[
https://issues.apache.org/jira/browse/TINKERPOP-3036?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17842459#comment-17842459
]
Stephen Mallette commented on TINKERPOP-3036:
---------------------------------------------
This was "fixed" with TINKERPOP-3056. In short, 3.6.7/3.7.2 prevents the case
where you can inadvertently update an {{Element}} by throwing an error as
follows:
{code}
gremlin> g.mergeV([(T.id):'v-1']).
......1> option(onCreate, [(T.label): 'Person', 'email':
'[email protected]', 'age': 12]).
......2> option(onMatch, sideEffect(property(single, 'email',
'[email protected]')).
......3> sideEffect(property(single, 'age',
22)).constant([:])).
......4> mergeV([(T.id):'v-2']).
......5> option(onCreate, [(T.label): 'Person', 'email':
'[email protected]', 'age': 13]).
......6> option(onMatch, sideEffect(property(single, 'email',
'[email protected]')).
......7> sideEffect(property(single, 'age',
23)).constant([:])).
......8> mergeV([(T.id):'v-3']).
......9> option(onCreate, [(T.label): 'Person', 'email':
'[email protected]', 'age': 14]).
.....10> option(onMatch, sideEffect(property(single, 'email',
'[email protected]')).
.....11> sideEffect(property(single, 'age',
24)).constant([:])).
.....12> id()
The incoming traverser for MergeVertexStep cannot be an Element
Type ':help' or ':h' for help.
{code}
In 3.7.2, you would use the [Cardinality
syntax|https://tinkerpop.apache.org/docs/current/upgrade/#_map_and_cardinality].
In 4.x, there is a breaking change that will consistently promote the matched
{{Element}} to the {{option()}}. We could not make this change in earlier
versions without severely breaking behavior folks rely on where they expect a
{{Map}} as the incoming traverser to {{option()}}. The {{Map}} behavior was
important but less expected as compared to the {{Element}}. In 4.x if you want
the {{Map}} behavior you would have to label that step and {[select()}} it like:
{code}
g.inject(xx1, xx1, xx2).
fold().as("m").
mergeE(__.select("m").limit(Scope.local,1)).
option(Merge.onCreate, __.select("m").range(Scope.local, 1, 2)).
option(Merge.onMatch, __.select("m").tail(Scope.local))
{code}
> mergeV updating incorrect vertices and other issues
> ---------------------------------------------------
>
> Key: TINKERPOP-3036
> URL: https://issues.apache.org/jira/browse/TINKERPOP-3036
> Project: TinkerPop
> Issue Type: Bug
> Affects Versions: 3.7.1
> Reporter: Kelvin Lawrence
> Priority: Critical
>
> The following Stack Overflow post (link below) reported some issues with
> mergeV when side effects and cardinality are used (as well as with 3 mergeV
> steps chained together). The issue reproduces on 3.7.1 using TinkerGraph.
> [https://stackoverflow.com/questions/77803485/batch-mergev-working-well-when-creating-items-but-yielding-unexpected-results-w]
>
> Reproduction steps
>
> {code:java}
> gremlin> g=TinkerGraph.open().traversal()
> ==>graphtraversalsource[tinkergraph[vertices:0 edges:0], standard]
> gremlin> g.mergeV([(T.id):'v-1']).
> ......1> option(onCreate, [(T.label): 'Person', 'email':
> '[email protected]', 'age': 12]).
> ......2> option(onMatch, sideEffect(property(single, 'email',
> '[email protected]')).
> ......3> sideEffect(property(single, 'age',
> 22)).constant([:])).
> ......4> mergeV([(T.id):'v-2']).
> ......5> option(onCreate, [(T.label): 'Person', 'email':
> '[email protected]', 'age': 13]).
> ......6> option(onMatch, sideEffect(property(single, 'email',
> '[email protected]')).
> ......7> sideEffect(property(single, 'age',
> 23)).constant([:])).
> ......8> mergeV([(T.id):'v-3']).
> ......9> option(onCreate, [(T.label): 'Person', 'email':
> '[email protected]', 'age': 14]).
> .....10> option(onMatch, sideEffect(property(single, 'email',
> '[email protected]')).
> .....11> sideEffect(property(single, 'age',
> 24)).constant([:])).
> .....12> id()
> ==>v-3
> gremlin> g.V().valueMap(true)
> ==>[id:v-2,label:Person,email:[[email protected]],age:[13]]
> ==>[id:v-1,label:Person,email:[[email protected]],age:[12]]
> ==>[id:v-3,label:Person,email:[[email protected]],age:[14]]
> gremlin> g.mergeV([(T.id):'v-1']).
> ......1> option(onCreate, [(T.label): 'Person', 'email':
> '[email protected]', 'age': 12]).
> ......2> option(onMatch, sideEffect(property(single, 'email',
> '[email protected]')).
> ......3> sideEffect(property(single, 'age',
> 22)).constant([:])).
> ......4> mergeV([(T.id):'v-2']).
> ......5> option(onCreate, [(T.label): 'Person', 'email':
> '[email protected]', 'age': 13]).
> ......6> option(onMatch, sideEffect(property(single, 'email',
> '[email protected]')).
> ......7> sideEffect(property(single, 'age',
> 23)).constant([:])).
> ......8> mergeV([(T.id):'v-3']).
> ......9> option(onCreate, [(T.label): 'Person', 'email':
> '[email protected]', 'age': 14]).
> .....10> option(onMatch, sideEffect(property(single, 'email',
> '[email protected]')).
> .....11> sideEffect(property(single, 'age',
> 24)).constant([:])).
> .....12> id()
> ==>v-3
> gremlin> g.V().valueMap(true)
> ==>[id:v-2,label:Person,email:[[email protected]],age:[24]]
> ==>[id:v-1,label:Person,email:[[email protected]],age:[23]]
> ==>[id:v-3,label:Person,email:[[email protected]],age:[14]]
> gremlin> {code}
> The query seems to be both updating values in the incorrect vertices as well
> as leaving some values unchanged.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
