open the file as usual, set binmode on the handle (optional on UNIX), and read the file into string(s), change the data and print it back.

Is there anything specific you're unsure about?

Even though binmode is indeed optional on *nix systems, it never hurts to use it anyway - on said systems it's just a no-op, but it makes your program more portable.

On the other hand, if you're not aiming for portability, well, yeah, you don't need to bother.

If you're talking about a executable file, you probably don't want to change the length of file. That is, you can't replace 'abc' by 'abcde'. Because some of the bytes will be pointers to other parts of the file, and lengthening the file by those extra 'de' bytes in the middle will throw everything out of kilter.

To be sure that you don't do this, you have two approaches. The main trick is to open the file with '+<', which opens the file for read and write (and doesn't truncate it upon opening). You can either seek to the position you want, and change the contents:

use strict;
use warnings;

my $file = shift or die "no file given\n";
my $offset = 12345;

open my $fh, '+<', $file or die "Cannot open $file for read/write: $!\
+n";

seek $fh, 0, $offset or die "Cannot seek to $offset in $file: $!\n";
print $fh "blah blah blah" or die "Cannot print to $file: $!\n";
close $fh;
[download]

Or, if you don't know exactly at what offset you need to make the change, you'll have to walk through the file until you find what you are looking for. For instance, in blocks of 32 bytes:

$/ = \32;
my $pos = tell($fh);
while( <$fh> ) {
    if ($_ =~ /something/) {
        my $cur = tell($fh);
        seek $fh, 0, $pos          or die "Can't seek: $!\n";
        print $fh "something else" or die "Can't print: $!\n";
        seel $fh, 0, $cur;
    }
    $pos = tell($fh) or die "Can't tell: $!\n;
}
[download]

In files built up out of variable-length blocks, it may well be that you read a small fixed header, which you decode to determine how much you have to read to pull in the rest of the block. In this case you would wind up setting $/ each time through the loop. This brings back memories of decoding WordPerfect files...

The strings don't have the same size. My OS is WinXP. Thanks

What kind of file are you trying to edit? If it's an executable (.exe, .dll), you're out of luck.

Aside from the info you've already received from the other residents of the monastery, here's what I do.

I used to occasionally have to ship executables with a serial number and licensee name in it. I'd never compile it in, as that would be a nuisance. Also, you never know how many (unintentional) cases of the string appear that you want to change.

So, what I used to do is to:

(a) precede the block of strings to patch with a preamble, and

(b) set each string to a known value, and

(c) provide plenty of expansion room in strings with variable length.

So if my fellow brethren will forgive me, here's the kind of code I'd use in C to hold my strings:

const char z_patchable_start[]="Start of patchable area";
const char serial_number[]="$$serial_number.....$$";
const char licensee[]="$$licensee..............................."
                      "........................................."
                      ".......................................$$";
const char z_patchable_end[]="End of patchable area";
[download]

Hope this helps...

--roboticus

Well, you got good answers to a very open ended and nebulous question. Generally however it helps if you provide a context and very often sample code. For a quick "Can Perl ..." question you will find that the Chatter Box is an excellent resource (you have to register, but PerlMonks doesn't require money to change hands).