I haven't got a solution to demonstrate what I want to do yet, because I'm still
looking at the current Result mechanism to see how it functions. However you're
right that it might not be worth the effort - I'm sure I'm not aware of all the
dependencies that tie into the mechanism. Implications for REST handling and for
OGNL and theme handling come to mind.
Here's a catalogue of current solutions to the absence of a 'view logic' layer:
* create a 'straight-through' Action mapping using just ActionSupport, and chain
the input result to it
* fetch data from model in your action and store it in the session
* use taglibs to retrieve data for the dropdowns or lists they render
* use <s:action> taglibs to retrieve the data and place it somewhere accessible
* use Preparable interceptor
* use Tiles interceptors
There were some fixes proposed too, extending Preparable, or implementing a
FLASH scope to make the process quicker and more efficient.
It's a tribute to the flexibility of the framework that it could be done in so
many different ways, but the purist and idealist in me thinks of the situation
where a struts newby is asking what to do in this situation. It would be nice to
say, 'This is where it goes, job done' rather than have to explain a work-around.
I'd like to see the idea taken on-board, even if it's just put on a back-burner.
If my speed of development picks up (it can't get much slower at the moment - I
find myself unintentionally at the wrong end of the learning curve again) then
I'll definitely put together something myself by way of a demo. At the moment
though I shall have to make do with an ugly child class extending TilesResult.
Best regards
Adam
Jeromy Evans on 10/01/08 01:32, wrote:
Adam Hardy wrote:
Is it right to assume that no-one is interested by the idea of extending
the S2 Result to allow extra configuration or classes at this location in
the architecture?
Hi Adam, I haven't been convinced yet that there's a strong argument for this
compared to alternative approaches yet. Or rather that the effort doesn't
seem justified for the amount I'd use it. However one related use-case I've
yet to find a good approach for is when using the REST plugin and a
ContentTypeHandler to determine which result handler is executed: For the
same result: - when the requested content type is HTML, a view is returned
that often does need access to additional data (eg. options for a select) -
when the requested content type is XML or JSON, the result does not need that
additional data to be prepared In this case it makes sense that only the HTML
result itself, or the view components themselves, should perform the
additional logic to prepare the HTML view. The action itself definitely
should not be doing it (but does).
Your proposal may be the best way to solve that but may need some more
thought within that context.
[snip]
Adam Hardy on 02/01/08 23:55, wrote:
That was the standard struts 1 approach, IIRC. My grudge with the other
solutions I think also applies to this: it's not a natural solution, it's
another square peg in a round hole.
This is why I'd like to see something in Struts2 that addresses this
fundamental duality of purpose in the HTTP request.
Adam
Brian Pontarelli on 02/01/08 22:50, wrote:
However, this still requires some composition or inheritance to support
multiple actions using the same preparation right?
-bp
Musachy Barroso wrote:
What I do is that I create an "input" map to the input method on the
same action class for which validation is failing. That method is
empty (from ActionSupport), and forwards to the page that will show
the data. On this page I use the normal tags backed by the action. If
multiple tags call the same method then I "cache" the results on the
action. I think your problem can easily be solved by making "input"
go thru an action, instead of forwarding to the page straight. (but
you already knew this :) ) <redundant-email />
regards musachy
On Dec 31, 2007 11:41 AM, Adam Hardy
<[EMAIL PROTECTED]> wrote:
Happy New Year everybody.
This issue is mostly due to the validation failure mechanism which
passes flow direct to the 'input' result without giving a chance
to code the data retrieval needed to get data for dropdowns,
associated lists, etc etc which the view/JSP will need.
Currently the assorted solutions to this all seem to be forcing
round pegs into square holes. For instance, I could make the
'input' result an action chain to go onto another action which does
the data reading. Or I could fetch the data via an <s:action> or a
taglib.
The S2 documentation says in various places things like:
"First, we need to change or query the application's state, and
then we need to present an updated view of the application. The
Action class manages the application's state, and the Result Type
manages the view."
Three or four years ago, this issue with the view was discussed
alot. There was talk of mechanisms termed 'view-controllers' and
concepts such as 'view logic'.
I'd love to see this accommodated for in S2.
There is a certain amount of coding I can do to achieve my goals in
the Results, but it may not be the best place for it - the name
'Result' implies more of a link between the Action and the View,
rather than a place for coding data retrieval.
Essentially I think there is a strong call for a Class or chain of
classes that can be tied to each particular View, whether Tile,
JSP, velocity or whatever.
This is obviously not what Results were designed for. I can do it
currently but the S2 config allows only one class per ResultType -
so effectively I'd need one ResultType per JSP, or some pattern for
it.
The sort of operations I'm thinking of:
- retrieving lists, sometimes parameterized (e.g. a list of items
allowed in a particular category - requires the categoryId)
- caching lists (countries in ApplicationScope, personalized data
in session for example)
- localization of dropdown beans (i.e. country names)
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
