Copying two words from a string into two variables

Question

I want to get the variables $color and $number from a string that in general is like this: "something, numColor (number)". The color might be W, U, B, R, G. If there is no color the variable color should be C if the string before the comma doesn't have the word land or L otherwise. If there is more than one color the variable $color should be M. Here are some examples of what the string may look like and what the variables should be:

• Sorcery, R (1) $color=R, $number=1
• Creature — Beast 5/3, 4G (5) $color=G $number=5
• Sorcery, 1WWU (4) $color=M $number=4
• Legendary Land $color=L $number=0
• Artifact, 0 $color=C $number=0
• Legendary Creature — Eldrazi 15/15, 15 (15) $color=C $number=15

Answer 1

You can do task by sed on-liner:

sed '/(\?\([0-9]\+\))\?$/s//; number=\1/
     t n
     s/$/; number=0/
     :n
     /^.*, [0-9]\?\([WURBG]\)/{
                               s//color=\1/
                               s/[WURBG]\{2,\}/M/
                              }
     /[Ll]and/s/^[^;]*/color=L /
     /color/!s/^[^;]*/color=C /' file

But I'd like to offer put commands in script file:

#!/bin/sed -f
/(\?\([0-9]\+\))\?$/s//; number=\1/
t n
s/$/; number=0/
:n
/^.*, [0-9]\?\([WURBG]\)/{
    s//color=\1/
    s/[WURBG]\{2,\}/M/
}
/[Ll]and/s/^[^;]*/color=L /
/color/!s/^[^;]*/color=C /

then do

sed -f script.file file

Please note that all above regex are tested on your example only and if you meet unproper works its can be settled by a little tune up.

Answer 2

While you can do this sort of thing by wrapping the more primitive Unix tools — grep, sed, awk, etc. — in a shell script, this sort of problem really wants to be handled in a full programming language that has a powerful regular expression system. Personally, I'd reach for Perl:

#!/usr/bin/perl -w
use strict;

my $line = 0;
my ($junk, $color, $number);

open my $data, '<', 'data.txt' or die "open: $!\n";
while (<$data>) {
    chomp;
    ++$line;

    if (m/Land/) {
        print "color=L, number=0\n";
    }
    else {
        ($junk, $color, $number) = m/, (\d+)?([WURBG]+) \((\d+)\)$/;
        if (defined $color and defined $number) {
            $color = 'M' if length($color) > 1;
            print "color=$color, number=$number\n";
        }
        else {
            ($junk, $number) = m/, (\d+)? ?\((\d+)\)$/;
            if (defined $number) {
                print "color=C, number=$number\n";
            }
            else {
                print "Line #$line is malformed!\n";
            }
        }
    }
}

data.txt contains this:

Sorcery, R (1)
Creature — Beast 5/3, 4G (5)
Sorcery, 1WWU (4)
Legendary Land
Artifact, (0)
Legendary Creature — Eldrazi 15/15, 15 (15)

There's just one difference from what you posted: the "Artifact" line in your question doesn't have parentheses around the 0 value, which would require an exception to be made in the parser. This can be added, but I don't see why that's better than fixing the format of the data file.

I assume you don't simply want the color and number values printed out, as this script does. You would put your own code in for each of the print lines.

The $junk bit comes from my assumption that the digit before the color letter(s) might be significant. I'm using it to help the parser do its thing. If you have a real use for this digit, you can rename the $junk variable to have semantic meaning. It's only "junk" to me, because I don't know what the value means.

---