CODEWITHSUNDEEP
vaticle-typedbvaticle-typeql
nydatamind007
nydatamind007

Reputation: 21

When defining a rule via the Grakn Java API, I get an error when I try and exclude uniary relations

type("overlap-rule").sub(Graql.Token.Type.RULE)
            .`when`(
                and(
                    `var`("t1").isa("trip"),
                    `var`("t2").isa("trip"),
                    `var`("t1").neq(`var`("t2"))
                )
            ).then(
                `var`().isa("trip-overlap")
                    .rel("overlapped-trip", "t1")
                    .rel("overlapped-trip", "t2")

            )

Gives an error: 

match $o isa trip-overlap; get;
Error: UNKNOWN: Invalid state in variable predicate [[$t1 !== $t2]] with answer [[$t1/V4136][$t2/V4136]]: either a concept is missing or not an attribute.. Please check server logs for the stack trace.

But when trying to define the same rule in graql, everything works: 

overlap-rule sub rule,
    when {
        $t1 isa trip;
        $t2 isa trip;
        $t1 != $t2;
    }, then{
        (overlapped-trip: $t1, overlapped-trip: $t2) isa trip-overlap;
    };

ekg> match $o isa trip-overlap; get;
{$o id V81928216 (overlapped-trip: id V4136, overlapped-trip: id V4296) isa trip-overlap;}

Does neq not work as expected?

Upvotes: 1

Views: 44

Answers (1)

flyingsilverfin
flyingsilverfin

Reputation: 450

There are actually two kinds of "not equals" in Graql! There is the != that you used in the graql rule definition, which states that two concepts cannot be the same.

There is another type of not equals, that is written !== in the language itself, representing the comparison of attribute concept values!

The trick is that !== is written using var1.neq(var2) in the Java query builder, whereas != is written using var1.not(var2) in the query builder ;)

Upvotes: 1

Related Questions