Help with Regular Expression

Question

I have been looking for a regular expression with Google for an hour or so now and can't seem to work this one out :(

If I have a number, say:

2345

and I want to find any other number with the same digits but in a different order, like this:

2345

For example, I match

3245 or 5432 (same digits but different order)

How would I write a regular expression for this?

Answer 1

The simplest regular expression is just all 24 permutations added up via the or operator:

/2345|3245|5432|.../;

That said, you don't want to solve this with a regex if you can get away with it. A single pass through the two numbers as strings is probably better: 1. Check the string length of both strings - if they're different you're done. 2. Build a hash of all the digits from the number you're matching against. 3. Run through the digits in the number you're checking. If you hit a match in the hash, mark it as used. Keep going until you don't get an unused match in the hash or run out of items.

Answer 2

I think it's very simple to achieve if you're OK with matching a number that doesn't use all of the digits. E.g. if you have a number 1234 and you accept a match with the number of 1111 to return TRUE;

Let me use PHP for an example as you haven't specified what language you use.

$my_num = 1245;
$my_pattern = '/[' . $my_num . ']{4}/'; // this resolves to pattern: /[1245]{4}/
$my_pattern2 = '/[' . $my_num . ']+/'; // as above but numbers can by of any length

$number1 = 4521;
$match = preg_match($my_pattern, $number1); // will return TRUE

$number2 = 2222444111;
$match2 = preg_match($my_pattern2, $number2); // will return TRUE

$number3 = 888;
$match3 = preg_match($my_pattern, $number3); // will return FALSE
$match4 = preg_match($my_pattern2, $number3); // will return FALSE

Something similar will work in Perl as well.

Answer 3

Put the digits of each number in two arrays, sort the arrays, find out if they hold the same digits at the same indices.

RegExes are not the right tool for this task.

Answer 4

I don't think a regex is appropriate. So here is an idea that is faster than a regex for this situation:

• check string lengths, if they are different, return false
• make a hash from the character (digits in your case) to integers for counting
• loop through the characters of your first string:
  • increment the counter for that character: hash[character]++
• loop through the characters of the second string:
  • decrement the counter for that character: hash[character]--
  • break if any count is negative (or nonexistent)
• loop through the entries, making sure each is 0:
  • if all are 0, return true
  • else return false

EDIT: Java Code (I'm using Character for this example, not exactly Unicode friendly, but it's the idea that matters now):

import java.util.*;

public class Test
{
    public boolean isSimilar(String first, String second)
    {
        if(first.length() != second.length()) 
            return false;
        HashMap<Character, Integer> hash = new HashMap<Character, Integer>();
        for(char c : first.toCharArray())
        {
            if(hash.get(c) != null)
            {
                int count = hash.get(c);
                count++;
                hash.put(c, count);
            }
            else
            {
                hash.put(c, 1);
            }
        }
        for(char c : second.toCharArray())
        {
            if(hash.get(c) != null)
            {
                int count = hash.get(c);
                count--;
                if(count < 0)
                    return false;
                hash.put(c, count);
            }
            else
            {
                return false;
            }
        }
        for(Integer i : hash.values())
        {
            if(i.intValue()!=0)
                return false;
        }
        return true;
    }

    public static void main(String ... args)
    {
        //tested to print false
        System.out.println(new Test().isSimilar("23445", "5432"));

        //tested to print true
        System.out.println(new Test().isSimilar("2345", "5432"));
    }
}

This will also work for comparing letters or other character sequences, like "god" and "dog".

Answer 5

You could do something like this to ensure the right characters and length

 [2345]{4}

Ensuring they only exist once is trickier and why this is not suited to regexes

(?=.*2.*)(?=.*3.*)(?=.*4.*)(?=.*5.*)[2345]{4}

Answer 6

There is an "elegant" way to do it with a single regex:

^(?:2()|3()|4()|5()){4}\1\2\3\4$

will match the digits 2, 3, 4 and 5 in any order. All four are required.

Explanation:

(?:2()|3()|4()|5()) matches one of the numbers 2, 3, 4, or 5. The trick is now that the capturing parentheses match an empty string after matching a number (which always succeeds).

{4} requires that this happens four times.

\1\2\3\4 then requires that all four backreferences have participated in the match - which they do if and only if each number has occurred once. Since \1\2\3\4 matches an empty string, it will always match as long as the previous condition is true.

For five digits, you'd need

^(?:2()|3()|4()|5()|6()){5}\1\2\3\4\5$

etc...

This will work in nearly any regex flavor except JavaScript.

Answer 7

Regular expressions are not appropriate for this purpose. Here is a Perl script:

#/usr/bin/perl

use strict;
use warnings;

my $src = '2345';
my @test = qw( 3245 5432 5542 1234 12345 );

my $canonical = canonicalize( $src );

for my $candidate ( @test ) {
    next unless $canonical eq canonicalize( $candidate );
    print "$src and $candidate consist of the same digits\n";
}

sub canonicalize { join '', sort split //, $_[0] }

Output:

C:\Temp> ks
2345 and 3245 consist of the same digits
2345 and 5432 consist of the same digits