Re: Regex matching on anywhere but the beginning or end of a line

[fruiture]

Well, you can achieve that without regexp:

my $string = q{"string (12" or 1 foot or 1') into 6" MySQL varchar"};
local $_ = 1;
while( ($_ = index($string,'"',$_)) !=-1
       and ++$_ < length($string)
   ){
      substr($string,$_++,0,'"');
}

--
http://fruiture.de

[xmath]

But why would you want to do that? Regexes are one of perl's major strengths, and they can deal with this situation very nicely.

Your code snippet (idx) also seems to be slower than both regex-based solutions given above (zwa, sub):

       Rate  idx  sub  zwa
idx 73430/s   --  -5% -15%
sub 77672/s   6%   -- -10%
zwa 86374/s  18%  11%   --

(Benchmarked the same way as this one)

[fruiture]

Although I agree that probably a regexp is more straight-forward, I get different results benchmarking the stuff.

my $string = q{"string (12" or 1 foot or 1') into 6" MySQL varchar"};

benchmark(
        shift,
        q{"string (12"" or 1 foot or 1') into 6"" MySQL varchar"},

        'substr' => sub {
                my $t = $string;
                substr($t, 1, -1) =~ s/"/""/g;
                $t
        },

        's///' => sub {
                my $t = $string;
                $t =~ s/(?<=.)"(?=.)/""/g;
                $t
        },

        'index' => sub {
                my $t = $string;
                local $_ = 1;
                while( ($_ = index($t,'"',$_)) !=-1
                               and ++$_ < length($t)
                                      ){
                              substr($t,$_++,0,'"');
                }
                $t
        },
);

benchmark() is just a wrapper around Benchmark::cmpthese() that checks the results of the routines.

# run with perl 5.6.1

EXPECTED   : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"
index      : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"
s///       : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"
substr     : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"

Benchmark: timing 1000000 iterations of index      , s///       , subs
+tr     ...
index      : 19 wallclock secs (19.66 usr +  0.02 sys = 19.68 CPU) @ 5
+0813.01/s (n=1000000)
s///       : 19 wallclock secs (19.85 usr +  0.00 sys = 19.85 CPU) @ 5
+0377.83/s (n=1000000)
substr     : 23 wallclock secs (24.13 usr +  0.00 sys = 24.13 CPU) @ 4
+1442.19/s (n=1000000)
               Rate substr      s///        index
substr      41442/s          --        -18%        -18%
s///        50378/s         22%          --         -1%
index       50813/s         23%          1%          --

# and with perl 5.8.0

EXPECTED   : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"
substr     : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"
index      : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"
s///       : "string (12"" or 1 foot or 1') into 6"" MySQL varchar"

Benchmark: timing 1000000 iterations of index      , s///       , subs
+tr     ...
index      : 18 wallclock secs (18.71 usr +  0.01 sys = 18.72 CPU) @ 5
+3418.80/s (n=1000000)
s///       : 33 wallclock secs (31.67 usr +  0.02 sys = 31.69 CPU) @ 3
+1555.70/s (n=1000000)
substr     : 35 wallclock secs (35.24 usr +  0.01 sys = 35.25 CPU) @ 2
+8368.79/s (n=1000000)
               Rate substr      s///        index
substr      28369/s          --        -10%        -47%
s///        31556/s         11%          --        -41%
index       53419/s         88%         69%          --

--
http://fruiture.de

[xmath]

Okay, I re-ran the benchmark with your exact code on my computer, and I got:
Read more... (1391 Bytes)