Was working on serializing JanusGraph predicates - geo, text - for
withRemote. Since those predicates become P, I had to borrow and modify the
TinkerPop P serializer and noticed that something's not like the other.

Robert Dale

On Thu, Jun 15, 2017 at 8:37 AM, Stephen Mallette <[email protected]>
wrote:

> Robert, how did you go about hitting that problem with P.inside()? It
> occurs to me now that this was so deadly a bug because I'm not sure we ever
> end up actually serializing an "inside".
>
> On Thu, Jun 15, 2017 at 6:23 AM, Stephen Mallette <[email protected]>
> wrote:
>
> > We do have a test for P.inside in the process tests but I didn't realize
> > that it doesn't compile to a P.inside at bytecode serialization time:
> >
> > gremlin> g.V(1).outE().has("weight", P.inside(0.0d,
> 0.6d)).inV().explain()
> > ==>Traversal Explanation
> > ============================================================
> > ============================================================
> > ===========================
> > Original Traversal                 [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> >
> > ConnectiveStrategy           [D]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > MatchPredicateStrategy       [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > FilterRankingStrategy        [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > InlineFilterStrategy         [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > IncidentToAdjacentStrategy   [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > AdjacentToIncidentStrategy   [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > RepeatUnrollStrategy         [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > RangeByIsCountStrategy       [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > PathRetractionStrategy       [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > LazyBarrierStrategy          [O]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > TinkerGraphCountStrategy     [P]   [GraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > TinkerGraphStepStrategy      [P]   [TinkerGraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > ProfileStrategy              [F]   [TinkerGraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> > StandardVerificationStrategy [V]   [TinkerGraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> >
> > Final Traversal                    [TinkerGraphStep(vertex,[1]),
> > VertexStep(OUT,edge), HasStep([weight.and(gt(0.0), lt(0.6))]),
> > EdgeVertexStep(IN)]
> >
> > We likely need more direct serialization tests of P, but I think those
> > already exist in master. Made a note to review further after release.
> >
> >
> > On Wed, Jun 14, 2017 at 3:57 PM, Robert Dale <[email protected]> wrote:
> >
> >> Fix pushed to tp32 and master.
> >>
> >> Robert Dale
> >>
> >> On Wed, Jun 14, 2017 at 3:50 PM, Stephen Mallette <[email protected]
> >
> >> wrote:
> >>
> >> > Well - now that the VOTE on 3.2.5 is cancelled we can now fix up these
> >> > couple of issues, specifically:
> >> >
> >> > 1. anyStepRecursively() bug (kuppitz is going to handle that)
> >> > 2. Gryo serialization of inside() (robert dale, you had the fix for
> >> that -
> >> > do you want to just CTR that in? though i'm also interested in why
> tests
> >> > didn't catch that problem)
> >> >
> >> > I'm going to leave out the other issue noted:
> >> >
> >> > https://issues.apache.org/jira/browse/TINKERPOP-1691
> >> >
> >> > as it is not user facing  - just something related to the test suite
> >> > (providers at least have a workaround for that if they hit problems as
> >> they
> >> > can @OptOut).
> >> >
> >> > I also don't intend to deploy another SNAPSHOT so i'm just going to
> >> keep us
> >> > on "3.2.5" and not revert to "3.2.5-SNAPSHOT". Let's just patch this
> up
> >> > then I'll start on a fresh release packaging tomorrow.
> >> >
> >> > Any other concerns?
> >> >
> >> >
> >> >
> >> > On Tue, Jun 6, 2017 at 6:27 AM, Robert Dale <[email protected]>
> wrote:
> >> >
> >> > > That will probably work too. I use https://wummel.github.io/linkc
> >> hecker/
> >> > >
> >> > >
> >> > > Robert Dale
> >> > >
> >> > > On Tue, Jun 6, 2017 at 6:20 AM, Daniel Kuppitz <[email protected]>
> >> wrote:
> >> > >
> >> > > > https://validator.w3.org/checklink
> >> > > >
> >> > > > On Tue, Jun 6, 2017 at 12:18 PM, Stephen Mallette <
> >> > [email protected]>
> >> > > > wrote:
> >> > > >
> >> > > > > huh - that's a neat idea. is there a specific tool you use?
> >> > > > >
> >> > > > > On Tue, Jun 6, 2017 at 5:48 AM, Robert Dale <[email protected]>
> >> > wrote:
> >> > > > >
> >> > > > > > Linkchecker passes.
> >> > > > > >
> >> > > > > > Robert Dale
> >> > > > > >
> >> > > > > > On Mon, Jun 5, 2017 at 6:38 AM, Stephen Mallette <
> >> > > [email protected]
> >> > > > >
> >> > > > > > wrote:
> >> > > > > >
> >> > > > > > > I published latest docs for 3.2.5-SNAPSHOT:
> >> > > > > > >
> >> > > > > > > http://tinkerpop.apache.org/docs/3.2.5-SNAPSHOT/
> >> > > > > > >
> >> > > > > > > and made another deployment to the Apache Snapshot Repo
> after
> >> > those
> >> > > > > > > TinkerFactory adjustments.
> >> > > > > > >
> >> > > > > > > On Fri, Jun 2, 2017 at 8:39 PM, Stephen Mallette <
> >> > > > [email protected]
> >> > > > > >
> >> > > > > > > wrote:
> >> > > > > > >
> >> > > > > > > > Just a reminder that code is frozen on the tp32 branch
> >> starting
> >> > > > > > tomorrow
> >> > > > > > > > (Saturday) and for the following week. We'll use this
> >> thread to
> >> > > > > discuss
> >> > > > > > > any
> >> > > > > > > > issues or problems on 3.2.5 that are found during testing.
> >> > There
> >> > > > are
> >> > > > > no
> >> > > > > > > > open pull requests and no outstanding issues that I'm
> aware
> >> of.
> >> > > > I've
> >> > > > > > > > published a TinkerPop 3.2.5-SNAPSHOT for providers to test
> >> > > against
> >> > > > > (or
> >> > > > > > > they
> >> > > > > > > > may build themselves - whatever is more convenient).
> >> > > > > > > >
> >> > > > > > > > Thanks,
> >> > > > > > > >
> >> > > > > > > > Stephen
> >> > > > > > > >
> >> > > > > > >
> >> > > > > >
> >> > > > >
> >> > > >
> >> > >
> >> >
> >>
> >
> >
>

Reply via email to