[geeklog-devel] caching template library

[Tony Bibbs]

I'd only point out the obvious, by delaying again and again moving to a more supported template system we now eat the responsibility of maintaining that code.  It may be that Joe's improvements to PHPLib aren't that much work to maintain (which would be good), but I still feel at some point it will be an inevitable move to something like Smarty, Flexy, et. al.  

The common argument is "what about all the themes out there that'd have to change?".  While it is a valid concern, we only support one theme now (a good decision I'd add) and I feel the rich features of some of the better supported template systems will encourage theme developers to convert or build new themes.

I guess, really, it's up to the vision in everybody's mind.  If the goal is to keep the GL codebase from having too many drastic changes then that's fine, though, I'd question that rational alone.  As much as we don't compare ourselves to other CMSs, this is an area that GL is woefully behind in and I'm concerned the proposed solution doesn't go far enough.  Sure, compilation is nice...but what about magically escaping output to help combat nastiness like XSS? Using method calls in templates?  Native translation support?  Simple control structures?

I don't pretend the move to a template library like one of those would be easy...it'll be a LOT of work.  But given the activity amongst developers lately this is surely something that could happen in a fair amount of time.  I think we'd have to considering supporting the most current stable release in addition to the release with the new template system to give time for people to make the move.  My point is the logistics of it all is manageable.

Generally I agree in doing small bits of work to incrementally improve a product but there are times when that just won't do and I'd simply ask if that time isn't now?

--Tony

----- Original Message ----
From: Michael Jervis <mjervis at gmail.com>
To: Geeklog Development <geeklog-devel at lists.geeklog.net>
Sent: Thursday, November 1, 2007 7:34:35 AM
Subject: Re: [geeklog-devel] caching template library

> I'd like to hear some comments, please, from those who initially tried
> to speed up or otherwise improve the template class. Mike and Vinny, I
> think, did the most work there. Guys, did you have a chance to play with
> Joe's changes?

I had a play with them in the initial state, and it did show a marked
improvement in performance, however, there were some functional issues
that Joe has now resolved and I haven't re-run my perf tests on it due
to lack of time.

My approaches were to totally replace the templating library with
allegedly faster alternatives, with an adapter pattern. This resulted
in slower geeklog. I then tried re-factoring chunks of geeklog (where
looping was used) with the faster alternatives (and an adapter
pattern) and got an improvement, but not a huge one.

Joe's approach is to improve the performance of PHPLIb by compiling
it's templates and caching them, it does speed things up


Mike
_______________________________________________
geeklog-devel mailing list
geeklog-devel at lists.geeklog.net
http://eight.pairlist.net/mailman/listinfo/geeklog-devel