ShapeMap specification candidate draft

[jimkont]

As decided in the last teleconference, all CG members are kindly requested to review and vote on the ShapeMap specification that Eric drafted here: https://shexspec.github.io/ShapeMap/

if no issues are identified we will publish it along with the rest documents we published in our community page https://www.w3.org/community/shex/

Please review by Monday evening US time

[labra]

It would be nice if we could add some convention to include also prefixed names in shape maps. Something like:

{ FOCUS rdf:type schema:Person }@START

which would be equivalent to

{ FOCUS <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://schema.org/Person> }@START

whenever there is some convention to find those prefix declarations in the data or the schema

[gkellogg]

Some grammar comments:

• shapeMap ends with a ';' which is not quoted; does this have some significance?
• The triplePattern production has a First/First conflict on '{' (a concern for context-free parsing), I rewrite as follows:

  [6]     triplePattern         ::= '{' (
                                        ("FOCUS" iri (objectTerm | '_'))
                                      | ((subjectTerm | '_') iri "FOCUS")
                                    ) '}'
  

[ericprud]

I added a triplePatternLL production which folks can compare with triplePattern (just above it). My pref is to favor the readability of triplePattern but can go either way.

[lucaswerkmeister]

Some grammar comments:

• shapeSelector (production 7) references ATSTART, but Terminals defines AT_START (production 17).
• The terminals reference two terminals that are not defined: PNAME_NS (production 71 and 15) and PNAME_LN (production 16).
• The anchor for the PNAME_NS href in production 71 is #term-PNAME_NS, while the anchor for the same terminal in production in production 15 is #term-PN_PNAME_NS (extra PN_). (The anchor for the PNAME_LN href in production 16 also has the extra PN_, but PNAME_LN is not referenced anywhere else.)
• AT_PNAME_NS and AT_PNAME_LN (production 15 and 16) appear unused. Are they unintentional duplicates of ATPNAME_NS and ATPNAME_LN (production 70 and 71)?

Also, the “ShapeMap usage” section feels incomplete to me. It describes how to convert a triple pattern to a SPARQL triple pattern, but not how to derive a fixed ShapeMap from it. To turn the SPARQL triple pattern into a full SPARQL query, I would think that you’d have to define the variable name used for a focus selector more specifically (instead of just a fresh variable), so that it can be SELECTed.

[ericprud]

@lucaswerkmeister , The terminals you pointed out were indeed vestiges of a larger language with support for prefixed names. I've commented them out in the editor's draft.

Re the usage section, I had failed to complete the thought. I added this to the section on fixing triple patterns:

For each triple T in G which matches P, the fixed ShapeMap has a shape association F where F.shape = A.shape and • If A's subject is a focus selector, F.node is T's subject. • If A's object is a focus selector, F.node is T's object.

I also changed

If A.node is an RDF graph, A is in the fixed ShapeMap.

to

If A.node is an RDF node, A is in the fixed ShapeMap.

You can add a laugh icon to your comment to signify that the above two commits address your comments. Tx for the careful review!

P.S. I'm happy to adopt another convention to indicate that a comment had been satisfied. I picked the laugh icon because it's kind of like a smiley face indicating contentment; also+1 would imply support for the issue itself rather than indicating that it has been addressed (I guess).