[Trinidad] StampState mismatched in tr:iterator

[Tibor Blénessy <bl...@gmail.com>]

Hello,

I am having troubles with tr:iterator component. We are using tr:iterator to
dynamically create form elements (there are several labels and text inputs
for each item in the collection). However transition to next page ends with
ClassCastException that HtmlScript cannot be cast to EditableValueHolder.
This cast happens in restoreRowState method of EVHState class. It seems to
me that StampStates are somehow mismatched during save and restore calls,
and that HtmlScript component pickups wrong StampState from the map (there
should be no stamp state for HtmlScript afaik, but there is instance of
EVHState with corresponding key, I was unable to track how this instance of
EVHState gets created, but it is in the map even before the first call to
the saveStampState method for that request).

We are using Trinidad 1.2.13, I found issue TRINIDAD-2047 in jira which
seemed to be related to the issue, so I build current 1.2.x branch, but
problem remains the same (althought 2047 helped a bit, bc stampstate map now
doesnt contain null values in key1 of DualKey class ).

I understand that's quite difficult to help with such a vague description,
however it's quite difficult to simulate the issue in some simple scenario.
I would be gratefull if somehow could explain how StampStates are supposed
to work, bc code looks quite messy.

We also use Spring Webflow 2.0 and Myfaces implementation 1.2.10 if it could
be of any relevance.

Thanks

Tibor Blenessy

Re: [Trinidad] StampState mismatched in tr:iterator

[Tibor Blénessy <bl...@gmail.com>]

Well, some other way how to index children would serve certainly better. I
encapsulated all children tags of tr:iterator into tr:group and it seems to
be working, bc then there is only one element in iterator's children
collection. However when using tr:group no stampstate is saved at all, so I
am not sure if this is working properly.

Tibor Blenessy

On Thu, Mar 10, 2011 at 14:39, Scott O'Bryan <da...@gmail.com> wrote:

> I think it may be a limitation of the current design.  It might be
> interesting if this could be enhanced to use some sort of 'stamp key'
> instead of index.
>
> On Mar 10, 2011, at 6:16 AM, Tibor Blénessy <bl...@gmail.com> wrote:
>
> > Hi,
> >
> > I found out that tr:iterator expects it's children List to remain always
> the
> > same, bc StampStates are indexed by position in children collection.
> However
> > this is not always true, during postback request UIInstruction components
> > generated by facelets are dropped from children collection (I think this
> > might be because they are not saved in the viewstate, maybe webflow and
> it's
> > serialization mechanism has something to do with this).
> >
> > Is this bug of Trinidad or is problem somewhere else? Maybe it would be
> > better if StampStates were indexed by component Id rather then by
> position?
> >
> > Any input much appreciated. Thanks
> >
> > Tibor
> >
> > On Tue, Mar 8, 2011 at 23:49, Tibor Blénessy <bl...@gmail.com> wrote:
> >
> >> Hello,
> >>
> >> I am having troubles with tr:iterator component. We are using
> tr:iterator
> >> to dynamically create form elements (there are several labels and text
> >> inputs for each item in the collection). However transition to next page
> >> ends with ClassCastException that HtmlScript cannot be cast to
> >> EditableValueHolder. This cast happens in restoreRowState method of
> EVHState
> >> class. It seems to me that StampStates are somehow mismatched during
> save
> >> and restore calls, and that HtmlScript component pickups wrong
> StampState
> >> from the map (there should be no stamp state for HtmlScript afaik, but
> there
> >> is instance of EVHState with corresponding key, I was unable to track
> how
> >> this instance of EVHState gets created, but it is in the map even before
> the
> >> first call to the saveStampState method for that request).
> >>
> >> We are using Trinidad 1.2.13, I found issue TRINIDAD-2047 in jira which
> >> seemed to be related to the issue, so I build current 1.2.x branch, but
> >> problem remains the same (althought 2047 helped a bit, bc stampstate map
> now
> >> doesnt contain null values in key1 of DualKey class ).
> >>
> >> I understand that's quite difficult to help with such a vague
> description,
> >> however it's quite difficult to simulate the issue in some simple
> scenario.
> >> I would be gratefull if somehow could explain how StampStates are
> supposed
> >> to work, bc code looks quite messy.
> >>
> >> We also use Spring Webflow 2.0 and Myfaces implementation 1.2.10 if it
> >> could be of any relevance.
> >>
> >> Thanks
> >>
> >> Tibor Blenessy
> >>
> >>
> >>
>

Re: [Trinidad] StampState mismatched in tr:iterator

[Scott O'Bryan <da...@gmail.com>]

I think it may be a limitation of the current design.  It might be
interesting if this could be enhanced to use some sort of 'stamp key'
instead of index.

