I solved "match strings with no unique characters", on domain ^.*$, in PCRE, in 33 characters (so obviously without enumerating the charset).

And it uses a widely applicable technique, easily generalizable to string-lists with no unique strings, for example.

Regexes available if you want to see them.

@Grimy I would like to see that please. (Unique characters. As for the perfect powers one, I'm going to try that later.)

^((.)((?<=(?=\n|\2.*\2|(?3)).)))*$

This is emulating variable-length look-behind by recursively look-behinding a single char

Hahaha! Nice. I would have guessed PCRE would be boring and protect against that somehow.

Why do you use \n instead of ^?

Not the same

\n never matches, since the domain is ^.*$

But without it, PCRE complains “recursive call could loop indefinitely”

There are lots of ways around that, including wrapping the entire (?=foo) as (?(?=foo)|(*F))

How about \0

That would give wrong results for strings that contain \0

I don't think PCRE can process \0 in a string, because it's written in C, which uses \0-terminated strings.

\n however, can be inside ^.*$

In non-/m mode

/m? /s is what causes . to match \n, and it’s not the default

By default . is a synonym for \N

/m is what causes ^ and $ to match on newlines in addition to the beginning and end of the string.

all /s does is make . match newlines, which you can do without it using \X or [\s\S] or [\d\D] etc.

Even with /m, you can still match the true beginning and end of the string with \A and \z.

My point is, without /s, ^.*$ cannot contain newlines

Since . doesn’t match newlines

I specified the domain as ^.*$ in order to exclude newlines

My point is, it's less universal than it can be. \0 would be more universal.

I challenge you to get PCRE to parse a \0 inside a string

[^\0]* isn’t any more universal than .*

Why not just use ^((.)((?<=(?=.^|\2.*\2|(?3)).)))*$

Although for full strings, even when not in /m mode, ^((.)((?<=(?=.^|\2.*\2|(?3)).)))*\z is needed

Oh right .^ is much shorter than .(?!)

@Grimy Published this as a method for matching a character occurring exactly once.

Yay