[
https://issues.apache.org/jira/browse/TINKERPOP-3295?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18144086#comment-18144086
]
ASF GitHub Bot commented on TINKERPOP-3295:
-------------------------------------------
Cole-Greer commented on code in PR #3802:
URL: https://github.com/apache/tinkerpop/pull/3802#discussion_r4558124054
##########
docs/src/upgrade/release-3.7.x.asciidoc:
##########
@@ -31,6 +31,52 @@ image::gremlin-zamfir.png[width=185]
Please see the
link:https://github.com/apache/tinkerpop/blob/3.7.8/CHANGELOG.asciidoc#release-3-7-8[changelog]
for a
complete list of all the modifications that are part of this release.
+=== Upgrading for Users
+
+==== Mutating Steps and Bulked Traversers
+
+Gremlin combines traversers that hold the same value into a single traverser
with a bulk count. This happens for
+duplicate values given to `inject()`, after `barrier()` and after the barriers
that `LazyBarrierStrategy` adds
+automatically. The `addV()`, `addE()`, `mergeV()`, `mergeE()` and `property()`
steps now perform their write once for
+every traverser that a bulked traverser represents, so the number of mutations
no longer depends on whether
+traversers happened to be bulked.
+
+[source,groovy]
+----
+// 3.7.7
+gremlin> g.inject(1, 1).addV('person').id()
+==>0
+==>0
+gremlin> g.V().count()
+==>1
+
+// 3.7.8
+gremlin> g.inject(1, 1).addV('person').id()
+==>0
+==>1
+gremlin> g.V().count()
+==>2
+----
+
+Traversals that write from duplicate traversers may now create more elements,
apply `onMatch` where it was skipped
+before, add more `list` cardinality values, or raise an error for a reused
user supplied `T.id`. Such traversals
+should be reviewed if they depended on the previous counts. Mutations made
inside the child traversals of steps such
+as `map()`, `sideEffect()` and `coalesce()` still occur once per bulked
traverser.
+
+See: link:https://issues.apache.org/jira/browse/TINKERPOP-3295[TINKERPOP-3295]
+
+=== Upgrading for Providers
+
+==== Graph System Providers
+
+===== Mutating Steps and Bulked Traversers
+
+Providers that replace `AddVertexStep`, `AddEdgeStep`, `MergeVertexStep`,
`MergeEdgeStep` or `AddPropertyStep` with
+their own implementations should perform the mutation once for each unit of a
traverser's bulk. New Gherkin tests in
+the `AddVertex`, `AddEdge`, `MergeVertex` and `MergeEdge` features validate
this behavior.
Review Comment:
Nit: I don't think a dedicated callout to providers is needed here, as this
is mostly a bug fix on our end. If any provider happens to have replicated our
bug, the above examples and new feature tests should be all they need.
> inject() and addV() has bulking issues
> --------------------------------------
>
> Key: TINKERPOP-3295
> URL: https://issues.apache.org/jira/browse/TINKERPOP-3295
> Project: TinkerPop
> Issue Type: Bug
> Components: process
> Affects Versions: 3.7.7
> Reporter: Stephen Mallette
> Priority: Critical
>
> {code}
> gremlin> g.inject(1,1).addV('X').id()
> ==>0
> ==>0
> gremlin> g.V().count()
> ==>1
> {code}
> the example is for {{addV()}} and the {{inject()}} start step but
> investigation should probably cover other variations of mutations like,
> {{addE()}}, {{mergeV/E()}} and {{drop()}} and steps and mid-traversal
> {{inject()}}.
> Should also cover cases like: {{g.inject(1, 1).addV('X').property(T.id,
> 'same')}}
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
