babel-eslint not compatible with no-unused-vars: [2, {vars: local}]

[kentor]

Ok I have this setup:

❯ tree -a -I node_modules
.
├── A
│   ├── .eslintrc
│   └── a.js
├── B
│   ├── .eslintrc
│   └── b.js
└── package.json

A/.eslintrc:

---
rules:
  no-unused-vars: [2, {vars: local}]  # Allow defining globals

A/a.js:

var a = 1;

B/.eslintrc:

---
parser: babel-eslint

B/b.js:

var b = 2;

package.json:

{
  "dependencies": {
    "babel-eslint": "^3.1.19",
    "eslint": "^0.24.0",
    "esprima": "^2.4.0",
    "esprima-fb": "^15001.1.0-dev-harmony-fb"
  }
}

Running eslint on dir A and B gives the expected result:

❯ $(npm bin)/eslint A B

B/b.js
  1:4  error  b is defined but never used  no-unused-vars

✖ 1 problem (1 error, 0 warnings)

however, running it on dir B and A gives an additional unexpected error on a.js:

❯ $(npm bin)/eslint B A

B/b.js
  1:4  error  b is defined but never used  no-unused-vars

A/a.js
  1:4  error  a is defined but never used  no-unused-vars

✖ 2 problems (2 errors, 0 warnings)

This only happens when using babel-eslint as the parser for directory B. I've tried the default parser, esprima, and esprima-fb and it does not throw the error for a.js. The result shouldn't change given the order of directories given to eslint right?

[hzoo]

I don't think this would be a babel-eslint since it's only the parser and changes some tokens etc - It doesn't define any rules or anything. I'm not sure what happens when you try to lint 2 directories with different configs other than in http://eslint.org/docs/user-guide/configuring#configuration-cascading-and-hierarchy - where it merges the config. Maybe that's what is happening here.

[hzoo]

@nzakas - this isn't really a babel-eslint issue right?

[nzakas]

If it works OK with the default parser, and Esprima, then it must be a babel-eslint issue.

[hzoo]

Ok it's because no-unused-vars: [2, {vars: local}] isn't compatible with babel-eslint - nothing to do with different directories/configs. Just do $(npm bin)/eslint B with the babel-eslint parser and the rule and it will error.

[hzoo]

Ok after doing some more looking it seems like it's because we are doing opts.sourceType = "module"; for escope. Based on that I tried

ecmaFeatures:
    modules: true
rules:
  no-unused-vars: [2, {vars: local}]

and it also will error with the regular parser.

[kentor]

I see.. the monkey patching basically affects all subsequent linting. For now I can just change my script to switch the order to have the directory that uses babel-eslint be the last.

[sebmck]

FWIW, Babel is probably going to get the ability to load ESLint rules rather than hacking around ESLint internal APIs and monkeypatching them.

[nzakas]

What would be necessary in the ESLint side to prevent you from needing to monkey patch or otherwise load rules directly?

[sebmck]

There's no way for people using Babel to ever use ESLint as a first class citizen. It'd mean putting all of this into ESLint core, escope etc.

[nzakas]

Yikes! Okay, at some point I'll dig in and see what's going on there.

[hzoo]

Yeah that part is comment attachment, supporting no-unused-vars for es7 (decorators, comprehensions) and flow by changing escope, changing estraverse visitors

[hzoo]

I guess we'll just close this for now with the workaround being what @kentor said:

For now I can just change my script to switch the order to have the directory that uses babel-eslint be the last.