Use expression character class instead of lookaround

[Michael Schmidt]

Motivation
Due to the lack of set operation for character classes, I wrote regexes like (?!\s)[\w\P{ASCII}] in the past (another example). They can be transformed to use set operations instead. Using set operation not only makes the intent clearer, it's also more performant. I did some performance test a while back, and (?!\s)[\w\P{ASCII}] was about 4x slower than [\w...whatever ranges are equivalent].

Description
Combine single-character lookarounds with character classes in regexes with the v flag. E.g. (?=[x])[y] => [y&&x] and (?![x])[y] => [y--x]. Regexes without the v should not be affected.

If the element after the lookaround is a character set, then the transformation should still be applied. E.g. (?!\s)\P{ASCII} => [\P{ASCII}--\s].

Examples

// good
/(?!\s)[\w\x80-\uFFFF]/;
/(?!\s)[\w\P{ASCII}]/u;

// bad
bad  = /(?!\s)[\w\P{ASCII}]/v;
good = /[[\w\P{ASCII}]--\s]/v;