Re: Long line of OR

At 07:16 2011-10-12, LuKreme wrote:
>* 9876543210^0 ! VAR ?? 
>(aa|apsumo|beni|candy|casa|cf|cpr|db|dbl|dhosts|fb|glue|gofobo|hita|ic|llc|logi|micr|netf|ocb|plus|pl|rcc|rk|rocb|scv|snap|spot|su|tor|tripa|twit|v365|vm|root|admin|rsmith|bsmith)
>
>and it occurred to me there might be a reason not to do that.

Questions generally end with a query mark (?).  I take it this posed 
as a question?

You're using maximal scoring, so as soon as the regexp evaluates 
true, it'll evaluate the condition as true and skip all other 
scorings.  However, the entire regexp has to be parsed.  You MIGHT 
get somewhat better performance by splitting it into multiple scoring 
lines, with more frequently encountered elements in the earlier 
conditions, but I expect you'd need to run a bunch of benchmarked 
tests to see a timing difference.  The length of VAR will no doubt 
have an impact.

You may however see more of an improvement by optimizing your regexp:

>(a(a|psumo|dmin)|b(eni|smith)|c(andy|asa|f|pr)|d(b|bl|hosts)|fb|g(lue|ofobo)|hita|ic|l(lc|ogi)|micr|netf|ocb|p(lus|l)|r(smith|oot|cc|k|ocb)|s(cv|nap|pot|u)|t(or|ripa|wit)|v(365|m))

Readability takes a big dive though.  Yes, there are some other 
optimizations tat could be made, but that's the broad strokes.

Again, youd need to run benchmarks - put recipes in a sandbox, run 
against a large corpus several times, omitting the longest and 
shortest times (owing to the effects of cache, etc), and would need 
to run it on a host which didn't have other demands on it.

---
  Sean B. Straw / Professional Software Engineering

  Procmail disclaimer: <http://www.professional.org/procmail/disclaimer.html>
  Please DO NOT carbon me on list replies.  I'll get my copy from the list.