note 77136 deleted from function.include by danbrown

Note Submitter: eeckart at gmail dot com ---- In reference to the fopen() and $use_include_path workaround for checking whether a include file exists as posted by [arnold at bean-it dot nl]... I did some benchmarks with 100 dirs (each with 1 file) in the include_path on Apache/2.0.55 (Win32) PHP/5.2.1. Here are the results: (note: in the success scenarios, I am include'ing the file in the LAST directory on the include stack. This is to maximize the seek time; fail scenarios simply use a non-existent file; all times are in seconds ) benchmarks --> plain vanilla INCLUDE (success) 0.05515718460083 0.054859161376953 0.053768157958984 (fail) 0.22402501106262 0.17378783226013 0.14510798454285 benchmarks --> fopen() and $use_include_path workaround (success) 0.060588836669922 0.069549798965454 0.056423902511597 (fail) 0.05295991897583 0.039775133132935 0.054499864578247 >>> CONCLUSION In cases involving MANY include paths, include() is -marginally- faster than the workaround (unsurprising - we incur the extra overhead of BOTH the fopen() call AND the $use_include_path seek). I say marginal in comparison to the "fail" scenarios, detailed next.. The most surprising thing is the inefficiency of include() when you are trying for a non-existent file. As compared to the workaround, the native include() method is roughly 3 times slower. Based on additional tests (50 paths only), I note that any decrease in seek time for either case is generally linear to the number of paths. --- THEREFORE, depending on the amount of paths you have, you might want to resort to different methods of optimizing the includes in your scripts, especially if you need to constantly test for the existence of include'd files (our organization uses a templating engine that relies on a virtual flat directory structure using the set_include_path() function). --- From a technical perspective, I have no idea why this is the case and I find the prospect of trawling the PHP source to be rather daunting. However, as an educated guess: I figure that the include() code assumes the file exists, and so wastes processor cycles making a system read call, which inevitably fails. For those interested in the benchmark script, you can email me at the email address provided above. No support, of course. :P