I'm going to answer that. But first, an introduction. Run these chunks of code as separate programs.

"japhy" =~ /a/;
eval q{ print $& };
[download]

The above code prints nothing.

"japhy" =~ /a./;
eval q{ print $& };
[download]

The above code prints "ap".

"japhy" =~ /a/;
print $&;
[download]

The above code prints "a".

Confused yet?

Here's the worst part:

@& = (1,2,3);
"japhy" =~ /a/;
eval q{ print $& };
[download]

That code prints "a".

---

From the above examples, you should extract some things:

• some regexes don't set $&
• other regexes do
• if Perl sees you use $&, it'll set $& all the time
• if Perl sees you use ANY variable named &, it'll set $& all the time

So what's going on? Well, the regex /a/ is too simple to be a regex, so Perl does a Boyer-Moore search instead. But the regex /a./ isn't a BM-type search, so it's a regex. And as long as it really goes through the regex engine, Perl sets up the $& functionality.

But if Perl sees you using $&, it makes everything that looks like a regex act like a regex, and set up the $& functionality. Bah. That's irritating, and "bad".

This just in: I patched gv.c so that only $& sets the PL_sawampersand flag to a true value. Now you can use @& without worry.

_____________________________________________________
Jeff[japhy]Pinyan: Perl, regex, and perl hacker.
s++=END;++y(;-P)}y js++=;shajsj<++y(p-q)}?print:??;