http://qs1969.pair.com?node_id=11138746


in reply to Re^4: Anyone with XS experience willing to create a high performance data type for Perl?
in thread Anyone with XS experience willing to create a high performance data type for Perl?

> I can switch languages if i need to. If you know of something Perl that can do it as fast as Ruby/C rbtree gem i will be very glad to use Perl.

I have to admit your overall approach feels completely alien to me. :) Is this a personal hobby project? Or for work? - if so, how many programmers at your company?

I don't switch languages lightly because it's not practical to maintain sharp skills across many different programming languages at the same time. In any case, I'd need to get approval from the other programmers in my team before doing so.

And I don't introduce dependencies lightly. What if your dependent module (or any of its dependencies) has a security vulnerability? What if its author abandons it? What if its license restricts you? How quickly can you isolate/troubleshoot a bug in its code? What if you later need to port your software to a platform (or a different Perl version) that your chosen 3rd party library does not support?

I tend to take a much more conservative approach. I'd start by solving the problem, in an easy to understand and maintain style, in plain Perl. If that was fast enough, problem solved. If not, I'd benchmark it to find where the bottleneck is and consider the options, including introducing third-party dependencies. Since both Perl and C++ are approved languages at work, I'd even consider writing the whole thing in plain portable ANSI C++ (again without third-party dependencies). Note, btw, that C++ std::map is usually implemented using red-black trees.

For some background on where I'm coming from see: Why Create Coding Standards and Perform Code Reviews?