[ 
https://issues.apache.org/jira/browse/TINKERPOP3-715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14577937#comment-14577937
 ] 

Daniel Kuppitz commented on TINKERPOP3-715:
-------------------------------------------

It's not the first time that this question comes up, but - Nah, please don't. 
{{.optional()}} never worked as expected. Your sample can be solved using 
{{.union()}}:

{code}
g.V(x).union(
    out('rated').aggregate('userLibrary').is(null), // .aggregate() will 
collect all vertices, but .is(null) will never emit an element
    outE('rated').has('stars', gte(4)).inV().       // union-traversals are 
processed in their given order, thus 'userLibrary' will be available in the 2nd 
union-traversal
    inE('rated').has('stars',gte(4)).inV().
    
outE('rated').has('stars',gte(4)).inV().where(not(within('userLibrary')))).groupCount()
{code}

Not tested, but it should work.

> [Proposal] Do we need a optional(traversal) like in TinkerPop2?
> ---------------------------------------------------------------
>
>                 Key: TINKERPOP3-715
>                 URL: https://issues.apache.org/jira/browse/TINKERPOP3-715
>             Project: TinkerPop 3
>          Issue Type: New Feature
>            Reporter: Marko A. Rodriguez
>
> In TinkerPop2 we had {{optional()}}. It allowed the traverser to go down the 
> {{optional}}-branch and its results were only used for their sideEffects. For 
> instance:
> {code}
> g.V(x).optional(out('rated').aggregate('userLibrary'))
>   .outE('rated').has('stars',gte(4)).inV()
>   .inE('rated').has('stars',gte(4)).inV()
>   .outE('rated').has('stars',gte(4)).inV().where(not(within('userLibrary')))
>     .groupCount()
> {code}
> To do this in TinkerPop3, we would do:
> {code}
> g.V(x).sideEffect(v -> 
> g.V(v.get()).out('rated').aggregate('userLibrary').iterate())
>   .outE('rated').has('stars',gte(4)).inV()
>   .inE('rated').has('stars',gte(4)).inV()
>   .outE('rated').has('stars',gte(4)).inV().where(not(within('userLibrary')))
>     .groupCount()
> {code}
> ... :/ ... should we add {{optional}} back in the mix? It would be like 
> {{.as('a').local(...).select('a')}} but it would not care if the nested 
> traversal returned a result or not.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to