You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@wicket.apache.org by Martin Grigorov <mg...@apache.org> on 2014/05/26 21:51:57 UTC

Re: ComponentResolvers tried all the way up the hierarchy

Hi Jesse,

I've applied your patches for WICKET-5569 and WICKET-5580.
Please create a new ticket with this improvement + a test.

Martin Grigorov
Wicket Training and Consulting


On Mon, Apr 28, 2014 at 11:07 PM, Martin Grigorov <mg...@apache.org>wrote:

> Hi Jesse,
>
> The improvement sounds good !
> I won't be able to work on it this week but I'm sure you will create a
> patch with a test by the time I'm back ;)
>
> Martin Grigorov
> Wicket Training and Consulting
>
>
> On Sun, Apr 27, 2014 at 10:50 PM, Jesse Long <jp...@unknown.za.net> wrote:
>
>> Hi Wicket Devs,
>>
>> ComponentResolvers#resolve() first tries ComponentResolvers#
>> resolveByComponentHierarchy().
>>
>> This tries the immediate container going all the way up the hierarchy
>> attempting to resolve the component is the container is also a
>> IComponentResolver.
>>
>> I'm worried that the wrong problem is being fixed in the wrong place here.
>>
>> Basically, I understand the need for this code as:
>> TransparentWebMarkupContainer inside TransparentWebMarkupContainer would
>> not work, because TransparentWebMarkupContainer just calls get(String) on
>> its parent. If its parent is also transparent, the get(String) is not going
>> to return the component, instead, resolve() should be called.
>>
>> Therefore, we must try the first one, and if that fails, continue up the
>> tree in case the parent is also a TransparentWebMarkupContainer.
>>
>> This smells like the ComponentResolvers class fixing
>> TransparentWebMarkupContainer bugs.
>>
>> Rather, we could just make TransparentWebMarkupContainer first try
>> get(String) on the parent, if that fails, check if the parent is a
>> IComponentResolver and if so, call resolve() on it.
>>
>> Like:
>>     @Override
>>     public Component resolve(MarkupContainer container, MarkupStream
>> markupStream, ComponentTag tag)
>>     {
>>         MarkupContainer parent = getParent();
>>         Component resolvedComponent = parent.get(tag.getId());
>>
>>         /*
>>          * If get(String didn't work, parent may be a IComponentResolver,
>> in which case we must call
>>          * resolve() on it.
>>          */
>>         if (resolvedComponent == null && parent instanceof
>> IComponentResolver)
>>         {
>>             resolvedComponent = ((IComponentResolver)parent).resolve(container,
>> markupStream, tag);
>>         }
>>
>>         if (resolvedComponent != null && getPage().wasRendered(
>> resolvedComponent))
>>         {
>>             /*
>>              * Means that parent container has an associated homonymous
>> tag to this grandchildren
>>              * tag in markup. The parent container wants render it and it
>> should be not resolved to
>>              * their grandchildren.
>>              */
>>             return null;
>>         }
>>         return resolvedComponent;
>>     }
>>
>>
>> When using a TransparentWebMarkupContainer, it is reasonable to expect
>> the component to be resolved from the component bound to the grandparent of
>> the component tag in the markup hierarchy, but with the current plan, the
>> component could be resolved from the great great great grandparent.
>>
>> Current behavior probably also leads to some confusing error messages
>> when a component cannot be resolved, where a component deep in the
>> hierarchy could "steal" a component which has not yet been rendered from an
>> ancestor, and the error message would be about how the ancestor is missing
>> a component or it was already rendered etc. User would have a hard time
>> figuring out that the problem is actually deeper inside the hierarchy.
>>
>> Thoughts?
>>
>> Thanks,
>> Jesse
>>
>
>

Re: ComponentResolvers tried all the way up the hierarchy

Posted by Jesse Long <jp...@unknown.za.net>.
Hi Martin,

Thanks for merging that.

I have been working on a patch for this, but got stuck with some errors 
introduced in HtmlHeaderContainer, Border and a few others. Will 
continue working on it this weekend.

