[
https://issues.apache.org/jira/browse/TINKERPOP-2986?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17756567#comment-17756567
]
ASF GitHub Bot commented on TINKERPOP-2986:
-------------------------------------------
li-boxuan opened a new pull request, #2212:
URL: https://github.com/apache/tinkerpop/pull/2212
SparkGraphComputer uses StarGraph to represent an in-memory graph.
EdgeFilter can be configured by users to drop unneeded edges, but the edge
properties aren't dropped. This fixes this bug and greatly reduces memory
footprint.
> StarGraph shall drop edge properties when dropping edges
> --------------------------------------------------------
>
> Key: TINKERPOP-2986
> URL: https://issues.apache.org/jira/browse/TINKERPOP-2986
> Project: TinkerPop
> Issue Type: Improvement
> Reporter: Boxuan Li
> Priority: Major
>
> SparkGraphComputer uses StarGraph and StarGraph::StarVertex to represent
> vertices and edges. The StarGraph::StarVertex::dropEdges method drops
> unneeded edges, but it doesn't drop their edge properties, which are stored
> in another data structure (StarGraph::edgeProperties).
> In my use case, I only need edges with a certain label - yet all edge
> properties are loaded and never get dropped. With my local fix, the memory
> footprint used by the Spark program drops by more than 90%.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
