"Vadim Gritsenko" <[EMAIL PROTECTED]> wrote: > Pier Fumagalli wrote: > ... > >> So, what do you think? (Surely I need some help to put together the brat) >> >> > > If it provides getKey and getValidity, than it's much better velocity > than velocity in the cocoon-land!
You talking about caching? Yes, I thought about that... A template is no more than a program itself, generating SAX events as its output. As long as the program instructions (the template sources themselves) and the input data (whatever data is available to the template) are valid, then we can safely predict that the output generated by the template will be valid as well. So, for instance, if the Context (a-la Jexl) contains only cachable variables (and therefore, for example, no "request" or "response" object), yes, we can cache the whole thing. That is one of the other few peculiarity in which Garbage differs from Velocity (VNU needs caching desperately)... > <disclamer value="could be irrelevant"> > Have you seen discussions on re-implementation of XSP which suppose to > have representation of XSP as series of SAX events plus some hooks into > programing logic. Idea was by Berin and I don't remember details to > present it here correctly, but it sounds like that if you to implement > yet another templating engine then those ideas might help you solve > efficiency & caching issues for this velocity garbage... > :) > > Ok, it was "callback style xml handlers", thread is here: > http://marc.theaimsgroup.com/?t=101620413700006&r=1&w=2 > </disclamer> Yes, in fact the Garbage tree and processor can be easily reused for this approach. The only thing different, at this point, would be the parser and event-tree builder. Pier
