[COFF] Re: Grep has gone overboard

[Bakul Shah <bakul@iitbombay.org>]

> On May 30, 2022, at 1:16 AM, Ralph Corderoy <ralph@inputplus.co.uk> wrote:
> 
> Hi Bakul,
> 
>> 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.
> 
> Being lazy, I tend to use a shell's brace expansion and then whittle it
> down to generate permutations.  I keep thinking it's surprising there's
> no Bell Labs program to produce them given there's things like
> factor(1).

For permutations I just use this k program:

$ cat perm.k
p:{:[1<x;,/(>:'(x,x)#1,x#0)[;0,'1+_f x-1];,!x]}
perm:{x@p[#x]}
$ k perm.k
  perm "abc"
("abc"
 "acb"
 "bac"
 "bca"
 "cab"
 "cba")
I won’t bother explaining but p n returns all permutations of
indices 0..n-1. Then perm is easy to construct. Similarly easy to
insert | between strings but your solution is much more in the spirit
of regexp and faster than my brute force solution!

Also, thanks for catching the anchoring error!

> 
> The brace expansion is wasteful as it's going from n**l down to n! and
> it can result in the old bash here not responding to SIGINT for a while
> if it's producing gazillions.
> 
>    $ printf '%s\n' {b,a,d,u,g,l,y}{b,a,d,u,g,l,y}{b,a,d,u,g,l,y}\
>> {b,a,d,u,g,l,y}{b,a,d,u,g,l,y}{b,a,d,u,g,l,y}{b,a,d,u,g,l,y} |
>> egrep -v '(.).*\1' |
>> fgrep -x -f - /usr/share/dict/words
>    ladybug
>    $
> 
> -- 
> Cheers, Ralph.