[jira] [Commented] (TINKERPOP-2627) Use of union() in second by() to group()

["Stephen Mallette (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/TINKERPOP-2627?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17445870#comment-17445870 ] 

Stephen Mallette commented on TINKERPOP-2627:
---------------------------------------------

This is also an issue with {{coalesce()}}:

{code}
gremlin> g.V().group().by(T.label).by(coalesce(fold(),constant(null)))
==>[software:[v[5]],person:[v[6]]]
{code}

The issue seems to be related to the fact that {{GroupStep}} looks for the first non-local {{Barrier}} in the traversal and if one is found processes it up to that point before doing a final reduction on the remainder of the traversal. Since situations like {{coalesce()}} and {{union()}} have the {{Barrier}} as child traversals the {{Barrier}} is never found and that important bit of code never executes. 

The workaround for {{coalesce()}} is similar to {{union()}}:

{code}
gremlin> g.V().group().by(T.label).by(values('age').fold().local(coalesce(unfold(),constant(null)).fold()))
==>[software:[null],person:[29,27,32,35]]
{code}

> Use of union() in second by() to group()
> ----------------------------------------
>
>                 Key: TINKERPOP-2627
>                 URL: https://issues.apache.org/jira/browse/TINKERPOP-2627
>             Project: TinkerPop
>          Issue Type: Bug
>          Components: process
>    Affects Versions: 3.4.12
>            Reporter: Stephen Mallette
>            Priority: Minor
>
> Using air routes dataset, you get some odd output if you `union(...).fold()` in the second `by()` and if the children to `union()` are reducing steps:
> {code}
> gremlin> g.V().hasLabel('airport').has('country','TZ').
> ......1>   group().
> ......2>     by('country').
> ......3>     by(union(values('elev').min(), values('elev').max()).fold()).
> ......4>   unfold()
> ==>TZ=[3745, 3745, 2932, 2932, 3763, 3763, 371, 371, 5600, 5600, 4550, 4550, 182, 182, 54, 54]
> {code}
> I was able to workaround it with:
> {code}
> gremlin> g.V().hasLabel('airport').has('country','TZ').
> ......1>   group().
> ......2>     by('country').
> ......3>     by(values('elev').fold().local(union(min(local),max(local))).fold()).
> ......4>   unfold()
> ==>TZ=[54, 5600]
> {code}



--
This message was sent by Atlassian Jira
(v8.20.1#820001)