http://www.perlmonks.org?node_id=1031672


in reply to Re^2: Selecting HL7 Transactions
in thread Selecting HL7 Transactions

"The parentheses are used for the repeat factor. All my research on the internet indicates a multi-character pattern that needs to be repeated multiple times should be enclosed in parentheses. Is this not correct?"

Here's a test showing clustering and capturing. Both match as expected. Capturing also sets $1. 

$ perl -Mstrict -Mwarnings -E '
    my $re1 = qr{PV1\|1\|O\|(?:[^|]*\|){3}\|};
    my $re2 = qr{PV1\|1\|O\|([^|]*\|){3}\|};
    my $x = q{PV1|1|O|F3|F4|F5|F6|F7};
    my $y = q{PV1|1|O|F3|F4|F5||F7};
    say "------- Clustering -------";
    say "Match in \$x" if $x =~ /$re1/;
    say $1 if $1;
    say "Match in \$y" if $y =~ /$re1/;
    say $1 if $1;
    say "------- Capturing -------";
    say "Match in \$x" if $x =~ /$re2/;
    say $1 if $1;
    say "Match in \$y" if $y =~ /$re2/;
    say $1 if $1;
'
------- Clustering -------
Match in $y
------- Capturing -------
Match in $y
F5|

[download]

-- Ken

Replies are listed 'Best First'.
Re^4: Selecting HL7 Transactions
by BillDowns (Novice) on May 02, 2013 at 01:02 UTC

    On a different note, since you seem quite knowledgeable about Perl, and again referring to google searches, non-greedy matching is usually defined in a manner that  (.*?\|) and  ([^|]*\|) should be equivalent. And I think so, too. Why are they not?

[reply]
[d/l]
[select]

      Ignoring the issue with newlines and the "s" modifier that I alluded to earlier, the heart of the matter is that "." matches any character while "[^|]" matches any character except the pipe character.

      Taken in isolation, /(.*?\|)/ and /([^|]*\|)/ may well produce the same result: 

      $ perl -Mstrict -Mwarnings -E '
    my $x = q{A|||||Z};
    my $dot_re = qr{(.*?\|)};
    my $cc_re = qr{([^|]*\|)};
    $x =~ $dot_re;
    say $1;
    $x =~ $cc_re;
    say $1;
'
A|
A|

      [download]

      The reasons they do this, however, are different. "A" is the least number [non-greedy] of zero or more of any characters (".*?") that match before a literal pipe character ("\|"). It just so happens that "A" is also the greatest number [greedy] of zero or more non-pipe characters ("[^|]*") that match before a literal pipe character ("\|"). So, in both cases "A|" is captured.

      Now consider the following where the capture groups are no longer in isolation: 

      $ perl -Mstrict -Mwarnings -E '
    my $x = q{A|||||Z};
    my $dot_re = qr{(.*?\|)Z};
    my $cc_re = qr{([^|]*\|)Z};
    $x =~ $dot_re;
    say $1;
    $x =~ $cc_re;
    say $1;
'
A|||||
|

      [download]

      Here, "A||||" is the least number [non-greedy] of zero or more of any characters (".*?") that match before a literal pipe character ("\|") that is immediately followed by a literal Z character: "A||||" plus "|" are captured. However, "" (i.e. nothing) is the greatest number [greedy] of zero or more non-pipe characters ("[^|]*") that match before a literal pipe character ("\|") that is immediately followed by a literal Z character: "" plus "|" are captured.

      I recommend you take a look at Regexp::Debugger which provides a visualisation of Perl's regular expression engine in action — I think you'll find it most enlightening.

      I'd also recommend you look at the Perl documentation (available online at http://perldoc.perl.org/perl.html) before reaching for an internet search engine. Here's a list of Perl regular expression documentation that you'll find linked from that page:

      • perlrequick — Perl regular expressions quick start
      • perlretut — Perl regular expressions tutorial
      • perlfaq6 — Perl frequently asked questions: Regexes
      • perlre — Perl regular expressions, the rest of the story
      • perlrebackslash — Perl regular expression backslash sequences
      • perlrecharclass — Perl regular expression character classes
      • perlreref — Perl regular expressions quick reference

      That's the order the links appear on that page: look at them in whatever order you want. To be honest, I was a little surprised there was so many; had I realised in advanced, I might not have chosen to start enumerating them all here.

      -- Ken

[reply]
[d/l]
[select]

        Frankly, I find the Perl documentation quite badly written years ago and I gave up on it.

        Back to the non-greedy matching, I understand what you are saying. It's just that's not the normal and usual understanding of the word. To my mind, and I would bet most of the world, since the expression was (*.?\|) that would mean the least number of characters before the next - stress that next - \|.

        That's what "non-greedy" should mean, IMO. Perl's implementation still greedy.

[reply]
Re^4: Selecting HL7 Transactions
by Anonymous Monk on May 02, 2013 at 00:46 UTC
    Thanks for that info - I did not know about clustering. It does not show up in the first half-dozen google searches on regular expressions. In this case, it doesn't matter - the utility script doesn't care about any capturing. It is simply qualifying.
[reply]
      Sorry - forgot to log in first. That was me :)
[reply]

In Section Seekers of Perl Wisdom