Page 2 of 2

Re: Conditional includes or...

Posted: Tue Jan 15, 2008 2:51 pm
by Christopher
I agree with your last statement. My point is that the statistics you quote are for some very specific configurations -- and other configurations report very different results. So it's not largely a performance question (as Greg says). Likewise as I note, the conclusions of that discussion were about solving PEAR2's problems, which are quite possibly very different than many users would have. The maintenance issues are to provide flexibilities for different kinds of styles -- which individuals and teams would not have by definition.

Re: Conditional includes or...

Posted: Tue Jan 15, 2008 3:01 pm
by Ambush Commander
Why can't people just do an analysis of how the low-level code actually works? There are a number of "expensive" fs calls the bytecode cache has to do, and some variation of how the OPs are generated based on class context, but nothing so complicated that a theoretical analysis wouldn't be possible. I know performance is largely a practical matter, but if empiric tests are giving different results, it seems sensible to me.
The maintenance issues are to provide flexibilities for different kinds of styles -- which individuals and teams would not have by definition.
Well, maybe big teams might...

Re: Conditional includes or...

Posted: Tue Jan 15, 2008 3:28 pm
by Christopher
I agree. I just want to be sure that people understand that PEAR2's problems with many include_onces called multiple times, and the need for file organizations flexibility, might not be their problems. Yahoo probably has a whole team dedicated to performance tuning, and its problems may or may not have much in common the people here. Both include_once and __autoload are reasonable solutions. There are pros and cons to both that can be weighed given the circumstances.