[
https://issues.apache.org/jira/browse/TINKERPOP-3201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18033389#comment-18033389
]
ASF GitHub Bot commented on TINKERPOP-3201:
-------------------------------------------
andreachild commented on code in PR #3248:
URL: https://github.com/apache/tinkerpop/pull/3248#discussion_r2467531060
##########
gremlin-core/src/main/java/org/apache/tinkerpop/gremlin/process/traversal/step/filter/TailGlobalStep.java:
##########
@@ -58,9 +59,13 @@ public Traverser.Admin<S> processNextStart() {
// If we are bypassing this step, let everything through.
return this.starts.next();
} else {
+ if (this.tail.isEmpty()) {
+ consuming = true;
+ }
// Pull everything available before we start delivering from the
tail buffer.
- if (this.starts.hasNext()) {
+ if (consuming && this.starts.hasNext()) {
this.starts.forEachRemaining(this::addTail);
+ consuming = false;
Review Comment:
Can this be replaced with just a check for `tail.isEmpty()`?
```suggestion
if (this.tail.isEmpty() && this.starts.hasNext()) {
this.starts.forEachRemaining(this::addTail);
```
> Fix tail inside repeat
> ----------------------
>
> Key: TINKERPOP-3201
> URL: https://issues.apache.org/jira/browse/TINKERPOP-3201
> Project: TinkerPop
> Issue Type: Bug
> Components: process
> Affects Versions: 3.8.0
> Reporter: Ken Hu
> Priority: Major
> Fix For: 3.8.0
>
>
> tail() step inside repeat() is not properly resetting per iteration.
> Consider the following queries on the modern graph on 3.7.x
> {code:java}
> gremlin>
> g.withoutStrategies(RepeatUnrollStrategy).V(1).repeat(__.both().order().tail(2)).times(2)
> ==>v[4]
> ==>v[6]
> {code}
> {code:java}
> gremlin> g.V(1).both().order().tail(2).both().order().tail(2)
> ==>v[5]
> ==>v[6]
> {code}
> The unrolled version doesn't match the repeat() version because tail() is
> recalculated on each call to processNextStart(). The tail() is being
> calculated too frequently even though there are still output traversers
> remaining from the previous calculation.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