On Mar 10, 2011, at 6:16 AM, Tibor Blénessy <bl...@gmail.com> wrote:

> Hi,
>
> I found out that tr:iterator expects it's children List to remain always the
> same, bc StampStates are indexed by position in children collection. However
> this is not always true, during postback request UIInstruction components
> generated by facelets are dropped from children collection (I think this
> might be because they are not saved in the viewstate, maybe webflow and it's
> serialization mechanism has something to do with this).
>
> Is this bug of Trinidad or is problem somewhere else? Maybe it would be
> better if StampStates were indexed by component Id rather then by position?
>
> Any input much appreciated. Thanks
>
> Tibor
>
> On Tue, Mar 8, 2011 at 23:49, Tibor Blénessy <bl...@gmail.com> wrote:
>
>> Hello,
>>
>> I am having troubles with tr:iterator component. We are using tr:iterator
>> to dynamically create form elements (there are several labels and text
>> inputs for each item in the collection). However transition to next page
>> ends with ClassCastException that HtmlScript cannot be cast to
>> EditableValueHolder. This cast happens in restoreRowState method of EVHState
>> class. It seems to me that StampStates are somehow mismatched during save
>> and restore calls, and that HtmlScript component pickups wrong StampState
>> from the map (there should be no stamp state for HtmlScript afaik, but there
>> is instance of EVHState with corresponding key, I was unable to track how
>> this instance of EVHState gets created, but it is in the map even before the
>> first call to the saveStampState method for that request).
>>
>> We are using Trinidad 1.2.13, I found issue TRINIDAD-2047 in jira which
>> seemed to be related to the issue, so I build current 1.2.x branch, but
>> problem remains the same (althought 2047 helped a bit, bc stampstate map now
>> doesnt contain null values in key1 of DualKey class ).
>>
>> I understand that's quite difficult to help with such a vague description,
>> however it's quite difficult to simulate the issue in some simple scenario.
>> I would be gratefull if somehow could explain how StampStates are supposed
>> to work, bc code looks quite messy.
>>
>> We also use Spring Webflow 2.0 and Myfaces implementation 1.2.10 if it
>> could be of any relevance.
>>
>> Thanks
>>
>> Tibor Blenessy
>>
>>
>>

Re: [Trinidad] StampState mismatched in tr:iterator

[Tibor Blénessy <bl...@gmail.com>]

Hi,

I found out that tr:iterator expects it's children List to remain always the
same, bc StampStates are indexed by position in children collection. However
this is not always true, during postback request UIInstruction components
generated by facelets are dropped from children collection (I think this
might be because they are not saved in the viewstate, maybe webflow and it's
serialization mechanism has something to do with this).

Is this bug of Trinidad or is problem somewhere else? Maybe it would be
better if StampStates were indexed by component Id rather then by position?

Any input much appreciated. Thanks

Tibor

On Tue, Mar 8, 2011 at 23:49, Tibor Blénessy <bl...@gmail.com> wrote:

> Hello,
>
> I am having troubles with tr:iterator component. We are using tr:iterator
> to dynamically create form elements (there are several labels and text
> inputs for each item in the collection). However transition to next page
> ends with ClassCastException that HtmlScript cannot be cast to
> EditableValueHolder. This cast happens in restoreRowState method of EVHState
> class. It seems to me that StampStates are somehow mismatched during save
> and restore calls, and that HtmlScript component pickups wrong StampState
> from the map (there should be no stamp state for HtmlScript afaik, but there
> is instance of EVHState with corresponding key, I was unable to track how
> this instance of EVHState gets created, but it is in the map even before the
> first call to the saveStampState method for that request).
>
> We are using Trinidad 1.2.13, I found issue TRINIDAD-2047 in jira which
> seemed to be related to the issue, so I build current 1.2.x branch, but
> problem remains the same (althought 2047 helped a bit, bc stampstate map now
> doesnt contain null values in key1 of DualKey class ).
>
> I understand that's quite difficult to help with such a vague description,
> however it's quite difficult to simulate the issue in some simple scenario.
> I would be gratefull if somehow could explain how StampStates are supposed
> to work, bc code looks quite messy.
>
> We also use Spring Webflow 2.0 and Myfaces implementation 1.2.10 if it
> could be of any relevance.
>
> Thanks
>
> Tibor Blenessy
>
>
>