TINKERPOP-1758 Apply RemoteStrategy before all DecorationStrategy instances

[spmallette]

https://issues.apache.org/jira/browse/TINKERPOP-1758

Played around with a few ways to test this, but none seemed especially good. Everything I try to do seems fairly contrived or redundant. In the end, I ended up feeling like it was safe to just rely on the TraversalStrategies sorting system with posts/priors. That's pretty solidly tested so perhaps that is enough...

Here's the result of a quick manual test:

gremlin> g = TinkerGraph.open().traversal().withRemote('conf/remote-graph.properties').withStrategies(SubgraphStrategy.build().vertices(identity()).create())
==>graphtraversalsource[tinkergraph[vertices:0 edges:0], standard]
gremlin> g.V().out().out().values('name').explain()
==>Traversal Explanation
=======================================================================================================================================
Original Traversal                 [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(OUT,vertex), PropertiesStep([name],value)]

RemoteStrategy               [D]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
SubgraphStrategy             [D]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
ConnectiveStrategy           [D]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
RangeByIsCountStrategy       [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
RepeatUnrollStrategy         [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
MatchPredicateStrategy       [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
FilterRankingStrategy        [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
InlineFilterStrategy         [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
IncidentToAdjacentStrategy   [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
AdjacentToIncidentStrategy   [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
PathRetractionStrategy       [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
LazyBarrierStrategy          [O]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
TinkerGraphCountStrategy     [P]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
TinkerGraphStepStrategy      [P]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
ProfileStrategy              [F]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]
StandardVerificationStrategy [V]   [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]

Final Traversal                    [RemoteStep(DriverServerConnection-localhost/127.0.0.1:8182 [graph=g])]

All tests pass with docker/build.sh -t -n -i

VOTE +1

[dkuppitz]

VOTE: +1

[robertdale]

Does this handle custom strategies?

[spmallette]

no - this only works by convention @robertdale - i don't think there's a way to do that short of scanning the classpath somehow for "decorations" and ordering them properly. or perhaps something more drastic that mucked with strategy ordering that always placed RemoteStrategy first, but i was hesitant to mess with that too much. i dunno - am i not thinking of another way to implement this?

[robertdale]

I don't know. Just asking for the record.
VOTE +1

[spmallette]

fair enough - i kinda wanted to just fix this problem since it was out there for so long. i'm hesitant to just hardcode the RemoteStrategy as "first" - something feels off about doing that. perhaps we wait until this approach presents as an issue before building lots of new infrastructure to support something more flexible.