Hi,
On Mon, Apr 12, 2021 at 12:13 PM Daniel Radünz wrote:
> Hello,
>
> I am facing a problem with our Wicket 8.10.0 application and I wonder if
> it's a
> bug or if I am using the MetaData feature in an unsupported or wrong way.
>
> I have a Behavior, which adds a MetaData entry to a component in the
> Behavior's
> "onConfigure(Component component)" method and later removes the same
> MetaData again
> from the component in the Behaviour's "detach(Component component)" method.
> (I'll skip explaining what it is used for, but if it's important to
> understand my
> use case I'll gladly explain it)
>
> Now upon detaching the component with said Behaviour, Wickets
> NotDetachedModelChecker
> encountered a not detached model in the component. After debugging the
> problem for
> a few hours I discovered the cause:
>
> Right before the detach phase starts, the components internal "data" field
> is an
> array with the following 4 entries:
> - [0] the component's Model
> - [1] the MetaDataEntry belonging to the described Behavior
> - [2] the Behavior itself
> - [3] an AttributeModifier with a StringResourceModel
>
> When the component's "detach()" method is called by Wicket, it internally
> calls
> the static "Behaviors.detach(Component component)" method. This method
> then iterates
> through all the entries in the aforementioned "data" array and detaches
> them, if
> they are Behaviours.
> Now when it detaches the Behavior (index 2), the MetaDataEntry (index 1)
> gets removed
> as well by our Behavior, shifting every entry in the data array by 1. The
> next loop
> iteration for index 3 then doesn't find anything anymore, since the
> AttributeModifier,
> which still needed to get detached, is now at index 2 and not 3 anymore,
> ultimately
> resulting in the observed error from the NotDetachedModelChecker.
>
> To fix the problem for now I set the MetaDataEntry to a known non-null
> "undefined"
> value in the Behavior's "detach(Component component)" method and then
> compare that
> value again in the Behavior's "onConfigure(Component component)" to treat
> it as null.
>
> Now I wonder, if you would consider this being a bug, if I'm just using it
> the wrong
> way or if I should rather save the information in a different way, like
> for example
> in a map (Component mapped to value) in the RequestCycle object.
>
It sounds like a bug to me!
Please create an issue in JIRA. With a failing test case would be perfect!
To work it around you can probably use RequestCycle's MetaData instead.
Regards,
Martin
>
> Thanks for your help
>
> Daniel
>
>
> -
> To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
> For additional commands, e-mail: users-h...@wicket.apache.org
>
>
