Stark Arya created TINKERPOP-2308:
-------------------------------------
Summary: IncidentToAdjacentStrategy conflict with
RepeatUnrollStrategy when traversal has pathStep
Key: TINKERPOP-2308
URL: https://issues.apache.org/jira/browse/TINKERPOP-2308
Project: TinkerPop
Issue Type: Improvement
Components: server
Affects Versions: 3.4.2
Reporter: Stark Arya
My dsl like:
{code:java}
g.V("gdb_sample_marko").repeat(__.bothE("gdb_sample_knows",
"gdb_sample_created",
"gdb_sample_create").otherV().hasLabel("gdb_sample_person",
"gdb_sample_software").simplePath()).times(4).hasId("gdb_sample_lop").path().limit(3);{code}
{code:java}
// code placeholder
I call RepeatUnrollStrategy, then EarlyLimitStrategy to optimize the traversal:
protected Traversal.Admin<?, ?> applyStrategy(GraphTraversal<?, ?> traversal) {
TraversalStrategies s = new DefaultTraversalStrategies();
s.addStrategies(RepeatUnrollStrategy.instance(),
EarlyLimitStrategy.instance());
s.toList().forEach(traversalStrategy -> {
System.out.println("before " + traversalStrategy.toString() + "
traversal - " + traversal);
traversalStrategy.apply(traversal.asAdmin());
});
//s.applyStrategies(traversal.asAdmin());
return traversal.asAdmin();
}
{code}
And after RepeatUnrollStrategy finished, the traversal became:
{code:java}
before EarlyLimitStrategy traversal - [GraphStep(vertex,[gdb_sample_marko]),
VertexStep(BOTH,[gdb_sample_knows, gdb_sample_created,
gdb_sample_create],vertex), HasStep([~label.within([gdb_sample_person,
gdb_sample_software])]), PathFilterStep(simple), NoOpBarrierStep(2500),
VertexStep(BOTH,[gdb_sample_knows, gdb_sample_created,
gdb_sample_create],vertex), HasStep([~label.within([gdb_sample_person,
gdb_sample_software])]), PathFilterStep(simple), NoOpBarrierStep(2500),
VertexStep(BOTH,[gdb_sample_knows, gdb_sample_created,
gdb_sample_create],vertex), HasStep([~label.within([gdb_sample_person,
gdb_sample_software])]), PathFilterStep(simple), NoOpBarrierStep(2500),
VertexStep(BOTH,[gdb_sample_knows, gdb_sample_created,
gdb_sample_create],vertex), HasStep([~label.within([gdb_sample_person,
gdb_sample_software])]), PathFilterStep(simple), NoOpBarrierStep(2500),
HasStep([~id.eq(gdb_sample_lop)]), PathStep, RangeGlobalStep(0,3)]
before Gr{code}
At this points , Bug occured: VertexStep and EdgeOtherVertexStep merge into
VertexStep。 so the path result will lost edge info。
Why ?
{code:java}
RepeatUnrollStrategy.apply has this line:
TraversalHelper.applySingleLevelStrategies(traversal, repeatTraversal,
RepeatUnrollStrategy.class);
the code will use traversal.asAdmin().getStrategies() to optimize the following
repeatTraversal:
[VertexStep(BOTH,[gdb_sample_knows,
gdb_sample_created,gdb_sample_create],edge),
EdgeOtherVertexStep,
HasStep([~label.within([gdb_sample_person, gdb_sample_software])]),
PathFilterStep(simple),
RepeatEndStep]
And traversal.asAdmin().getStrategies() list may be:
[ConnectiveStrategy, IncidentToAdjacentStrategy, EarlyLimitStrategy,
MatchPredicateStrategy, FilterRankingStrategy, InlineFilterStrategy,
AdjacentToIncidentStrategy, RepeatUnrollStrategy, CountStrategy,
PathRetractionStrategy, LazyBarrierStrategy]
So IncidentToAdjacentStrategy will be appled before RepeatUnrollStrategy.
so the error occured.
{code}
How to fix?
IncidentToAdjacentStrategy.apply add following slice and result just ok
{code:java}
@Override
123 public void apply(final Traversal.Admin<?, ?> traversal) {
124
125 if ( (!(traversal.getParent() instanceof EmptyStep)) &&
TraversalHelper.hasStepOfAssignableClassRecursively(INVALIDATING_ST EP_CLASSES,
traversal)){
126 return;
127 }
128... origin logic code{code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
