Re: [SMARTY-DEV] Replacing include with include_once in Smarty::fetch()?

This is an interesting idea...but are you using a PHP accelerator? That should speed-up includes considerably. Further, the scenario you describe sounds hardly typical and even if you are generating many emails in bulk (hmmmmmmmmmm) I would assume that you wouldn't be facing many concurrent users but rather simply having to sporadically endure a longer running process. In other words, optimizing here seems unwarranted. Never-the-less, I think it would be possible to simulate such an effect as you describe using messju's really cool compiler plugin that you can find in this thread http://www.phpinsider.com/smarty-forum/viewtopic.php?t=291 (which is on recursion, but you need not use it recursively). I don't mean to be impolite, but I sure hope this isn't some spam related thing ;) xo boots --- Matthias Pigulla <mp@webfactory.de> wrote: > Dear smarty-developers, > > I'm extending a mailing application to allow for sending HTML emails. > I > use Smarty as template engine to build personalized e-mail messages > (Smarty outputs html as well as text mime parts). > > Right now, I'm doing some performance tests and have to call > $smarty->fetch a few thousand times (once for every personalized > e-mail > being sent), which is a very realistic - if not even small - setup > for > my purposes. > > The problem is that the include($_smarty_compile_path) around line > 1258 > in Smarty.php makes up a huge share of total execution time and also > consumes a lot of memory, although it just includes the same compiled > template again and again. I could imagine that the PHP parser will > analyze/precompile the file upon every inclusion? I'm not that > familiar > with PHP internals... > > As a quick experiment, I wrote a postfilter to add "<?php function > foo(&$this) { ?>" and "<? } ?>" around the compiled template and > changed > Smarty::fetch() to "include_**once**(...); foo($this);". > > That way I could reduce the execution time spent in that code region > to > a fraction of 1/8th, memory consumption is reduced by about 3.5MB. I > suppose that longer template code and more repetitions would even > yield > better improvements. > > I know that usually you would hardly fetch the same template > repeatedly > during the same request, so maybe that problem only occurs with my > very > specific setup. > > However, what do you think of automatically placing template code in > a > function body? The function name could be derived from the resource > or > file name. Note that this would require calling the function instead > of > just including the compiled template; however, as long as > Smarty::fetch > is the only section of code including templates, that should not be a > problem. > > Any ideas what else might break? Possible benefits? > > Best regards, > Matthias