> I've heard many times that it is best to minimize the
> amount of Python code that executes in the critical
> paths of a program; instead, we will get better
> performance by using native libraries (like the re
> module).
Keep in mind that, while the 're' module itself may be written in C, it is in effect an interpreter for another language (regexps). You can easily build a regular expression that will happily backtrack and capture its way into a performance black hole, esp. since you can't trace the individual sub-patterns using normal Python profiling tools.
Benchmarking the two implementations, as you suggest, is of course the only way to be sure in your case.
Benchmarking the two implementations, as you suggest, is of course the only way to be sure in your case.