← Back to team overview

randgen team mailing list archive

Performance improvements for the RQG



I have been disappointed that the RQG seems to take all the CPU while it is running, so I set forward to profile the script using Devel::NYTProf and fix any hotspots.

I have now pushed a few commits on the subject. Query generation should be 50% faster than it was before and _string(N) generation should be twice as fast.

The only drawback is that I had to refactor random string generation to use strings as buffers and not arrays. This means that only a-z strings can be generated from now on. I could not find any locations in the code or grammars where non-ascii strings were being used, so I assume it is a worthy trade-off for the major speed improvement that was achieved.

If you have observed any other performance issue that you would like to see fixed, let me know.

Philip Stoev

Follow ups