This might be much faster if the overlaps are considerably smaller than the total files.

And it avoids any semipredicate problem with $marker.°

(Not heavily tested, please check edge-cases)

use strict;
use warnings;

my $file1 = join "\n", qw( a b c d c );
my $file2 = join "\n", qw(     c d c x );


my $content = "$file2\n$file1";

$content =~ /^(.*)\n.*\1$/s;



(substr $file2,0,length $1)=$file1;

print $file2;
[download]

a
b
c
d
c
x
[download]

°) unfortunately it doesn't, prove left to the interested reader

I added the line start anchor as I wanted to match whole lines.

Agreed, assuming text files, a more "binary" marker is better.

Currently I feel the regex solution is interesting, but still not my first choice. I'll dig deeper if I start profiling.

> Currently I feel the regex solution is interesting, but still not my first choice. I'll dig deeper if I start profiling.

For completeness.

Another approach would be combining a binary search with string equality eq if your priorities tend towards performance.

Of course you could change to a division factor other than 0.5 depending on heuristics.

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}