Comments on: One More Erlang Wide Finder

By: Matt

Matt — Mon, 22 Oct 2007 11:22:27 +0000

Good stuff. I’ve just started digging into Erlang, and have been doing tons of random code-reviews at work ;-)

It is interesting to see the work you are doing with Erlang. Keep it up!

By: steve

steve — Sun, 21 Oct 2007 15:15:52 +0000

Or better yet, see tbray16.erl and wfbm4.erl — they're faster than everything else I've posted to date.

By: steve

steve — Sun, 21 Oct 2007 04:44:53 +0000

Matt: in my previous comment I was wrong that the bug you found didn’t affect the results. It required another fix as well, which decreases the performance of this solution by about 50%, which obviously is significant. Thus, all the more reason to go with the newer tbray15.erl and wfbm3.erl as mentioned in the previous comment.

By: steve

steve — Sun, 21 Oct 2007 01:38:01 +0000

Matt: a most excellent catch. Yes, that is a typo. It’s funny — I stared at that a few times because it looked funny, and yet never caught it.

For this test case, it doesn’t change the results or the performance. Comparisons are not done against that string; it’s used only to find shift values for advancing through the file. For the o1000k.ap dataset, it doesn’t matter because you get the same answer either way.

I’ve fixed this in the posted wfbm.erl file. But you want to grab tbray15.erl and wfbm3.erl instead anyway; they’re about 50% faster.

Thanks, Matt!

By: Matt

Matt — Sat, 20 Oct 2007 22:40:15 +0000

Quick question:

Is this a typo?

-define(STR, “GET /ongoing/When”).
-define(REVSTR, “mehW/gniogno/ TEG”).

“mehW” != “When” reversed because “m” != “n”. If you really are doing a comparison with REVSTR, I’d expect it to fail the match every time.