[
https://issues.apache.org/jira/browse/WICKET-5569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jesse Long updated WICKET-5569:
-------------------------------
Attachment: WICKET-5569-1.patch
WICKET-5569-1.patch
Patch against Wicket 6. This solution uses the visitor pattern to traverse all
children of the markup container to find descendants which are both
MarkupContainers and IComponentResolvers. It then attempts to retrieve markup
for the component from the MarkupContainer. If markup is found, a the
IComponentResolver interface of the MarkupContainer is used to resolve the
Component for that markup to ensure that we get the correct markup fragment.
The exact implementation is supposed to match that of
MarkupContainer#renderNext(). Please note that MarkupContainer#renderNext() is
not final. I think it should be.
I added a comment asking whether or not it is safe to assume that
MarkupStream#atTag() would return true. I'm not sure, please check and fix if
my assumption is wrong.
Included are tests.
> Unable to find markup for children of deeply nested IComponentResolvers
> during Ajax response
> --------------------------------------------------------------------------------------------
>
> Key: WICKET-5569
> URL: https://issues.apache.org/jira/browse/WICKET-5569
> Project: Wicket
> Issue Type: Bug
> Components: wicket
> Affects Versions: 6.14.0
> Reporter: Jesse Long
> Priority: Minor
> Attachments: WICKET-5569-1.patch
>
>
> Component hierarchy: Page -> WebMarkupContainer -> IComponentResolver (that
> uses Page to resolve) and Page -> Panel.
> Markup hierarchy: Page -> WebMarkupContainer -> IComponentResolver -> Panel.
> When rendering whole page, it works, because it is markup driven. Wicket
> encounters ComponentTag for Panel and resolves the Panel using
> IComponentResolver, which retrieves the Panel from the Page.
> When you add the Panel to an AjaxRequestTarget, the render is component
> driven. In order to render the Panel, we must retrieve the markup for the
> Panel from its parent MarkupContainer, which happens to be the Page.
> Markup.java around line 230 skips to closing tags of ComponentTag, so when
> Page gets to the opening tag of the WebMarkupContainer, it skips to the
> closing tag of the WebMarkupContainer, and so passes over the ComponentTag
> for Panel without noticing it. There is actually another check, in
> DefaultMarkupSourcingStrategy, that tries to fetch from all the "transparent"
> components in the markup container, but this is not good enough, because in
> our example, the IComponentResolver is not actually a direct child of the
> Panel's parent, to it is never used to try find the markup.
> One solution might be to traverse the tree, and attempt to find the markup
> from all IComponentResolving MarkupContainers, but we should be careful. I'm
> a bit concerned at how various parts of Wicket just assume that an
> IComponentResolver is transparent and resolves from its direct parent only.
> If we do go down the route of traversing the tree to find
> IComponentResolvers, then try find the markup from each of them, we really
> should add a check in
> AbstractMarkupSourcingStrategy#searchMarkupInTransparentResolvers() to check
> that the Component that the IComponentResolver resolves for the markup id is
> the same component for which we are looking for markup.
> This is a difficult one. I am working around it for the mean time, just
> recording the difficulty here. Will try make a patch when I can.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
