[
https://issues.apache.org/jira/browse/TINKERPOP3-781?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14652164#comment-14652164
]
Matt Frantz commented on TINKERPOP3-781:
----------------------------------------
The path history should include the common portion of the path, i.e. the part
leading up to the {{local}} step. That is how {{map(Traversal)}} works, for
example. Even though {{map}}'s argument is a traversal containing a
{{ReducingBarrierStep}}, it preserves the path prefix.
Aggregating the paths for the reduced portion of the traversal is another issue
to consider. How would you express that this additional computation was
desired? Here's a case where you do NOT need the path:
{noformat}
g.V(1).as('a').local(out().as('b').count()).as('c')
.select('a','c')
==>[a:v[1], c:3]
{noformat}
What if you asked for the "b" vertex, too? You would expect the same
cardinality result. That means "b" should be reduced with a default function,
e.g. make a list.
{noformat}
g.V(1).as('a').local(out().as('b').count()).as('c')
.select('a','b','c')
==>[a:v[1], b:[v[3],v[2],v[4]], c:3]
{noformat}
You could use {{by}} to modulate the value that is reduced.
{noformat}
g.V(1).as('a').local(out().as('b').count()).as('c')
.select('a','b','c')
.by('name').by('name').by()
==>[a:'marko', b:['lop','vadas','josh'], c:3]
{noformat}
Perhaps a second {{by}} argument could provide the reducing function?
Complicated.
{noformat}
g.V(1).as('a').local(out().as('b').count()).as('c')
.select('a','b','c')
.by('name').by('name', { acc, it -> acc += it.get().length() }).by()
==>[a:'marko', b:12, c:3]
{noformat}
Does this all become sugar for {{group}} of {{path}}?
{noformat}
g.V(1).as('a').out().as('b').path()
.group()
.by()
.by('name')
.by(/* Path reducing logic goes here */))
{noformat}
> Local aggregation should not destroy path
> -----------------------------------------
>
> Key: TINKERPOP3-781
> URL: https://issues.apache.org/jira/browse/TINKERPOP3-781
> Project: TinkerPop 3
> Issue Type: Improvement
> Components: process
> Reporter: Matt Frantz
>
> Currently, if we do an aggregating step (e.g. {{fold}}, {{groupCount}}, etc.)
> within the {{local}} step, we lose the path information.
> {noformat}
> gremlin> g.V(1).local(both().fold()).path()
> ==>[[v[3], v[2], v[4]]]
> {noformat}
> It would be better if the preceding portion of the path were still retained
> like so:
> {noformat}
> gremlin> g.V(1).local(both().fold()).path()
> ==>[v[1], [v[3], v[2], v[4]]]
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
