|Perl: the Markov chain saw|
derby wrote: and also ensure you use the 'o' flag so you're not spinning your wheels always compiling the regexes.
AFAIK, the 'o' flag is only useful if your regex contains variables. If your regex is a constant at compile time, it will only be compiled once in any case.
This regex may not do exactly what you want it to. Specifically, it's not going to remove trailing spaces from the line. The (.*) is greedy, and the following \s* allows a match of 0 whitespace characters, so the whitespace will always end up being matched in the ()'s.
There are many ways to fix this, but the easiest might be to simply match (\S*) instead of the . which some people believe should be avoided at all costs. However, I like derby's solution, to replace all of the regex's with one more general regular expression.
Update: Looking more closely at your code, your biggest bottleneck is almost definitely in the database and not in the perl. If you can avoid interspersing selects and inserts by keeping a copy of data you've already inserted, you'd be able to remove the table indices until it's built, and gain a lot of speed. Alternately, 26k records aren't really that many... you might consider preprocessing the records in memory into a list or hash, and then insert them all at once instead of processing the data in the order it is in the input file.