[
https://issues.apache.org/jira/browse/TINKERPOP3-679?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15008935#comment-15008935
]
Marko A. Rodriguez commented on TINKERPOP3-679:
-----------------------------------------------
Perhaps this is what an {{ExplainStep}} could do. I'm not too versed in what
{{EXPLAIN}} does in other query languages, but I could see it as not a "real
step" but a "terminal/action step" that looks like this:
{code}
g.V().outE("knows").inE().count().explain()
{code}
This would return a {{TraversalExplain}} object much like [~rjbriody]'s
{{TraversalMetrics}} object. The API for {{TraversalExplain}} could be:
{code}
TraversalExplain.getStrategies() // the sorted list of strategies that were
applied
TraversalExplain.getTraversal(int strategy) // the Traversal after being
strategy i was applied, where getTraversal(0) is the user constructed
concatenated traversal.
TraversalExplain.getTraversals() // the sorted list of traversals after each
strategy was applied.
{code}
You could then do this:
{code}
gremlin> g.V().outE("knows").inE().count().explain().getTraversals()
==>[NoStrategy,[GraphStep, VertexStep[edges,OUT,knows],
VertexEdgeStep[IN],GlobalCountStep]]
==>[AdjacentIncidentStrategy,[GraphStep, VertexStep[vertices,OUT,knows],
GlobalCountStep]]
==>[TinkerGraphStepStepStrategy,[TinkerGraphStep,
VertexStep[vertices,OUT,knows], GlobalCountStep]]
{code}
We could probably make a {{toString()}} which is basically the data in
{{getTraversals}} but all pretty like {{ProfileStep}}. This seems pretty easy
to do actually (a morning session to have a fully functional example).
> Debuggable Traversal
> --------------------
>
> Key: TINKERPOP3-679
> URL: https://issues.apache.org/jira/browse/TINKERPOP3-679
> Project: TinkerPop 3
> Issue Type: Improvement
> Components: process
> Affects Versions: 3.0.2-incubating
> Reporter: stephen mallette
>
> I'd like to be able to put a {{Traversal}} into "debug mode".
> {{TraversalStrategy}} application is a bit too mysterious, making it hard to
> figure out why a {{Traversal}} is failing. In a recent debug session, the
> problem was only sorted out after we looked at the order in which strategies
> were applied and even then it was a bit tricky to really understand what was
> happening.
> If the {{Traversal}} were in this "debug mode" it could give us insight into
> that strategy application process which would mean at least the following:
> * Display the order in which strategies were applied
> * For each strategy applied, display the {{toString()}} of the {{Traversal}}
> at that stage.
> A developer could then logically see what's in the magic and recognize that
> strategy 7 is ovewriting what strategy 2 is doing or that strategy 4 is
> executing before strategy 3 somehow.
> Not sure what the best way is to implement this. A straightforward solution
> with an environment variable check and println in {{applyStrategies}} seemed
> too "cheap" though kinda nice because very little would change and that
> output would come as part of the standard {{Traversal}} execution. Other
> (better) ways?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
