prefer-const should autofix, but it didn't

[aladdin-add]

let acorn = require("acorn"),
    assert = require("chai").assert;

to

const acorn = require("acorn"),
    assert = require("chai").assert;

[platinumazure]

I think we already have autofix in this rule. There are some conditions that prevent autofix.

Is there a particular use case that you think should be autofixed, but that isn't being autofixed?

[aladdin-add]

yes, I created an example

should this be an enhancement, or a bugfix?

[platinumazure]

It looks like the issue is that the code doesn't want to autofix if there are multiple declarators (under the assumption that if one declarator needs fixing but others don't, it's not clear how the fix should be applied).

However, we do have logic for autofixing if a destructuring assignment is completely convertible (all variables can be const).

So I could see this being treated as a bugfix, but I'll defer to others in the team.

[kaicataldo]

This feels more like an enhancement to me (the current behavior isn’t wrong—it could just be made smarter). But either way, seems like a good addition to make the change @platinumazure suggested.

[sstern6]

Would it be ok to take a look at this?

[sstern6]

Looking into this, as @platinumazure said the rule guards against autofixing multiple variable declarations found: https://github.com/eslint/eslint/blob/master/lib/rules/prefer-const.js#L324. How does it sound to update the logic of the rule to if there are multiple declarations that are ALL require statements to allow autofixing? To avoid changing the rule too much, and I think that in some situations having the dev decide how to split up the declarations is a good idea.

Does this sound ok?

[kaicataldo]

@sstern6 Can you clarify how requires factor into this? I'd be hesitant to add functionality that only affects assignments that are requires, since this rule should really only care about reassignment.

[sstern6]

Sure, maybe I misunderstood, but for example. Currently this code errors out but isnt autofixed because the rule checks the length of the block and says if there are multiple declarations then let the dev decide how to handle and eslint will not autofix.

let acorn = require("acorn"),
    assert = require("chai").assert;

Here is where I am misunderstanding, for this ticket, we want to update the rule to autofix in all cases?

My propsal was to autofix the following because all var declarations were require statements.

let acorn = require("acorn"),
    assert = require("chai").assert;

and to continue to allow devs to decide how to handle examples like the following.

let acorn = require("acorn"),
    assert = require("chai").assert,
    foo = 'foo';

or

let a ='a',
    b = 'b';

Maybe could you clarify what the intended behavior is?

Thanks @kaicataldo

[kaicataldo]

@sstern6 I think the ask here is to autofix multiple declarators (not to specifically fix assignments that are requires). The rule doesn't currently differentiate whether the assigned value is a require or not, and I don't think it should.

[kaicataldo]

To provide more context, I believe the ask is to add an autofixer for the following case:

let a = 1,
    b = 2;

to

const a = 1,
    b = 2;

instead of just warning like it does now.

Check out the demo link @aladdin-add provided above to see the warnings that are generated and how the autofix doesn't fix them (when we should be able to).

[sstern6]

Kk working on this. Will have a PR over the next few days. Thanks for the clarification.