Read it line by line, but keep track of which section you're in and if the section hasn't change, just concatenate. Then, when the section changes, go ahead and process the last section, then set the current section to the new one.

my ($section, @value);
while (<FILE>) {
    chomp;
    my ( $temp_section, $temp_value ) = /^(?:([^:]+):)?\s*(.*)/;

    if ( $temp_section ) {
        if ( $section ) {
            # Process the old value somehow.
        }
        $section = $temp_section;
        @value = ( $temp_value );
    }
    else {
        push @value, $temp_value;
    }
}
[download]

Don't slurp 600M - that translates to roughly 3G+ of RAM.

Not if you slurp to a scalar.

---

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

Lingua non convalesco, consenesco et abolesco.

Rule 1 has a caveat! -- Who broke the cabal?

It's still 600M + 41 bytes. :-)

---

My wife's blog

I hadn't considered slurping for that exact reason. The regex is what gave me some ideas. I've never done read ahead regexes. I figured someone here at PM would steer me in the right direction.

You sample code is exactly what I am looking for. Thanks!

If your files are small, then slurping and using a regex with /s and a lookahead is the easiest way:

#! perl -slw
use strict;

my $data = do{ local $/; <DATA> };

print "'$1'$2'\n" while $data =~ m[([a-z-]+):(.*?)\n(?=[a-z-]+:|$)]sg;

__DATA__
aut-num:       AS19710
as-name:       ASN
descr:         S4R
admin-c:       SNE1
tech-c:        SNE1
import:        from AS3356 63.215.71.1 at 63.215.71.2 action pref=20; 
+med=50;
               from AS3356 63.215.86.133 at 63.215.86.134 action pref=
+50; med=150;
               accept ANY
import:        from AS3847 action pref=10;
               accept ANY
export:        to AS3847
               announce AS19710
export:        to AS3356
               announce AS19710
notify:        nwcontact@email
mnt-by:        S4R
changed:       andy@email 20010502
source:        LEV
[download]

At each iteration of the while loop, $1 will be the section header, and $2 the body of the section with all the whitespace intact. You can further process $2 to remove or reduce the whitespace as required.

P:\test>448390
'aut-num'       AS19710'

'as-name'       ASN'

'descr'         S4R'

'admin-c'       SNE1'

'tech-c'        SNE1'

'import'        from AS3356 63.215.71.1 at 63.215.71.2 action pref=20;
+ med=50;
               from AS3356 63.215.86.133 at 63.215.86.134 action pref=
+50; med=150;
               accept ANY'

'import'        from AS3847 action pref=10;
               accept ANY'

'export'        to AS3847
               announce AS19710'

'export'        to AS3356
               announce AS19710'

'notify'        nwcontact@email'

'mnt-by'        S4R'

'changed'       andy@email 20010502'

'source'        LEV'
[download]

The files can be large, up to 600MB in one case. I hadn't considered slurping.

Thanks for the ideas!