Fixable rules

[sindresorhus]

I'm aware I could use Remark to modify the Markdown, but I'm looking for a way to add autofixing to remark-lint rules, similar to ESLint.

See how ESLint solves it: http://eslint.org/docs/developer-guide/working-with-rules#applying-fixes

Could be a --fix flag to remark-lint CLI.

[wooorm]

That would be really awesome! And it’s a pretty smart idea, to do it like that. Not entirely sure about al the methods though, why not just start/end position, and new text, so more like:

function finalNewline(ast, file) {
  var contents = file.toString();
  var last = contents.length - 1;
  var message;

  if (last > -1 && contents.charAt(last) !== '\n') {
    message = file.warn('Missing newline character at end of file');
    message.fix = [last, last, '\n']; // start offset, end offset, replacement text
  }
}

From what I gather, ESLint/xo allows running with --fix to apply all fixes. Does it also have a prompt base interface where users are asked whether to apply each fix? Something like:

> remark -u lint --fix-promt

Yields:

- asdasdasd
+ asdasdasd\n
Missing newline character at end of file
Want to apply this fix? (yN)

[sindresorhus]

Not entirely sure about al the methods though, why not just start/end position, and new text, so more like:

The ESLint fixer works on AST, not offsets. This one should do the same.

From what I gather, ESLint/xo allows running with --fix to apply all fixes.

It tries to apply as many fixes as possible:

Note that the fix is not immediately applied and may not be applied at all if there are conflicts with other fixes. If the fix cannot be applied, then the problem message is reported as usual; if the fix can be applied, then the problem message is not reported. - http://eslint.org/docs/developer-guide/working-with-rules#applying-fixes

Does it also have a prompt base interface where users are asked whether to apply each fix?

No. The goal should be to just fix everything without user interaction. I guess it could prompt for ambiguous cases.

I would strongly recommend digging into more deeply how ESLint does it. Discussions in the issue tracker, etc. They've spent a lot of time on it.

[wooorm]

The ESLint fixer works on AST, not offsets. This one should do the same.

Not really, the fix() function returns a {range: [index, index], text: text} (through those helper functions), you can also see them in the json formatter.

No. The goal should be to just fix everything without user interaction. I guess it could prompt for ambiguous cases.

That works really well for some of the rules, like inserting some white-space or a few characters, but not so much for bigger changes (as ESLint mentions in there docs), as they often interfere with other changes.

Plus it’d be cool for fenced-code-flag to try to detect the code language and suggest that, or alternatively use something the user types in!

[dirkroorda]

I do not use —fix because of the existence of prettier.
It seems prettier is also underway for markdown, using the remark-parser, but it currently lacks sufficient output options. It does do word wrap for long lines.
Currently, I use first prettier and then remark for fixing my markdown. It works on the command line, and from within vim, with a single keypress ( the ALE plugin).
So I think fixing based on individual rules might be a waste of time. But you do need a fair amount of options that control how to format an AST.