[
https://issues.apache.org/jira/browse/WICKET-7041?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17710185#comment-17710185
]
ASF GitHub Bot commented on WICKET-7041:
----------------------------------------
theigl opened a new pull request, #572:
URL: https://github.com/apache/wicket/pull/572
This PR avoids unnecessary allocation of `AppendingStringBuffer` for a
`StringResponse` that is never written to.
In our case, 99% of the allocations in `Component#internalRenderHead` can be
avoided in this way.
I'm not sure if we should add `LazyStringResponse` as a separate class, or
simply make the existing `StringResponse` lazy. It is used in a lot of places,
so we might avoid more allocations if we changed the default implementation.
See https://issues.apache.org/jira/browse/WICKET-7041
> Reduce allocations when rendering component headers
> ---------------------------------------------------
>
> Key: WICKET-7041
> URL: https://issues.apache.org/jira/browse/WICKET-7041
> Project: Wicket
> Issue Type: Improvement
> Components: wicket-core
> Affects Versions: 9.12.0
> Reporter: Thomas Heigl
> Assignee: Thomas Heigl
> Priority: Major
> Attachments: image-2023-04-10-16-56-52-061.png
>
>
> We noticed a *very* substantial amount of allocations coming from
> Component#internalRenderHead:
> !image-2023-04-10-16-56-52-061.png|width=716,height=612!
> This is the relevant part of the code:
> https://github.com/apache/wicket/blame/9b4c5fe9a26174a06d612837297eca90ef8711cc/wicket-core/src/main/java/org/apache/wicket/Component.java#L2642
> {code:java}
> StringResponse markupHeaderResponse = new StringResponse();
> Response oldResponse = getResponse();
> RequestCycle.get().setResponse(markupHeaderResponse);
> try
> {
> // Make sure the markup source strategy contributes to the header first
> // to be backward compatible. WICKET-3761
> getMarkupSourcingStrategy().renderHead(this, container);
> CharSequence headerContribution = markupHeaderResponse.getBuffer();
> if (Strings.isEmpty(headerContribution) == false)
> {
> response.render(StringHeaderItem.forString(headerContribution));
> }
> }
> finally
> {
> RequestCycle.get().setResponse(oldResponse);
> }
> {code}
> The code *always* allocates a new StringResponse for rendering potential
> markup header contributions. Internally, this creates a new
> AppendingStringBuffer with capacity 128.
> In my application, we do not use <wicket:head> tags at all. So 99.9% of these
> allocations are completely unnecessary.
> Ideally, I'd like to get rid of the temporary StringResponse, but I don't
> know if thats possible without breaking things.
> Another solution would be to create a LazyStringResponse that initializes the
> internal String buffer on first use.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
