Fix #227: treat "any single character" token as special in groupSingleCharsToCharClass

[Maxim-Mazurok]

Fixes #227
and fixes sindresorhus/eslint-plugin-unicorn#1231

[DmitrySoshnikov]

@Maxim-Mazurok thanks for the PR. Can we confirm all the tests pass, and do we need to handle other symbols?

[Maxim-Mazurok]

Can we confirm all the tests pass

Confirming that all tests are passing:

do we need to handle other symbols?

I couldn't really think of anything that would have similar characteristics...

For example /(\d|[\n\r])/gm gets optimized into /([\d\n\r])/gm which is correct. Basically, we only need to care about symbols that mean different things depending on [] context.

ChatGPT suggested to check ^, - and \ which all seem to work fine. For example, - works fine because of this if I understand correctly:

regexp-tree/src/optimizer/transforms/char-class-classranges-merge-transform.js

Lines 91 to 102 in 091d4e9

	/**
	* @param {Object} expression - Char or ClassRange node
	* @returns {number}
	*/
	function getSortValue(expression) {
	if (expression.type === 'Char') {
	if (expression.value === '-') {
	return Infinity;
	}
	if (expression.kind === 'control') {
	return Infinity;
	}

So no, I don't think we need to handle other symbols.

[DmitrySoshnikov]

Looks good, thanks for the contribution!

[Maxim-Mazurok]

Hi @DmitrySoshnikov , looks like this PR was supposed to be included in v0.1.25 however I can't find expected "meta" word in https://www.unpkg.com/[email protected]/dist/optimizer/transforms/group-single-chars-to-char-class.js and the issue is still reproducible in v0.1.25.

Perhaps it needs to be rebuilt and republished? Thank you!

[DmitrySoshnikov]

@Maxim-Mazurok thanks for the heads up, yes, seems the build script is broken and didn't recompile the code (in fact, npm run build doesn't work anymore, I'll take a look).

[DmitrySoshnikov]

@Maxim-Mazurok should be fixed in v0.1.27. Please let me know if you see any issues.