TINKERPOP-1867 union() can produce extra traversers

[dkuppitz]

https://issues.apache.org/jira/browse/TINKERPOP-1867

Implemented reset() in ComputerAwareStep to reset the previousIterator, and reset the finalStep in DefaultTraversal.

This fixes the false behavior, mainly seen in reducing barrier steps.

$ docker/build.sh -t -i
...
[INFO] ------------------------------------------------------------------------
[INFO] Reactor Summary:
[INFO] 
[INFO] Apache TinkerPop .................................. SUCCESS [1:45.623s]
[INFO] Apache TinkerPop :: Gremlin Shaded ................ SUCCESS [21.575s]
[INFO] Apache TinkerPop :: Gremlin Core .................. SUCCESS [57.690s]
[INFO] Apache TinkerPop :: Gremlin Test .................. SUCCESS [6.146s]
[INFO] Apache TinkerPop :: Gremlin Groovy ................ SUCCESS [3:03.207s]
[INFO] Apache TinkerPop :: Gremlin Groovy Test ........... SUCCESS [4.203s]
[INFO] Apache TinkerPop :: TinkerGraph Gremlin ........... SUCCESS [2:59.389s]
[INFO] Apache TinkerPop :: Gremlin Benchmark ............. SUCCESS [7.549s]
[INFO] Apache TinkerPop :: Gremlin Driver ................ SUCCESS [1:31.627s]
[INFO] Apache TinkerPop :: Neo4j Gremlin ................. SUCCESS [0.936s]
[INFO] Apache TinkerPop :: Gremlin Server ................ SUCCESS [15:29.995s]
[INFO] Apache TinkerPop :: Gremlin Python ................ SUCCESS [2:40.883s]
[INFO] Apache TinkerPop :: Gremlin.Net ................... SUCCESS [2.247s]
[INFO] Apache TinkerPop :: Gremlin.Net - Source .......... SUCCESS [44.422s]
[INFO] Apache TinkerPop :: Gremlin.Net - Tests ........... SUCCESS [44.304s]
[INFO] Apache TinkerPop :: Hadoop Gremlin ................ SUCCESS [6:30.134s]
[INFO] Apache TinkerPop :: Spark Gremlin ................. SUCCESS [15:04.061s]
[INFO] Apache TinkerPop :: Giraph Gremlin ................ SUCCESS [2:23:46.498s]
[INFO] Apache TinkerPop :: Gremlin Console ............... SUCCESS [4:00.598s]
[INFO] Apache TinkerPop :: Gremlin Archetype ............. SUCCESS [0.080s]
[INFO] Apache TinkerPop :: Archetype - TinkerGraph ....... SUCCESS [33.522s]
[INFO] Apache TinkerPop :: Archetype - Server ............ SUCCESS [15.100s]
[INFO] Apache TinkerPop :: Archetype - DSL ............... SUCCESS [3.588s]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 3:21:02.647s
[INFO] Finished at: Wed Jan 10 01:21:22 UTC 2018
[INFO] Final Memory: 173M/631M
[INFO] ------------------------------------------------------------------------

VOTE: +1

[spmallette]

Nice job. Does this also fix TINKERPOP-1629?

[dkuppitz]

I was hoping so, but no, it doesn't. At least not in tp32/. Maybe we're gonna be surprised once it's merged into master/.

[spmallette]

that sorta seems worrisome - i was thinking those two were highly related.....

[dkuppitz]

I've merged this branch into master/ (locally), but it didn't change anything (which is actually a good thing, as I don't like surprises after a merge :)). We still get the kinda unexpected empty list in OLTP and the whole bunch of empty lists in OLAP:

gremlin> g = TinkerFactory.createModern().traversal().withoutStrategies(StandardVerificationStrategy)
==>graphtraversalsource[tinkergraph[vertices:6 edges:6], standard]
gremlin> g.V(1).repeat(aggregate("a").fold().store("x").unfold().both().where(without("a")).dedup()).cap("x").next()
==>[]
==>[v[1]]
==>[v[3],v[2],v[4]]
==>[v[6],v[5]]
gremlin> g = TinkerFactory.createModern().traversal().withComputer().withoutStrategies(StandardVerificationStrategy)
==>graphtraversalsource[tinkergraph[vertices:6 edges:6], graphcomputer]
gremlin> g.V(1).repeat(aggregate("a").fold().store("x").unfold().both().where(without("a")).dedup()).cap("x").next()
==>[v[1]]
==>[]
==>[]
==>[]
==>[]
==>[]
==>[v[2],v[4],v[3]]
==>[v[5],v[6]]

The results are the same in tp32/ and master/, nothing's changed.

So, although the issue looks similar, it's probably a whole different thing that just happens to produce the same kind of unexpected result entries.

[spmallette]

I was hoping I could come up with some other form of testing for this change, but I can't seem to think of anything...........

VOTE +1 (pending changelog entry)

[robertdale]

VOTE +1