also a way to create SPARQL based targets using a SPARQL query. If a query is a pattern that may be re-used in different contexts, you could declare a custom type by parametrizing the query as described here https://w3c.github.io/shacl/shacl-af/#SPARQLTargetType

This would be your own type of target, using your namespace, not sh:.

The query, for example, could have predicate and object as parameters:

ex:MyTarget
a sh:SPARQLTargetType ;
rdfs:subClassOf sh:Target ;
sh:parameter [
sh:path ex:predicate ;
sh:nodeKind sh:IRI ;
] ;
      sh:parameter [
sh:path ex:object ;
sh:nodeKind sh:IRI ;
] ;
sh:prefixes ex: ;
sh:select """
SELECT ?this
WHERE {
?this $predicate $object .
}
""" .

Then, when you assign this custom target type, you would provide values for the predicate and object. I have assumed above that objects are resources.

This is just a quickly sketched example. I have not tried it. Read the spec for more details.

--
You received this message because you are subscribed to the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/4cc9fc07-6309-4c53-aaee-7073dd62ac2cn%40googlegroups.com.

Matt Goldberg

unread,
Nov 2, 2020, 11:01:16 PM11/2/20
to TopBraid Suite Users
Hello-

Thanks for the reply. I have considered doing exactly this, and may end up trying this further. My reservations with this method are that this requires Advanced Features which may not be supported by all SHACL engines, and with every usage of this Target Type the shape must specify itself as a parameter and I was hoping there would be a way to make it a bit smarter to avoid doing that. However, if that's the best option, I'll experiment with it.

Holger Knublauch

unread,
Nov 3, 2020, 2:06:55 AM11/3/20
to topbrai...@googlegroups.com

Just to add I don't see a solution to your specific issue either. Custom target types can only access their own parameters, not the context shape or the properties of that.

Holger

Matt Goldberg

unread,
Jul 21, 2021, 11:11:23 AMJul 21
to TopBraid Suite Users
Hello-

I'm continuing this thread to propose a potential feature to add to some future version of SHACL related to this discussion. This feature would be a special type of target that would enable additional focus nodes for a shape to be specified by a property in a domain ontology that connects the focus node to the shape. This could be configured such that the subject would be a focus node and the object would be the shape, or such that the subject is the shape and the object is a focus node, or  perhaps using a SHACL path which would enable more complex behavior than just a single predicate. I realize that the second case is like sh:targetNode, except the idea is to enable an existing domain property to have similar functionality to sh:targetNode. This would be useful for domains in which data may contain the specification of data requirements. As discussed earlier, there is currently no easy way to do this.

For example, consider this function ontology. If you look at the descriptions for fno:Parameter and fno:Output they look very similar to sh:PropertyShape (or perhaps sh:Parameter) in spirit, and the class fno:Function is therefore like sh:NodeShape (or perhaps sh:Function). If these classes were additionally modeled as subclasses of the corresponding SHACL class, then the focus nodes of an instance of fno:Function (and sh:NodeShape) would be the instances of fno:Execution connected to it via fno:executes. Therefore, it would be convenient to be able to create a target that effectively says "for all triples with fno:executes as predicate, the subject is a focus node of the object".

I realize that this is targeting behavior is different than the other targets; shapes effectively say what the focus nodes for that shape are with existing targets while this is a target that could apply to multiple shapes. However, I think it could enable SHACL to be a lot more flexible, help enable simpler integration with existing ontologies (especially those in which data requirements are part of the domain), and help reduce the amount of metamodeling required to get extra SHACL functionality added to existing models.

Matt Goldberg

Holger Knublauch

unread,
Jul 21, 2021, 11:38:06 PMJul 21
to topbrai...@googlegroups.com

Hi Matt,

sounds interesting. Do you have a draft spec for this someplace, or a worked out example?

One problem or design constraint with SHACL targets is that they should be executable in two modes:

    1) for a given shape, find all target nodes

    2) for a given node, find all shapes that target it

If a constraint types makes computing 2) hard then tools will not be efficiently able to validate a given instance, e.g. after a user has made changes on a form.

I am not saying this applies here but wanted to put this out as a thought to consider.

Holger

Matt Goldberg

unread,
Aug 1, 2021, 12:58:35 PMAug 1
to topbrai...@googlegroups.com

I've finally had some time to think about this. Another potential use case for this would be some future state of the W3C Data Cube ontology. Data Structure Definitions could be changed to be represented as Node Shapes, and the focus nodes of one of these would be the Observations that are part of the DataSet(s) that have that Data Structure Definition. This would require the use of a path instead of a single predicate.

I've thought of two possible implementations. The first is the creation of a new Target Type. Here's a (non-functional) prototype:

sh:PathFromShapeTarget
a sh:TargetType ;

rdfs:subClassOf sh:Target ;
sh:parameter [
    sh:path sh:path ;
sh:description "The path connecting one or more shapes to their focus nodes" ;
] ;
sh:select """
SELECT ?this ?currentShape
WHERE {
?shape $PATH ?this
FILTER EXISTS {?shape a/rdfs:subClassOf sh:NodeShape} # This may or may not be necessary
}
""" ;
.

Targets then take a SHACL path to identify the path desired and injects it in the SPARQL query like the path in a sh:SPARQLSelectValidator(if implemented in SPARQL). So for the example function ontology:

fno:FunctionTarget
a sh:PathFromShapeTarget ;
sh:path [
sh:inversePath fno:executes ;
] ;
.

