Forgot to add, you can solve this another way  (taking the abc example):
  grep ‘^...$’ | grep ‘[^a]*a[^a]*’ | grep ‘[^b]*b[^b]*’ | grep ‘[^c]*c[^c]*’
This will scale better for large N and easier to grok and probably what
Dave had in mind!

> On May 29, 2022, at 11:35 PM, Michael Kjörling <michael@kjorling.se> wrote:
> 
> On 29 May 2022 21:47 -0700, from bakul@iitbombay.org (Bakul Shah):
>> You can write a program to generate all permutations and use that as your regexp.
>> For example abc maps to abc|acb|bac|bca|cab|cba. You can rearrange it as
>> a(bc|cb)|b(ac|ca)|c(ab|ba) to see how an n letter permutation is computed from
>> permutations of n-1 letters. I don’t think you can do better.
> 
> I was trying to do something very similar using procmail regexes some
> time ago, only with strings instead of individual characters (for
> parsing and matching on arbitrarily-ordered, comma-separated values
> within a header). I never did come up with a good solution.
> 
> -- 
> Michael Kjörling • https://michael.kjorling.se • michael@kjorling.se
> “Remember when, on the Internet, nobody cared that you were a dog?”
>