[
https://issues.apache.org/jira/browse/TINKERPOP-3009?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17781895#comment-17781895
]
ASF GitHub Bot commented on TINKERPOP-3009:
-------------------------------------------
spmallette opened a new pull request, #2322:
URL: https://github.com/apache/tinkerpop/pull/2322
https://issues.apache.org/jira/browse/TINKERPOP-3009
Note that TraverserGenerators changed a bit because of the filter patterns
that were added, thus the assertion adjustments. Basically, rather than adding
and(inV()..., outV()...) for every single edge step it's done more
strategically such that only the direction not already traversed gets filtered.
Also avoids some retesting of vertices if they were already handled.
Before:
{code}
gremlin> g.withStrategies(SubgraphStrategy.build().
......1> vertices(hasLabel('x','y','z')).create()).
......2> V().out().profile()
==>Traversal Metrics
Step Count
Traversers Time (ms) % Dur
=============================================================================================================
TinkerGraphStep(vertex,[~label.within([x, y, z])])
0.160 58.50
VertexStep(OUT,edge)
0.017 6.53
AndStep([[EdgeVertexStep(IN), HasStep([~label.w...
0.017 6.19
EdgeVertexStep(IN)
0.000
HasStep([~label.within([x, y, z])])
0.000
EdgeVertexStep(OUT)
0.000
HasStep([~label.within([x, y, z])])
0.000
EdgeVertexStep(IN)
0.064 23.59
HasStep([~label.within([x, y, z])])
{code}
and after:
{code}
gremlin> g.withStrategies(new SubgraphStrategy(vertices: __.has("name",
P.within("josh", "lop", "ripple")))).V().out().profile()
==>Traversal Metrics
Step Count
Traversers Time (ms) % Dur
=============================================================================================================
TinkerGraphStep(vertex,[name.within([josh, lop,... 3
3 8.231 95.09
VertexStep(OUT,vertex) 2
2 0.337 3.90
HasStep([name.within([josh, lop, ripple])]) 2
2 0.087 1.01
>TOTAL -
- 8.656 -
{code}
> SubgraphStrategy produces excessive filtering when multiple labels are
> filtered upon
> ------------------------------------------------------------------------------------
>
> Key: TINKERPOP-3009
> URL: https://issues.apache.org/jira/browse/TINKERPOP-3009
> Project: TinkerPop
> Issue Type: Bug
> Components: process
> Affects Versions: 3.6.2
> Reporter: Brian O'Keefe
> Priority: Critical
>
> See the example profile below. Notice the traversal includes _has _filters
> everywhere repeatedly applied for vertices which are already filtered before
> too.
> {noformat}
> gremlin> g.withStrategies(SubgraphStrategy.build().
> ......1> vertices(hasLabel('x','y','z')).create()).
> ......2> V().out().profile()
> ==>Traversal Metrics
> Step Count
> Traversers Time (ms) % Dur
> =============================================================================================================
> TinkerGraphStep(vertex,[~label.within([x, y, z])])
> 0.160 58.50
> VertexStep(OUT,edge)
> 0.017 6.53
> AndStep([[EdgeVertexStep(IN), HasStep([~label.w...
> 0.017 6.19
> EdgeVertexStep(IN)
> 0.000
> HasStep([~label.within([x, y, z])])
> 0.000
> EdgeVertexStep(OUT)
> 0.000
> HasStep([~label.within([x, y, z])])
> 0.000
> EdgeVertexStep(IN)
> 0.064 23.59
> HasStep([~label.within([x, y, z])])
> 0.014 5.18
> >TOTAL -
> - 0.274 -
> {noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
