@Grimy Did you notice that I linked a text file above?

Oh right, I noticed and then forgot to look at it x)

I implemented a brute force search for that, but it only works up to 7 chars so far because 1) RME doesn't optimize things like (x*)*z yet, and 2) Once you start getting {NNN} and larger, without generalizing the treatment of it somehow, there would be too many sets of too large a size.

It didn't find anything that I hadn't already found manually.

Oh so up to 7 is definitely solved, cool

Err, clearly not very well solved though, since it’s missing \B

Which matches ~{1} in 2 chars

You currently have ~{1} in 5 chars with ^$|xx

Yes, I hadn't added \b and \B yet to the brute forcer, and I just added them and it found that :) and then I came to this window and saw you noticed that too

\b i’m gonna conjecture is never needed

\B should find some other uses

The brute forcer doesn't find anything else with \B up to 7

@Grimy How can you say that, when one of your own regexes uses \b? (Floor logarithm in base 2)

Granted, that's a full function, not just true/false, but it could be used within a larger regex

Oh, maybe not, because in a larger regex you just want to capture something, not return it as a match

Hmmm

Yeah I’m not very confident about \b not being needed

z should be provably never needed for non-empty sets, though

z+ = z, z* = empty, z? = empty, (z) = z

And by recursively applying those rules it’s possible to convert any pattern containing a z into either a pattern that doesn’t contain z or just z

I guess all the circumstances in which (?!) or (*F) are needed in PCRE are not applicable to ECMA.

Yeah backtracking verbs are what makes (?!) useful

If there are any backtracking verbs in P, P z is no longer equivalent to just z

Also recursion would make it much harder to prove anything with the above method

Oh, I missed a 9 char one: ^(?!x{N})

Oops never mind that

yeah

there’s shorter to match this set

I already realized that was covered by ^x{N,M}$

Probably should note things like that in the file

@Grimy BTW, I verified your 379 length abundant numbers regex up to 10000 :)

What is the shortest regex to match even perfect numbers?

@H.PWiz I think it would check for the corresponding Mersenne prime.

Find a k for which 2^k-1 is prime (using logarithm in base 2) and 2^(k-1)*(2^k-1) equals the input.

Yeah, I had a go, but I still don't have the hang of multiplication. Mine was60 long

i have 57

and 379 to match all perfect numbers

wait no 53 not 57

^(?=(x(x*?))(\1\1)+$)((x*)(?=\5$))+(?!(xx+)\6+$)\1\2$

Oh, of course log2 is not needed. I hadn't given it much thought. Very nice 53. I see how it works. :)

I would suggest posting it as a noncompeting answer

Damn chat won't let me paste that as code, so it ate the spaces that align the comments

Better commented version: regex for matching even perfect numbers.txt