Thanks,
Jesse

On 26/05/2014 21:51, Martin Grigorov wrote:
> Hi Jesse,
>
> I've applied your patches for WICKET-5569 and WICKET-5580.
> Please create a new ticket with this improvement + a test.
>
> Martin Grigorov
> Wicket Training and Consulting
>
>
> On Mon, Apr 28, 2014 at 11:07 PM, Martin Grigorov <mg...@apache.org>wrote:
>
>> Hi Jesse,
>>
>> The improvement sounds good !
>> I won't be able to work on it this week but I'm sure you will create a
>> patch with a test by the time I'm back ;)
>>
>> Martin Grigorov
>> Wicket Training and Consulting
>>
>>
>> On Sun, Apr 27, 2014 at 10:50 PM, Jesse Long <jp...@unknown.za.net> wrote:
>>
>>> Hi Wicket Devs,
>>>
>>> ComponentResolvers#resolve() first tries ComponentResolvers#
>>> resolveByComponentHierarchy().
>>>
>>> This tries the immediate container going all the way up the hierarchy
>>> attempting to resolve the component is the container is also a
>>> IComponentResolver.
>>>
>>> I'm worried that the wrong problem is being fixed in the wrong place here.
>>>
>>> Basically, I understand the need for this code as:
>>> TransparentWebMarkupContainer inside TransparentWebMarkupContainer would
>>> not work, because TransparentWebMarkupContainer just calls get(String) on
>>> its parent. If its parent is also transparent, the get(String) is not going
>>> to return the component, instead, resolve() should be called.
>>>
>>> Therefore, we must try the first one, and if that fails, continue up the
>>> tree in case the parent is also a TransparentWebMarkupContainer.
>>>
>>> This smells like the ComponentResolvers class fixing
>>> TransparentWebMarkupContainer bugs.
>>>
>>> Rather, we could just make TransparentWebMarkupContainer first try
>>> get(String) on the parent, if that fails, check if the parent is a
>>> IComponentResolver and if so, call resolve() on it.
>>>
>>> Like:
>>>      @Override
>>>      public Component resolve(MarkupContainer container, MarkupStream
>>> markupStream, ComponentTag tag)
>>>      {
>>>          MarkupContainer parent = getParent();
>>>          Component resolvedComponent = parent.get(tag.getId());
>>>
>>>          /*
>>>           * If get(String didn't work, parent may be a IComponentResolver,
>>> in which case we must call
>>>           * resolve() on it.
>>>           */
>>>          if (resolvedComponent == null && parent instanceof
>>> IComponentResolver)
>>>          {
>>>              resolvedComponent = ((IComponentResolver)parent).resolve(container,
>>> markupStream, tag);
>>>          }
>>>
>>>          if (resolvedComponent != null && getPage().wasRendered(
>>> resolvedComponent))
>>>          {
>>>              /*
>>>               * Means that parent container has an associated homonymous
>>> tag to this grandchildren
>>>               * tag in markup. The parent container wants render it and it
>>> should be not resolved to
>>>               * their grandchildren.
>>>               */
>>>              return null;
>>>          }
>>>          return resolvedComponent;
>>>      }
>>>
>>>
>>> When using a TransparentWebMarkupContainer, it is reasonable to expect
>>> the component to be resolved from the component bound to the grandparent of
>>> the component tag in the markup hierarchy, but with the current plan, the
>>> component could be resolved from the great great great grandparent.
>>>
>>> Current behavior probably also leads to some confusing error messages
>>> when a component cannot be resolved, where a component deep in the
>>> hierarchy could "steal" a component which has not yet been rendered from an
>>> ancestor, and the error message would be about how the ancestor is missing
>>> a component or it was already rendered etc. User would have a hard time
>>> figuring out that the problem is actually deeper inside the hierarchy.
>>>
>>> Thoughts?
>>>
>>> Thanks,
>>> Jesse
>>>
>>