is around thirty times faster

Actually this dependent on the size of @patterns. When I benchmark them against each other i get more like a 100% speedup. Precalculating a hash of the patterns and then doing a hash lookup is infinitely faster.

Testing 100
Benchmark: running bigrex, forex, hash, each for at least 1 CPU second
+s...
    bigrex:  1 wallclock secs ( 1.09 usr +  0.02 sys =  1.11 CPU) @ 19
+103.70/s (n=21186)
     forex:  1 wallclock secs ( 1.11 usr +  0.00 sys =  1.11 CPU) @ 44
+43.24/s (n=4932)
      hash:  2 wallclock secs ( 1.14 usr +  0.00 sys =  1.14 CPU) @ 16
+32288.34/s (n=1862441)
            Rate  forex bigrex   hash
forex     4443/s     --   -77%  -100%
bigrex   19104/s   330%     --   -99%
hash   1632288/s 36636%  8444%     --
Testing 10000
Benchmark: running bigrex, forex, hash, each for at least 1 CPU second
+s...
    bigrex:  1 wallclock secs ( 1.06 usr +  0.00 sys =  1.06 CPU) @ 83
+.80/s (n=89)
     forex:  2 wallclock secs ( 1.03 usr +  0.00 sys =  1.03 CPU) @ 42
+.68/s (n=44)
      hash:  1 wallclock secs ( 1.05 usr +  0.00 sys =  1.05 CPU) @ 13
+97043.89/s (n=1464102)
            Rate    forex   bigrex     hash
forex     42.7/s       --     -49%    -100%
bigrex    83.8/s      96%       --    -100%
hash   1397044/s 3273428% 1666934%       --
Testing 100000
Benchmark: running bigrex, forex, hash, each for at least 1 CPU second
+s...
    bigrex:  1 wallclock secs ( 1.08 usr +  0.00 sys =  1.08 CPU) @  7
+.42/s (n=8)
     forex:  2 wallclock secs ( 1.17 usr +  0.00 sys =  1.17 CPU) @  4
+.27/s (n=5)
      hash:  2 wallclock secs ( 1.03 usr +  0.00 sys =  1.03 CPU) @ 12
+19285.16/s (n=1257083)
            Rate     forex    bigrex      hash
forex     4.27/s        --      -43%     -100%
bigrex    7.42/s       74%        --     -100%
hash   1219285/s 28579944% 16429768%        --
[download]

So id say you learned something new today too. :-) (use a hash...)

---
demerphq

_{<Elian> And I do take a kind of perverse pleasure in having an OO assembly language...}