Looks like the strings you are trying to match are utf-16, but burried in a binary file. I'd recommend you use binmode on the file handle you are using to read the data and then you can:

use warnings;
use strict;
use Encode;

my $binstr = "\x{00}\x{01}\x{02}\x{03}\x{04}\x{05}" .
  "\x{00}A\x{00}u\x{00}t\x{00}h\x{00}o\x{00}r\x{00}" .
  "\x{80}\x{90}\x{a0}\x{b0}\x{c0}\x{d0}\x{e0}";
  
my $matchStr = encode ('utf16be', 'Author');

if ($binstr =~ /(\Q$matchStr\E)/) {
    my $match = decode ('utf16be', $1);
    
    print "Found $match\n";
}
[download]

Prints:

Found Author
[download]

Note that this assumes big endien which seems to match your example, but could be little endien which is native for Windows systems and normal for the net.

I've tried variously, use utf8; use various encodings but my matches don't capture the strings I'm seeking.

But grandfather is most likely right, you're trying to find Unicode strings inside a binary file, so treating the whole file as 16-bit Unicode, using binmode or open to set the encoding of the filehandle to 'ucs2le', for example using

    open IN, '<:encoding(ucs2le)', $file
[download]

So you could try grandfather's approach, which is a very sensible one, or you could do the inverse, and convert the strings you're searching for into UCS-2LE, and search the binary file using that.

Actually, I suspect that if indeed Unicode strings start at odd file (or buffer) positions, grandfathers method will fail to find them.

BTW A plain Perl, non Encode way to convert plain Latin-1 to UCS-2LE is using pack/unpack:

    $ucs2 = pack 'v*', unpack 'C*', $text;
[download]

Actually, I suspect that if indeed Unicode strings start at odd file (or buffer) positions, grandfathers method will fail to find them.

Interesting thought. However I checked it out with the sample code by inserting an extra byte before the 'Author' string and the match string was still found.

On reflection Perl doesn't know anything special about either the match string or the buffer being matched so the fact that there is meta information (the fact that it is actually utf-16) associated with the data is of no consequence.

---

DWIM is Perl's answer to Gödel

Many thanks to all; I'll give those a try. I don't think I tried UCS-2, certainly not UCS-2LE.

Donald.

I am now looking at what I had finally written, and I've clarified my question in my own mind to ask "how does the R.E. engine handle the metacharacters in a non-text environment."

Grandfather's example's \Q...\E led me to enlightment in the perlreref

Many thi^Hanks
Donald.

IDENT(1) -- 1993/11/09 -- GNU

NAME
    ident - identify RCS keyword strings in files

SYNOPSIS
    ident [ -q ] [ -V ] [ file ... ]

DESCRIPTION
    ident   searches   for   all   instances  of   the   pattern
    $keyword: text $  in the  named files  or, if  no  files are
    named, the standard input.
[download]

Ident is exactly what I am trying to emulate on this "platform." Unfortunately, ident doesn't seem to find these double-octet encoded strings as I thought it would, so I turned to Perl. ident does work fine on the source code, of course. (I use a _lot_ of cygwin to get me through the day in this Micros**t shop.) Maybe I did something wrong, though with ident that's hard to do :)

Thanks for the thought; I'll try again.

Donald.