[
https://issues.apache.org/jira/browse/BEAM-9578?focusedWorklogId=417156&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-417156
]
ASF GitHub Bot logged work on BEAM-9578:
----------------------------------------
Author: ASF GitHub Bot
Created on: 06/Apr/20 21:54
Start Date: 06/Apr/20 21:54
Worklog Time Spent: 10m
Work Description: ihji commented on pull request #11205: [BEAM-9578]
Enumerating artifacts is too expensive in Java
URL: https://github.com/apache/beam/pull/11205#discussion_r404411951
##########
File path:
runners/core-construction-java/src/main/java/org/apache/beam/runners/core/construction/Environments.java
##########
@@ -214,24 +220,90 @@ public static Environment createProcessEnvironment(
pathsToStage.addAll(stagingFiles);
}
- ImmutableList.Builder<ArtifactInformation> filesToStage =
ImmutableList.builder();
+ ImmutableList.Builder<Supplier<ArtifactInformation>> lazyArtifactsBuilder =
+ ImmutableList.builder();
for (String path : pathsToStage) {
Review comment:
This for loop is fairly cheap and from stream of Suppliers we can easily get
additional performance benefits by creating parallelStream. When we consider
parallelizing expensive computations, some boilerplate codes are needed anyway
in `getNonDeferredArtifacts()`. I think building a stream is a nice way to
abstract them out.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 417156)
Time Spent: 6.5h (was: 6h 20m)
> Enumerating artifacts is too expensive in Java
> ----------------------------------------------
>
> Key: BEAM-9578
> URL: https://issues.apache.org/jira/browse/BEAM-9578
> Project: Beam
> Issue Type: Bug
> Components: sdk-java-core
> Reporter: Luke Cwik
> Assignee: Heejong Lee
> Priority: Critical
> Labels: portability
> Fix For: 2.21.0
>
> Time Spent: 6.5h
> Remaining Estimate: 0h
>
> There are a lot of places (e.g. *ParDoTranslation#getParDoPayload*) which
> effectively call *Environments#createOrGetDefaultEnvironment* which causes
> [artifacts to be
> computed|https://github.com/apache/beam/blob/fc6cef9972780ca6b7525d4aadd65a8344221f1b/runners/core-construction-java/src/main/java/org/apache/beam/runners/core/construction/Environments.java#L114].
> This leads to zipping directories for non-jar dependencies.
> Similar problems may exist for Python/Go.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
