Re: Regex Help

Yes, some of this is shell dependant. I prefer using [x] over \x for escaping characters in regular expressions -- I just think it is easier to read (though it might make the regex slightly slower if the regex compiler doesn't optimize single-char classes -- I hadn't really worried about that before). It is unfortunate that the convoluted rules for detecting variable names inside regexes think [$] is trying to expand the variable $] so you can't use this trick to escape a dollar sign in a regexp.

In most Unix shells, you can switch between quoting styles to make your life easier:

perl -pi.orig -e 's#<% $flat->purge'"[(]'dynamix','outburst'[)] %>#"'$
+CONFIG{DYNAMIX_OUTBURST}#ge' www/index.html
[download]

Under Win32, you'd use:

perl -pi.orig -e "s#<% $flat->purge[(]'dynamix','outburst'[)] %>#$CONF
+IG{DYNAMIX_OUTBURST}#ge" www/index.html
[download]

- tye (but my friends call me "Tye")

Comment on Re: Regex Help Select or Download Code

Replies are listed 'Best First'.
REx optimization of char class by dchetlin (Friar) on Sep 23, 2000 at 05:25 UTC
Unfortunately, Perl doesn't optimize `[x]`: [~] $ perl -Mre=debug -wle'"x" =~ /x/' Freeing REx: `,' Compiling REx `x' size 3 first at 1 rarest char x at 0 1: EXACT <x>(3) 3: END(0) anchored `x' at 0 (checking anchored isall) minlen 1 Omitting $` $& $' support. EXECUTING... Guessing start of match, REx `x' against `x'... Found anchored substr `x' at offset 0... Guessed: match at offset 0 Freeing REx: `x' [~] $ perl -Mre=debug -wle'"x" =~ /[x]/' Freeing REx: `,' Compiling REx `[x]' size 10 first at 1 1: ANYOF[x](10) 10: END(0) stclass `ANYOF[x]' minlen 1 Omitting $` $& $' support. EXECUTING... Matching REx `[x]' against `x' Setting an EVAL scope, savestack=3 0 <> <x> \| 1: ANYOF[x] 1 <x> <> \| 10: END Match successful! Freeing REx: `[x]' [download] And in case that doesn't convince you: `[~] $ perl -MBenchmark -we'timethese(-5,{x=>q{"x"=~/x/},"[x]"=>q{"x"=~ +/[x]/}})' Benchmark: running [x], x, each for at least 5 CPU seconds... [x]: 4 wallclock secs ( 5.23 usr + 0.00 sys = 5.23 CPU) @ 85 +2787.57/s (n=4460079) x: 6 wallclock secs ( 5.25 usr + 0.00 sys = 5.25 CPU) @ 14 +99188.19/s (n=7870738)` [download] P.S. I think you missed a backwhack in your RExen above; Perl is going to try to interpolate `$flat`.	[reply] [d/l] [select]

Replies are listed 'Best First'.

REx optimization of char class
by dchetlin (Friar) on Sep 23, 2000 at 05:25 UTC

doesn't

[x]

[~] $ perl -Mre=debug -wle'"x" =~ /x/'
Freeing REx: `,'
Compiling REx `x'
size 3 first at 1
rarest char x at 0
   1: EXACT <x>(3)
   3: END(0)
anchored `x' at 0 (checking anchored isall) minlen 1 
Omitting $` $& $' support.

EXECUTING...

Guessing start of match, REx `x' against `x'...
Found anchored substr `x' at offset 0...
Guessed: match at offset 0
Freeing REx: `x'
[~] $ perl -Mre=debug -wle'"x" =~ /[x]/'
Freeing REx: `,'
Compiling REx `[x]'
size 10 first at 1
   1: ANYOF[x](10)
  10: END(0)
stclass `ANYOF[x]' minlen 1 
Omitting $` $& $' support.

EXECUTING...

Matching REx `[x]' against `x'
  Setting an EVAL scope, savestack=3
   0 <> <x>               |  1:  ANYOF[x]
   1 <x> <>               | 10:  END
Match successful!
Freeing REx: `[x]'
[download]

And in case that doesn't convince you:

[~] $ perl -MBenchmark -we'timethese(-5,{x=>q{"x"=~/x/},"[x]"=>q{"x"=~
+/[x]/}})'
Benchmark: running [x], x, each for at least 5 CPU seconds...
       [x]:  4 wallclock secs ( 5.23 usr +  0.00 sys =  5.23 CPU) @ 85
+2787.57/s (n=4460079)
         x:  6 wallclock secs ( 5.25 usr +  0.00 sys =  5.25 CPU) @ 14
+99188.19/s (n=7870738)
[download]

P.S. I think you missed a backwhack in your RExen above; Perl is going to try to interpolate $flat.

[reply]
[d/l]
[select]