Hehe....Yes calling the render methods is still not ideally efficient...

It is still easy to test a component/page..Or at least "fairly" easy.
Under org.apache.tapestry.test.util (I think) there is a Creator class
that will let you create an instance of a component/page with
properties (even of your abstract methods) setup via a map or array
parameter.

You could then pass in a mock request cycle, with a NestedMarkupWriter
instance, which will allow you to capture the string data written to
it's internal buffer and then assert whatever you want to assert with
that data. I plan on adding some documentation on this as well :)

The service idea is a possible workaround, but depending on how speedy
all of these things get expedited I really do want to attack the
direct update + component identity problems as the first order of
business, if at all possible. It does affect me as well :)

In the interim though, there is enough infrastructure available in
tacos alone that I can make PartialFor work with things they way that
we want. I didn't want to do it before because it would be another one
of those "you can use it - but.." scenerios, but it would be a good
chance to play with stuff in a smaller-impact environment anyways.

Does this sound good/reasonable?
On 11/22/05, Leonardo Quijano Vincenzi <[EMAIL PROTECTED]> wrote:
> Jesse Kuhnert wrote:
> > You are mostly understanding correctly, besides the listener reloading
> > the entire page in the response, sort of... :)
> >
> Yes but no... The listener isn't reloading the entire page, but it is
> indeed *rendering* the whole page.  Am I right?
> Two points:
>
> 1) Direct component refresh is not reliable. That's until the "use with
> care" warning is removed, hehe...  Now if we have a Direct refresh that
> 'just works' (I can put it in a @For or anywhere and it would work)
> that'd be great. But the "component service" solution is also useful.
>
> 2) I'm thinking of in-container component testing. I'd call the
> "component direct" service, or whatever is called, and use some HTML
> testing tool to check my components. Now, components really *shouldn't*
> have any external dependencies besides the ones define in its interface
> (the JWC file, annotations, etc), right? Or I could try to manually test
> a component by calling it directly and setting the parameters on some
> kind of component-test web interface.
>
> If the component update is rendering the whole page, isn't this process
> expensive? Even if it's cached...
> (and if it's not cached, the devoleper time is also valuable.. you know)
>
> --
> Ing. Leonardo Quijano Vincenzi
> Director Técnico
> DTQ Software
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to