Or for the Data Cube example:

qb:DataStructureDefinitionTarget
a sh:PathFromShapeTarget ;
sh:path [
sh:inversePath (
qb:dataSet
qb:structure
) ;
] ;
.

Now, I realize that following the current behavior of Node Shapes and Targets, each and every fno:Function or qb:DataStructureDefinition would require an extra triple connecting it to the above example Target via sh:target. While that could work, it feels like that the Target itself really captures the meaning that the specified path connects shapes to focus nodes independently of the specific shape and that requiring that extra triple to exist every time feels redundant.

Therefore, it would be nice if it were possible to enable that functionality, which is why I added the ?currentShape variable in the query; either a shape could be bound to ?currentShape get focus nodes, or a (potential) focus node could be bound to ?this to obtain the shapes that apply to that node via the path.

A second possible implementation would be to create a new Constraint Component that functions like sh:node, using a property perhaps called sh:nodePath. However, instead of specifying the URI of a Node Shape that focus nodes must also conform to, it specifies a SHACL path pointing to Node Shape(s) that focus nodes must also conform to.

This would enable the following additions for the function ontology:

fno:Execution
sh:nodePath fno:executes ;
.

Or these additions for the Data Cube ontology:

qb:Observation
sh:nodePath (
qb:dataSet
qb:structure
) ;
.

This approach seems a bit cleaner, more practical, and it has the benefit of applying to all instances of qb:Observation regardless of the specific Data Structure Definition relevant to a given Observation. Note that this doesn't apply a targeting rule based on a path independently; if there were multiple classes/shapes that would also conform to a shape at the same path, that would have to be expressed multiple times unlike the first approach (potentially). That's not a dealbreaker though, just a comment.

My main reservation with this approach is that I'm not a huge fan of how if sh:node fails validation, the error message states just that validation failed and not why it failed (like how the original SHACL Playground example says "Value does not have shape schema:AddressShape" instead of the actual error message "Value is not >= 10000"). The sh:node error messaging is not super helpful, and I would hope that the error messaging for this feature would show the expected helpful messages (and that in the future the error messages from sh:node could be propagated through to the final report as well).

I haven't had the time to dig into SHACL validator implementation details yet, so I'm not sure how feasible either of these options are to actually implement in a current SHACL validator. I'm curious to see what you think.


Virus-free. www.avg.com

You received this message because you are subscribed to a topic in the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/topbraid-users/JS6jfJikuBk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/a879b561-4cf5-2104-3dee-b763b6e6ce49%40topquadrant.com.

Virus-free. www.avg.com

Holger Knublauch

unread,
Aug 4, 2021, 1:25:24 AMAug 4
to topbrai...@googlegroups.com

Hi Matt,

without understanding all consequences and requirements, some comments below.

This approach would stretch the current use of target types quite a bit. For example, sh:parameter values cannot easily be "deep" blank node structures, but only individual nodes. Property shape validators use $PATH as a special, hard-coded parameter to avoid such complications. Before such a pseudo-generic mechanism is added only to cover this particular use case, I think it would be cleaner to introduce a proper target type for a SHACL 1.1 Core, as below.

A second possible implementation would be to create a new Constraint Component that functions like sh:node, using a property perhaps called sh:nodePath. However, instead of specifying the URI of a Node Shape that focus nodes must also conform to, it specifies a SHACL path pointing to Node Shape(s) that focus nodes must also conform to.

This would enable the following additions for the function ontology:

fno:Execution
  sh:nodePath fno:executes ;
.

Or these additions for the Data Cube ontology:

qb:Observation
  sh:nodePath (
    qb:dataSet
    qb:structure
  ) ;
.

This approach seems a bit cleaner, more practical, and it has the benefit of applying to all instances of qb:Observation regardless of the specific Data Structure Definition relevant to a given Observation. Note that this doesn't apply a targeting rule based on a path independently; if there were multiple classes/shapes that would also conform to a shape at the same path, that would have to be expressed multiple times unlike the first approach (potentially). That's not a dealbreaker though, just a comment.

My main reservation with this approach is that I'm not a huge fan of how if sh:node fails validation, the error message states just that validation failed and not why it failed (like how the original SHACL Playground example says "Value does not have shape schema:AddressShape" instead of the actual error message "Value is not >= 10000"). The sh:node error messaging is not super helpful, and I would hope that the error messaging for this feature would show the expected helpful messages (and that in the future the error messages from sh:node could be propagated through to the final report as well).

The property sh:detail was introduced to capture those nested violations. The SHACL API has a flag to activate those nested violations, yet the SHACL Playground is just that - a simple playground. Don't expect it to be the ultimate truth.

I haven't had the time to dig into SHACL validator implementation details yet, so I'm not sure how feasible either of these options are to actually implement in a current SHACL validator. I'm curious to see what you think.

Moving forward, you may elect to open a ticket similar to https://github.com/w3c/data-shapes/issues/137 with a proposal on what is missing. This is the place to talk about future versions of SHACL where such requirements could be addressed better. Without at least some basic acknowledgement by people outside of TopQuadrant (so that other implementations would understand such extensions), or show-stopping use cases by paying customers, it would be hard for me to spend much time on such extensions on my own.

Holger


Reply all
Reply to author
Forward
0 new messages