merlyn's on-target comment aside, I'd normally do that by assigning the repeated pattern to a variable and interpolating that into the regex pattern.

# original: my $pat = "\d\s+foo";
# fixed:
my $pat = "\\d\\s+foo";

my $regex = qr/$pat[^\r\n]*$pat/;
[download]

Not sure about PCRE, but whatever programming language you use, I suspect you can construct a string using variables similarly to build up to your pattern.

Update: Doh! Thanks, ikegami. I was trying to show the subpattern with strings for the cross-language analogy and slipped up.

That should be

my $pat = "\\d\\s+foo";
[download]

or better yet

my $pat = qr/\d\s+foo/;
[download]

Your pat incorrectly matches "dssssfoo", and doesn't match "1 foo" as it should.

my $pat = "\d\s+foo";
print("$pat\n"); # ds+foo
[download]

Wow, you guys are amazingly quick to reply and helpful in those replies. That said, since I need to do this outside of Perl, in a pure PCRE expression, neither this nor ikegami's post is helpful.

Seeing as how I just looked and found no PCREmonks, however, does anyone know if such a forum actually exists and I'm just bad at Googling, or if a forum with such a purpose exists under some different name/URL?

Thanks Again,
Alex

Wow, you guys are amazingly quick to reply and helpful in those replies. That said, since I need to do this outside of Perl, in a pure PCRE expression, neither this nor ikegami's post is helpful.

I disagree. We didn't use anything Perl-specific. For example, here it is in JScript:

var pat = "\\d\\s+foo";
var regexp = "(" + pat + ")[^\r\n]*(" + pat + ")";
var re = new RegExp(regexp);
re.exec(some_string);
[download]

(Untested)

At least you were nice enough to announce this as a PCRE problem. We've had past inquisitors ask about PHP and Java regexen without prior disclosure, which only leads to broken useless answers because we (sensibly) presume a Perl environment.

Oops. You're looking for PCREmonks, not Perlmonks.

Darn. I knew it was too easy to reply with:

my $regex = qr/(\d\s+foo)/;
$some_string =~ /$regex[^\r\n]*$regex/;
[download]

-QM
--
Quantum Mechanics: The dreams stuff is made of

That doesn't capture properly. Fix:

my $regex = qr/\d\s+foo/;
$some_string =~ /($regex)[^\r\n]*($regex)/;
[download]

Also, your solution doesn't scale (if capturing is desired). He can't say, for example, the following while capturing what $regexp matches:

my $regex = qr/\d\s+foo/;
$some_string =~ /(?:$regex[^\r\n]*)*($regex)/;
[download]

Oh...I didn't realize PCREmonks existed. I'll go ask them. Thanks for being friendly in the meantime. :-)

Alex

Is there a way of taking, say,

/(\d\s+foo)[^\r\n]*(\d\s+foo)/
[download]

What about:

my $re = qr/\d\s+foo/;
sub match {
  local $_ = shift;
  /(\d\s+foo)[^\r\n]*\1/x       and print "1\n";
  /\d\s+foo  [^\r\n]*\d\s+foo/x and print "2\n";
  /($re)     [^\r\n]*\1/x       and print "3\n";
  /$re       [^\r\n]*$re/x      and print "4\n";
}
match '4  foo4  foo';   ## easily matched
match '4  foo5  foo';   ## a bit trickier

__OUTPUT__

1
2
3
4
2
4
[download]

We can see that all 4 regexps matched the first text but in order to match the second, we can't use the backreference and must resort to other things. I think isolating the common part and using it where necessary ("4") is best.

Taking advantage of specific features of your regexp,

@matches = map { /^(\d\s+foo)/; $1 } /(?:\d\s+foo[^\r\n]*)/g;
[download]

This will validate and return matches. If you just wish to validate, the following will suffice:

# Specific case:
/(?:\d\s+foo[^\r\n]*)/;

# General case:
$repeated = qr/\d\s+foo/;
/(?:$repeated[^\r\n]*)$repeated/;
[download]

Update: My capture version has two issues. And the pattern is still repeated.

You really do want a parser for this:

list  : term list_  { [ $item[0], $item[1], @{$item[2]} ] }

list_ : sep term    { [ $item[1], @{$item[2]} ] }
      |             { [ ] }
[download]