1. TT parses the page and does all tags, except those containing a
priority
2. I get, say, an array ref of hash refs, corresponding to the tags
for an entire page (since priorities don't necessarily match
layout order),
3. my stuff handles the priorities and make the calls in parallel,
waiting for calls to finish appropriately
4. I pass the array ref back to TT for interpolation, with variables
set in my code available to TT
It would all be much simpler if step 1 was that your code looks at some kind of user preferences data to see what should be on the page, and TT didn't enter the picture until step 4. That would also let you build clean code for generating the content data, which could make smart use of caching for things like your "geo" data. It would be an ordinary forking and IPC programming exercise.
Likely way ahead of me, but if I have ten tags that each take a second, I have a ten second+ load time. Do the same thing in parallel and I am around a second.
You'd be very lucky if that turns out to be true, but it could be close if these tasks are mostly just waiting on I/O from some other source.
Incidentally, sites like my.yahoo.com do all of this ahead of time and just do includes of the cached content when you hit the server. And my.yahoo.com makes heavy use of Perl.
- Perrin
_______________________________________________ templates mailing list [email protected] http://lists.template-toolkit.org/mailman/listinfo/templates
