regexp: extracting info

[jeanluca]

Dear Monks,

I've the following issue:

$str = abc dddd    10 ;
$str1 = "abc dddd ee    10" ;

I need to extract "10" however there might be 2 or 3 words preceding. Somehow I cannot construct the regexp. My quess was something like:
$number = $str =~ /abc\sdddd\s[ee]\s+(\d+)/gs ;
but it didn't work:(
Furthermore I realized that I don't understand the ?:, ?= etc patterns. I tried to read the perlre but that didn't help very much
Any suggestions for a good tut or an other post ?

Thanks in advance
Luca

[prasadbabu]

If i understood your question correctly, this will work. Recently Roy Johnson has explained about this in his tutorial Using Look-ahead and Look-behind which will help you to understand '?=' etc.

$str = 'abc dddd    10 ';
($number) = $str =~ /(\d+)\s*$/;
print "$number";

[Joost]

Any suggestions for a good tut or an other post ?

Well, there's always perlrequick and perlretut. :-)

[tirwhan]

You almost have it. Two things are wrong:

1. You need to put $number into list context to retrieve $1
2. Read up the bit on character classes again in perlre, they allow you to match a range of single characters, you still need to quantify them if you want to match more than once. [ee] and [e] are equivalent and will both match a single character "e".

this will work:

($number) = $str =~ /abc\sdddd\s+(?:ee\s+)?(\d+)/;

As for your question about the two extended operators, they influence capturing of an expression in brackets. perlre really explains it the best way I know, though you may want to take a look at Mastering Regular Expressions which is a very worthwhile book to read when learning regular expressions.

Update:Removed "gs" as they're not necessary (thanks to prasadbabu for pointing that out)

[ikegami]

Noone clearly explained your problem, IMHO, so allow me.

1) While square brackets (as in [ee]) usually refer to optional components in BNF and command usage syntax, they mean something else in regexps. Append a question mark to make an atom (a non-special character, a character escape, something in brackets or something in parenthesis) optional (e.g. a?, (?:regexp)?). In this case, [ee] should be changed to (?:ee)?.

2) Regexps do not return captured values in a scalar context.
my $number = $str =~ /.../;
should be
my ($number) = $str =~ /.../;

3) Did you really mean to say /g? Also, /s is useless since there's no "." in your regexp.

 

Here are some solutions (probably already mentioned):

From the front:
/^(\w+\s+){2,3}(\d+)/
From the back:
/(\d+)\s*$/
If it's the only number:
/(\d+)/

[ptum]

Well, your description of the extraction rules and what you expect to encounter are somewhat sparse, but if the number will always be at the end of the string, something simple might do the trick for you:

use strict;
my $str = "abc dddd     10 ";
my $number;
if ($str =~ /^.*?(\d+)\s*$/) {
  $number = $1;
}
else {
  # do something with unexpected record
}

Alternatively, you may be looking for two or three words, like this:

if ($str =~ /^(\w+\s+){2,3}(\d+)/) {
  $number = $2;
}

(code not tested)

[doctor_moron]

Furthermore I realized that I don't understand the ?:, ?=

Perl for Dummies's book said
if for some odd reason you want to suppress $1 or $2 to prevent them from getting assigned, add ?: after the left paranthesis.

example

$test = 'NESTING';
$test =~ /(?:N.)S(.I)/;
print "$1 \n";

when you run this program, Perl displays "TI"

zak

[planetscape]

These are my most frequently used resources for constructing/understanding regexes:


Mastering Regular Expressions
perlrequick
perlretut
perlre
YAPE::Regex::Explain
GraphViz::Regex
The Regex Coach (Win32 only, IIRC)

HTH,

[jeanluca]

The problem I have with this string is a little bit miss understood. Assume you have a very large multi line string, and only one line looks likt describe above!
Thanks for the tut suggestions!

Luca

[pKai]

If you want to match a line in a multiline string, you need the m-modifier and the line-begin (^) and line-end ($) indicator.
You can read about those in the perldocs already meantioned.

Going with your specification so far:

I need to extract "10" however there might be 2 or 3 words preceding.

and

a very large multi line string, and only one line looks likt describe

Going with this, i would suggest:

$str =~ /^(?:\w+\s+){2,3}10$/m

There might be other restrictions you want to place on your match (more specific whitespace preceding the 10? optional whitespace (or even other words) after the 10?)

We don't know. You have to make up your mind, what the appropriate pattern is.