Monday, May 14, 2012

Next Talk: Lies, Damn Lies, and Benchmarks (part 2)

PM's have brought up issues with using perl on the jobs. We've all heard some version of "Do it in shell,
there's too much overhead in starting up perl", "Perl takes too long to write", or something similar.

This all applies until you have to actually debug ten levels of plumbing in the shell. Or until you actually
look at the kernel overhead of passing all of the piped data through the shell.

There are some things that are more easily done in shell, but many people avoid perl for the wrong reasons.

This month we will look at how to benchmark Perl and perl, a few things to look for when you run
the perl compiler on your system, and how to manage the tradeoff between multi-layer shell plumbing and perl startup.

One of the issues is benchmarking shell accurately using, of course, perl and Benchmark. Another is
looking whether it really takes "too long" to write non-trivial Perl vs. shell.