– DRAFT –
SPARQL 1.2 Task Force

Scribe?

Depending on attendance/availability: round table on opinions and criteria

andys: olaf is not present despite his request for the round table, so we will skip it for today

<pfps> sounds good to me

andys: let us look at the tests for queries which should cause syntax errors and get an update from james an peter on testing

<william-vw> +1

Tests of queries that should be syntax errors: ( issue 2 )

andys: introduces issue 2

<AndyS> FILTER EXISTS { BIND ( :e AS ?z ) }

pfps: is that description (quoted from the title) adequate.

pfps: it says nothing about scoping

pfps: we need a clearer description of the issue

pfps: some substitution would be syntactically illegal, but there is another issue, that there can be scoping violations, which is a diffferent issue.

<pfps> One problem is that the current spec can transform something like BIND ( :e AS :f )

<AndyS> https://github.com/afs/SPARQL-exists/blob/main/tests/exists-2-def-var/exists-bad-syntax-01.rq

<pfps> Another problem is that a BIND is illegal if the binding variable is in-scope, so ?z :a :b FILTER EXISTS { BIND ( :c as ?z ) } might be illegal

james: it would help to have an example from the tests. andys suggestede the reference above.

andys: if one were to substitute the bind would be invalid sysntax

<william-vw> ah so the problem is that one can only bind to an unused variable?

james: are we at liberty either to say that this is invalid syntax, or may we change the exists interpretation rules so that it is valid.

andys: we are at liberty to do either because the current interpetation bows up

<AndyS> { :s :p1 ?o BIND("1" AS ?o) }

william-vw: is the restriction that one cannot bind already in scope variables?

pfps: yes.

pfps: there is a type in the spec as what it says about the soping rules is (at leat) confusing if not incorrect.

pfps: a fix for exists may actually make the working correct, but i will put in an issue

<AndyS> https://github.com/afs/SPARQL-exists/blob/main/tests/exists-2-def-var/exists-bad-syntax-04.rq

james: how to demonstrate alternative interpretations

andys: the most important thing is to have tests which demonstrate the spec.

andys: if there are interesting ways that they can be illegal (where these are negative syntax tests) if we decide on a solution for them, then we can reuse these tests.

andys: before we introduce additional tests we should disucss opionoins and criteria.

andys: syntax-4 test has a projection. (which adequately covers thos case?)

pfps: this is illegal, depending on the solution. there has been a proposal to do no injection where the exists runs independently

andys: that solution would change queries which are curently legal.

pfps: agreed, that would not satisfy one of the reasobale criterie (presever existing queries)

andys: the select form (#4) is no different from the bind case

<gb> #4

<AndyS> https://github.com/afs/SPARQL-exists/blob/main/tests/exists-2-def-var/exists-bad-syntax-03.rq

andys: bad-syntax-3 is related in that it could be legal for some interpretations, but for substitution semantics it is invalid syntax.

andys: bound is a special functional form

james: do we collect the alternative results for different implementations?

andys: they _could_ server as informative guidance

<AndyS> https://github.com/afs/SPARQL-exists/blob/main/tests/exists-2-def-var/exists-good-syntax-01.rq

andys: #1 is a minimal query which one might one to get to work.

<gb> CLOSED Issue 1 No activity (nor even README) since WG approval in August (by TallTed)

pfps: that remains fin in any imaginable solution?

andys: in that specific form, yes, but in other combinations it could become invalid

andys: threr was somehting in sql9 where a select queryy could appear in a project position, which should return one rwo.
… : there is not complete consistency across implementation, but there is convergence around lateral.
… : not certain it is in the spec.

andys: the sql spec does define everything bottom up

andys: i wanteed to bring up these as covering the cases.

pfps: i agree, except that ... i have to look again.

andys: the outcomes in the manifest should be treated with skepticism.

Tests process update

james: how to best presnt correlated results. pfps has some results that could be presented
… or use to show results not as test outcomes (yes/no)
… could change EARL reports to also accept inline results for the acceptable criteria to operate in the same way
… https://lists.w3.org/Archives/Public/public-rdf-star-wg/2025May/0031.html
… do the endpoitns pfps is using, allow data to be uploaded?

pfps: fixed graph
… I have been running systems on my server

james: no public GSP?
… you run 3/4 implementation locally?

pfps: yes

james: can we change (extend) the EARL reports?

pfps: want a proper result format - ideally JSON output maybe XML.

(discussion of processing query execution results)

pfps: the implementations do not offer remote sparql services that james' script expected.

pfps: it would be possible to rerun the tests and save the results as json documents.

Topics for next time

andys: next week we can discuss the criteria. will try te get gregory in attendance as well, althoug that is early for him.

andys: it is an important discussion and we need the right people herer.
… othr wise, what test area shoudl we look at?

pfps: will look at qt as the ranswer property. it is just a matter to run everything.

william: what is the process to submit another issue?

william: something about negated property sets in paths. the behaviour is unintutive.

andys: if you think it is errate, we need a test case.

william: i found it confusing enough that it was hard to formulate a test cse.

andys: try an open endpoint to develop the problem description