[Patch] PluginListParser.parse amounts to 80% of allocations during startup

[simon04]

What steps will reproduce the problem?

java -agentpath:async-profiler-2.0-rc-macos-x64/build/libasyncProfiler.so=start,file=profile.html,event=alloc -jar josm-latest.jar

​https://github.com/jvm-profiling-tools/async-profiler

What is the expected result?

PluginListParser.parse does not contribute significantly to heap allocations

What happens instead?

PluginListParser.parse amounts to 80% of allocations during startup

Please provide any additional information below. Attach a screenshot if possible.

Relative:URL: ^/trunk
Repository:UUID: 0c6e7542-c601-0410-84e7-c038aed88b3b
Last:Changed Date: 2021-03-01 23:51:35 +0100 (Mon, 01 Mar 2021)
Revision:17551
Build-Date:2021-03-02 02:30:57
URL:https://josm.openstreetmap.de/svn/trunk

Identification: JOSM/1.5 (17551 en) Mac OS X 10.16
OS Build number: macOS 11.2.1 (20D74)
Memory Usage: 280 MB / 2048 MB (108 MB allocated, but free)
Java version: 15.0.2+7, Azul Systems, Inc., OpenJDK 64-Bit Server VM
Look and Feel: com.formdev.flatlaf.FlatLightLaf
Screen: Display 1 1680×1050 (scaling 2,00×2,00) Display 2 3008×1692 (scaling 2,00×2,00)
Maximum Screen Size: 3008×1692
Best cursor sizes: 16×16→16×16, 32×32→32×32

Plugins:
+ flatlaf (35703)
+ utilsplugin2 (35691)

Map paint styles:
+ https://josm.openstreetmap.de/josmfile?page=Styles/PublicTransport&zip=1

[Don-vip]

is the html file ok? I can't open it

[simon04]

The file is okay on my disk, but attachment:profile.html.gz​​ is broken. I'll upload the non-gz variant.

[Don-vip]

So much better. I wonder why we didn't think of storing the attributes directly.

[simon04]

In 17571/josm:

fix #20563 - PluginListParser.parse amounts to 80% of allocations during startup

[Don-vip]

It seems these tests, or plugins, need an update:

org.openstreetmap.josm.plugins.PluginHandlerTest.testPluginInformationAction()
org.openstreetmap.josm.actions.downloadtasks.PluginDownloadTaskTest.testUpdatePluginCorrupt

See jenkins/job/JOSM/jdk=JDK8/lastCompletedBuild/testReport/

[simon04]

In 17573/josm:

see #20563 - Fix PluginHandler, PluginInformation

Regression of r17571.

[simon04]

Err, concerning the first test, the class Attributes, starting with Java 11,

This map and its views have a predictable iteration order, namely the order that keys were inserted into the map, as with LinkedHashMap.
​https://docs.oracle.com/en/java/javase/11/docs/api/java.base/java/util/jar/Attributes.html

PluginDownloadTaskTest found an actual glitch in my commit...

[Don-vip]

Replying to simon04:

PluginDownloadTaskTest found an actual glitch in my commit...

That's why we have them :)