Rule Change: Add onlyOneSimpleParam: boolean option to no-confusing-arrow

[JLHwung]

What rule do you want to change?

no-confusing-arrow

What change to do you want to make?

Generate fewer warnings

How do you think the change should be implemented?

A new option

Example code

() => 1 ? 2 : 3;
(a, b) => 1 ? 2 : 3;
({ a }) => 1 ? 2 : 3;
([ a ]) => 1 ? 2 : 3;
(...a) => 1 ? 2 : 3;

What does the rule currently do for this code?

Warn "Arrow function used ambiguously with a conditional expression."

What will the rule do after it's changed?

I propose a new onlyOneSimpleParam option with false default value.

When this option is set to true, the rule will not report any errors if one of the following is satisfied:

1. There are more than 1 arguments or no arguments
2. The argument is not an identifier

The intention of no-confusing-arrow is to check whether a BinaryExpression is incorrectly written as an arrow expression. When the arrow contains no argument, the expected BinaryExpression () >= 1 is never valid so we don't have to throw at all. Actually I think this behaviour should be enabled unconditionally.

When the arrow contains more than one argument: the expected BinaryExpression (a, b) >= 1 ? 2 : 3 is rare use case so we don't throw

When the arrow param is non-simple: the expected BinaryExpression ({ a, b }) >= 1 ? 2 : 3 is even rarer so we don't throw.

Note that all the current rule example code gives only an arrow expression with exactly one simple parameter. We could enable this option by default in future majors.

Participation

• I am willing to submit a pull request to implement this change.

Additional comments

Context: babel/babel#14201 (comment)

[nzakas]

I think such an option would make sense.

@eslint/eslint-tsc thoughts?

[btmills]

Just to clarify, the idea is that with this option, the rule would only report arrow functions with a single argument that is either an identifier or a primitive literal?

I’d be fine not warning for 0-argument arrows by default, either as part of this or a separate change.

[Gautam-Arora24]

I would like to work on this issue :)

[JLHwung]

@Gautam-Arora24 Sure. I didn't have bandwidth yet. Please feel free to claim it.

[nzakas]

Just to clarify, the idea is that with this option, the rule would only report arrow functions with a single argument that is either an identifier or a primitive literal?

That’s correct.

I’d be fine not warning for 0-argument arrows by default, either as part of this or a separate change.

For simplicity, I’d say let’s do it as part of the new option for now. We can tinker with defaults later if we think it’s necessary.

[btmills]

Makes sense to me. Since mdjermanovic left a 👍 on the original post, marking as accepted. You're good to go, @Gautam-Arora24!