unicorn/better-regex autofix breaks "match all" in multiline regexps

[FloEdelmann]

This regexp is autofixed by unicorn/better-regex like this:

- /lorem(?:.|\n)*?ipsum/m
+ /lorem[.\n]*?ipsum/m

This breaks the regex:

• (.|\n) means "any character, including a newline character"
• [.\n] means "a period character or a newline character"

Instead, it should be fixed to [^], according to https://stackoverflow.com/a/16119722/451391:

- /lorem(?:.|\n)*?ipsum/m
+ /lorem[^]*?ipsum/m

[RunDevelopment]

While this is definitely a bug, your proposed expected fix is incorrect. In JS RegExp, the dot (.) is equivalent to [^\r\n\u2028\u2029] (unless the s flag is enabled), so (?:.|\n) is equivalent to [^\r\u2028\u2029]. So the correct fix would be:

- /lorem(?:.|\n)*?ipsum/m
+ /lorem[^\r\u2028\u2029]*?ipsum/m

(Whether this fix is desirable is a different question.)

[sindresorhus]

This should be reported at https://github.com/DmitrySoshnikov/regexp-tree, which is what does the regex transformations.

If you can figure out which transform that causes this, it might be possible to disable that specific transform in the mean time:

eslint-plugin-unicorn/rules/better-regex.js

Line 34 in ade13ac

optimized = optimize(original, undefined, {blacklist: ignoreList}).toString();

[FloEdelmann]

I've opened an upstream issue: DmitrySoshnikov/regexp-tree#217