fast string parser: regex versus substr

mce has asked for the wisdom of the Perl Monks concerning the following question:

Hi All,

I want to find a way to parse a string in a real performant way. This is what I came up with.
I found 2 methods so far, but there must be a better way to do it.
The string is position delimited, i.e. from the 3the to the 16the it contains something, and so far. Now, I want only the text in these fields, not the blanks.
Let me show you

our @data=<DATA>;
 
# some code comes here...
sub dosubstr {
   foreach my $i ( 0..$#data ) {
      my $line=$data[$i];
      my $jcpu=substr($line,2,16);
      my $j=substr($line,18,48);
      my $s=substr($line,290,16);
      $jcpu =~ s/\s//g;
      $s =~ s/\s//g;
      $j =~ s/\s//g;
      #  warn "$jcpu $j $s";
      # ..store the values in a hash, but that is not important here
      }
   }
 
sub doregex {
   foreach my $i ( 0..$#data ) {
      my $line=$data[$i];
      $line =~ m/^04(?=(\S+)).{16}(?=(\S+)).{40}.{216}.{16}(?=(\S+)).{
+16}/ ;
      my $jcpu=$1;
      my $j=$2;
      my $s=$3;
      #  warn "$jcpu $j $s";
      # ..store the values in a hash, but that is not important here
      }
   } 

__DATA__
04A12345          RELEASE
 
                A12345          RELEASE         A12345
04FTOP            DD_BUIL+
 
        
        FTOP            DD_REKL+        FTOP
04FTOP            DD_PLAN+
 
                FTOP            DD_REKL+        FTOP
[download]

Now, in a simple benchmark study, the doregex function is 3 times faster than the substr. But it just look so complex doesn't it.
So, I am asking the wisdom for my fellow monks to make it more performant.
I am talking about a data of thousands of lines and every second counts, as my operators don't like to wait for webpages :-) Thanks in advance,
Update: fixed substr value to correct Abigail-II comment
---------------------------
Dr. Mark Ceulemans
Senior Consultant
BMC, Belgium

Comment on fast string parser: regex versus substr Download Code

Replies are listed 'Best First'.
Re: fast string parser: regex versus substr by liz (Monsignor) on Nov 20, 2003 at 16:18 UTC
You might want to have a look at unpack. Benchmarks between the use of unpack() and substr() were discussed previously in Unexpected: unpack slower than several substr's. Why?. Liz	[reply]
Re: fast string parser: regex versus substr by dragonchild (Archbishop) on Nov 20, 2003 at 16:19 UTC
Read up on unpack. For fixed-width data, it's the fastest Perl has to offer. ------ We are the carpenters and bricklayers of the Information Age. The idea is a little like C++ templates, except not quite so brain-meltingly complicated. -- TheDamian, Exegesis 6 ... strings and arrays will suffice. As they are easily available as native data types in any sane language, ... - blokhead, speaking on evolutionary algorithms Please remember that I'm crufty and crochety. All opinions are purely mine and all code is untested, unless otherwise specified.	[reply]
Re: fast string parser: regex versus substr by Abigail-II (Bishop) on Nov 20, 2003 at 16:22 UTC
Well, at least one of the approaches is incorrect. I uncommented the warnings, turned them into prints and added the following to your program: `dosubstr; print "---\n"; doregex;` [download] Running your program gives me: `A12345 RELEASE A1 2345RELEASEA12345 FTOP DD_BUIL+ FT OPDD_REKL+FTOP FTOP DD_PLAN+ FT OPDD_REKL+FTOP ---` [download] Abigail	[reply] [d/l] [select]
Re: fast string parser: regex versus substr by mce (Curate) on Nov 20, 2003 at 16:55 UTC
Thanks all, I came up with. `sub dopack { foreach my $i ( 0..$#data ) { my $line=$data[$i]; my ($jcpu,$j,$s)=(unpack('@2A16A40A232A16', $line))[0,1,3]; warn "$jcpu $j $s"; # ..store the values in a hash, but that is not important here } }` [download] And it won the benchmark competition :-) I never really understood pack and unpack, but it am getting to like it. --------------------------- Dr. Mark Ceulemans Senior Consultant BMC, Belgium	[reply] [d/l]
Re: Re: fast string parser: regex versus substr by bart (Canon) on Nov 20, 2003 at 22:24 UTC
`my ($jcpu,$j,$s)=(unpack('@2A16A40A232A16', $line))[0,1,3];` [download] I think you'd be better off using the "x" template to ignore the third field (with index 2): `my ($jcpu,$j,$s)= unpack('@2A16A40x232A16', $line);` [download] You can replace the leading '@2' with 'x2'', too. Or, just the reverse, replace the 'x232' with '@290'. I don't think it'll make much difference speedwise.	[reply] [d/l] [select]
Re: fast string parser: regex versus substr by Art_XIV (Hermit) on Nov 20, 2003 at 16:47 UTC
unpack, as the other monks have stated, is almost undoubtedly the way to go if your data is fixed-width. It's a moot point, but your benchmarks may have been spurious since you didn't s/// whitespace in the doregex function like you did w/ dosubstr. Hanlon's Razor - "Never attribute to malice that which can be adequately explained by stupidity"	[reply]


Perl-Sensitive Sunglasses
	PerlMonks