Re: how to read multiple line from a file

[TJPride]

If you really need a line-based approach rather than just loading the whole thing into memory (most of the time, the latter approach will work just fine), then you need to cache the previous lines and do a search of the cache:

use strict;
use warnings;

my $csize = 2;   ### Number of lines to save
my (@cache, $str, $c);

my $file = 'data.txt';

open (IN, $file) or die "Can't open $file for read.";
while (<IN>) {
    shift @cache if $#cache == $csize - 1;
    push @cache, $_;
    $c++;

    $str = join '', @cache;
    if ($str =~ /jumped\nover/) {
        print "String found in $file at lines " . ($c - $#cache) . "-$
+c:\n";
        print $str;
        last;
    }
}

Input file data.txt:

the
quick
brown
fox
jumped
over
the
unfortunate
dog

Output:

String found in data.txt at lines 5-6:
jumped
over

[educated_foo]

Just a minor note from the C programmer in me: the easiest and most efficient way to do a fixed-size queue is to use the mod operator:

while (<>) {
    $buf[$. % $size] = $_;
}

[LanX]

As a minor note of a functional programmer, I'll put the logic in an extra iterator like here: Re: Split string after 14 Line Feeds?.

This makes the code much more readable. =)

Otherwise I'm not sure where you want the processing logic to happen in your code, maybe something like unless ($. % $size) { ...} within the loop?

I think If a sliding window is needed, shift and push might be better.

[educated_foo]

If a sliding window is needed, shift and push might be better.

This leads to extra memmove()s if you're lucky (I believe Perl does this), and endless memory consumption otherwise. If you use the mod operator when indexing your array, you get a ring buffer without excess copying. You could probably abstract this out nicely with a bit of extra work.

[ww]

How does this adapt to variable length records/lines/data?

[educated_foo]

It keeps a fixed number of records (as defined by $/). Length has nothing to do with it.

[skyworld_chen]

thanks for all of your replies. I will check these ideas again.