in reply to Finding Line numbers in a file

According to my Camel, "each time a pattern successfully matches (including the pattern in a substitution), it sets the $`, $&, and $' variables to the text left of the match, the whole match, and the text right of the match."

That sounds useful. 

my $text = <<'END_OF_TEXT';
line 1 apple
banana line 2
line cherry 3
END_OF_TEXT
    ;
while ( $text =~ m/(apple|banana|cherry)/ig ) {
    my $word = $1;
    my $prelines = ( $` =~ tr/\n// );
    printf qq{Word "%s" found on line %d\n}, $word, $prelines + 1;
}
__END__
Word "apple" found on line 1
Word "banana" found on line 2
Word "cherry" found on line 3

[download]

If you use English, the $` variable is called $PREMATCH (see perlvar, which notes that using this variable "imposes a considerable performance penalty on all regular expression matches").

Replies are listed 'Best First'.
Re^2: Finding Line numbers in a file
by reasonablekeith (Deacon) on Apr 04, 2007 at 15:29 UTC
    Nice, but inefficient, and gets worse the bigger the text file is.

    Do not do this, use the others, they increase in a linear proportion with the size of the text file, and do not require entire file to be loaded into memory.

    ---
    my name's not Keith, and I'm not reasonable.
[reply]
      You are possibly right. You are just as possibly wrong. There are several things that we don't know, such as:
      • Average line length. Shorter lines means more lowlevel iterations.
      • Average file length. Longer files will require more memory - but that is about all.
      • Average hit count. How often is the string found in the file.
      • Average hit placement. How often does the string end up at the beginning or the end.
      • Implementation issues. Is the string passed in already in one chunk or do we have access to a file handle.
      There are just too many unknowns to use blanket statements as to which algorithm is best.

      But one thing that is a major issue is that the special regex capture variables shouldn't be used. They impose too much penalty. Instead though you can use @- and @+ which have no penalty. As in the following: 

      my $str = "1 one
2 two
3 one
4 four
5 one
6 five";
my $last_pos = 0;
my $newlines = 1;
while ($str =~ /(one)/g) {
    $newlines += substr($str, $last_pos, $-[0] - $last_pos) =~ tr/\n//
+;
    $last_pos = $-[0];
    print "Found on line $newlines\n";
}
# prints 
# Found on line 1
# Found on line 3
# Found on line 5

      [download]


      Notice the optimization that only counts newlines from the previous match.

      my @a=qw(random brilliant braindead); print $a[rand(@a)];
[reply]
[d/l]
[select]
      Dear kyle and reasonablekeith,
      Thanks for suggestion and warning also. This is making me think in new directions.
[reply]

In Section Seekers of Perl Wisdom