I'd like to match all b1
, where no node b2
exists as depicted below.
This is some kind of a 2-hop asymmetry. Note that the non-existent
b2
must link to the same c:C
as b1
, so other (a2) --> (b3) --> (a1)
can be ignored as long as b3
has no edge to c:C
.
I tried the following...
graph.traversal().V()
.match(
as("c").hasLabel("C"),
as("a").hasLabel("A")
.out().hasLabel("B").as("b1")
.where(out().as("c"))
.out().hasLabel("A").as("a2"),
not(
as("a2")
.out().hasLabel("B").as("b2")
.where(out().as("c")
.out().as("a1")
)
)
...but second MATCH step throws an exception:
java.lang.IllegalStateException: The provided match pattern is unsolvable: [[MatchStartStep(a1), HasStep([~label.eq(A)]), VertexStep(OUT,vertex), HasStep([~label.eq(B)]), WhereTraversalStep([WhereStartStep, VertexStep(OUT,vertex), WhereEndStep(c)])@[b1], VertexStep(OUT,vertex), HasStep([~label.eq(A)]), MatchEndStep(a2)], [MatchStartStep(a2), HasStep([~label.eq(B)]), WhereTraversalStep([NotStep([WhereStartStep, VertexStep(OUT,edge)])]), WhereTraversalStep([WhereStartStep, VertexStep(OUT,vertex)@[c], VertexStep(OUT,vertex), WhereEndStep(a1)])@[b2], MatchEndStep, MatchEndStep]]
I can't even match this is Cypher, unfortunately.
Suddenly, I found a solution which avoids scope problems:
graph.traversal().V()
as("a").hasLabel("A")
.out().hasLabel("B").as("b1")
.where(out().hasLabel("C"))
.out().hasLabel("A").as("a2")
.where(
not(
out().hasLabel("B")
.where(out().hasLabel("C"))
.out().as("a")
)
)