Integrate SADL List grammar with Jena builtins

[crapo]

See RuleWithList.sadl in TestSadl3Ide example project.

[crapo]

@agabaldon , @AbhaMoitra , I would like to make some changes to lists in SADL. I'm not sure how much you have used this recently implemented functionality or set of built-in functions, so I'm bringing it up for discussion before committing changes.

1. It was a mistake to call the JenaReasonerPlugin's list built-in "length" as that is a term that is a) a keyword in SADL, and b) a common term that modelers might like to be able to use in SADL without having to always prefix its (QName) because it is in the SadlBuiltinModel (at least for Jena-based reasoners). Rather it should be named "listLength", like the Jena-provided built-in for RDF lists.
2. The "listLength" built-in function can easily handle both SADL typed lists and RDF lists. In fact, all of the list built-ins could easily be enhanced to handle either type of list.
3. There is a bigger potential change. I've probably gotten carried away with the imagined importance of typed lists. The real value of typed lists, presumably, is just that--they are typed, enabling type checking of their content. There is no reason that the SADL typed list needs to have its own "first" and "rest" properties; it could use the rdf:first and rdf:rest properties and simply add the property restrictions on these properties when the list is typed. It may be possible to tweak the SADL grammar to allow regular RDF lists to be created in SADL--that would require some experimentation to determine. If that is possible then it would allow a modeler to create lists in SADL, as well as reason on any kind of list, that contain a mixture of any kind of items--numbers, instances of different classes, etc.

Thoughts? I will not make changes before hearing from you on this topic.

[AbhaMoitra]

@crapo : I am ok with these changes and they make sense.

@agabaldon : can you also take a look at this and share your views.

[AbhaMoitra]

@crapo : I am trying to use 'length of a list" and I am not able to get it working. Maybe my syntax is in correct? I am using

and len = builtinfunctions:^length(n2)

I also tried

and len = ^length(n2)

[crapo]

@AbhaMoitra , in my test case I use:
Rule ListLength:
if l1 is a YooHooList and len is length of l1 then print("Length of l1: ", len).

which translates to the Jena rule:
[ListLength: (http://sadl.org/UsingListExpression.sadl#l1 rdf:type http://sadl.org/UsingListExpression.sadl#YooHooList), listLength(http://sadl.org/UsingListExpression.sadl#l1, ?len) -> print('Length of l1: '^^http://www.w3.org/2001/XMLSchema#string, ?len)]

but this is on a branch of the code where I have changed the name of the built-in to listLength. I'm not sure I have a test case where built-in "length" was used as an explicit function rather than with the SADL grammar.

[agabaldon]

@crapo: the change to listLength makes sense. I think the flexibility you describe in (3) would also be good to have in the future.

What branch should I install to test listLength?

[crapo]

@agabaldon , I just created a PR on DefaultValues branch that has both default values and new "listLength".

[AbhaMoitra]

@crapo : Here is my rule - rewritten to use your syntax style above (I am using https://github.com/ml4ai/grasen-data/blob/master/SemanticModels/GraSEN

Rule MeasuredSignalCandidate // error signal = reference signal - measured signal
if node is a Function
and nodes of node is n1
and children of n1 is n2
and len is ^length of n2
then measuredSignalCandidate of node is true
and print("in MeasuredSignalCandidate, node is ", node).

In the Rules file, the syntax for "length" clause is

(?n2 http://sadl.org/builtinfunctions#length ?len)

Is that what you expect?

[crapo]

@AbhaMoitra , that syntax indicates that the translation is treating "length" as a property.

I'll restate my rule as I would expect it if the function name were "length" rather than "listLength".

[ListLength: (http://sadl.org/UsingListExpression.sadl#l1 rdf:type http://sadl.org/UsingListExpression.sadl#YooHooList), length(http://sadl.org/UsingListExpression.sadl#l1, ?len) -> print('Length of l1: '^^http://www.w3.org/2001/XMLSchema#string, ?len)]

For your rule, I'd expect:

... length(http://sadl.org/UsingListExpression.sadl#l1, ?len)...

I'm not sure what version of the code you are using. However, I see that Alfredo just merged the latest PR, which uses "listLength" as the Jena built-in function. Perhaps you should get the latest version of the SADL plugins that will be created from this merge into the development branch.

Note that you should delete the SadlBuiltinFunctions.sadl file from the ImplicitModel folder and let it be replaced with a new one so that function names are updated.

[agabaldon]

@crapo: after pulling the most recent changes, using listLength is not working but length is. Are we missing some update?
Also, I see some message about "Derivations written to..." Should we turn some setting off?

[crapo]

@agabaldon , I'm going to guess that this is because the old SadlBuiltinFunctions.sadl has not been deleted and still contains "length" rather than "listLength". It is also possible that the configuration.rdf file, which contains the fully qualified names of built-in function classes for Jena-based rules might be causing an error message. Currently the SadlBuiltinFunctions.sadl file is replaced when reasoner and/or translator selection is changed. I don't see any downside to deleting it as part of a clean as build will replace it with a new instance freshly created by the selected translator/reasoner pair.

The "Derivations written... " is a regression bug. I will fix it.

[crapo]

@agabaldon , @AbhaMoitra , just created a PR that addresses this and several other issues.