Re: how to benchmark the production code ?

On Thu, Apr 03, 2003 at 06:52:04PM +0200, luro@ze-ro.net wrote: > hello, > > I have installed and tested the Benchmark package of Pear. > I enjoyed it and think it can be a very useful tool. > > I'd like a more "high-end api" for benchmark, relying on the pear classes. > (could be evolution of the benchmark package) > Here is briefly the thing I'd like to do, but for wich I can't find a solution > (may be someone could help me) : > > I'd like to implement Markers in ALL my code. But I'd like to be able to > activate or deactivate those markers because I don't want the pear classes to > work all time, but only when measures are required to solve a specific problem > that as been suspected. > I also don't want to touch the code itself. > > This could be seen as a precompilation level either processing the benchmark > statement or not. > > For some specific scripts there are many work around possible for that. > As an example it's possible to code the pear into specific commented tags that > could be parsed and deleted before benchmark runs. > I did this in some small tests. > I have a benchmark script that call the production script, parse the comment > and then activate benchmark statements, and finally evaluate the php code > (either with eval() or output buffering and include statements) and handle the > benchmarks results. > > This is absolutely "Work Around" and it is not really clean. It may not work > for all scripts, for several reasons (include statements, path and so on ...) > > Does anyone has already been looking for this kind of benchmark behaviors and > have any ideas or possible solutions ? i like dbg: http://dd.cron.ru/dbg/ see http://dd.cron.ru/dbg/dnld/prof_results2.zip for an example of accessing the gathered profiling-info. > ze-ro