New rule proposal: no-dupe-else-if

[mdjermanovic]

Please describe what the rule should do:

Disallows duplicate test conditions in if-else-if chains.

What category of rule is this? (place an "X" next to just one item)

[X] Warns about a potential error (problem)

Provide 2-3 code examples that this rule will warn about:

if (a > 1) {
    foo();
} else if (a > 1) {
    bar();
}

if (a === 1) {
    foo();
} else if (a === 2) {
    bar();
} else if (a === 3) { 
    baz();
} else if (a === 2) {
    quux();
} else if (a === 5) {
    quuux();
}

Why should this rule be included in ESLint (instead of a plugin)?

It's a possible error. A statement with a duplicate condition can never execute (unless it's an unusual use case with side effects).

This rule is similar to the existing no-duplicate-case. The error could be due to a copy & paste and a long chain.

The test nodes would be compared by their tokens.

Are you willing to submit a pull request to implement this rule?

Yes.

[mdjermanovic]

One more 👍 to accept this.

[mysticatea]

Now accepted 🎉

[mdjermanovic]

I'm working on this :-)

[aladdin-add]

just a thought: for consecutive if, whether or not it should check, e.g.

if(foo === 0) {...}
if(foo === 0) {...} // to warn, or not?

I think it's good to cover it, or behide an option as an enhancement.

[mdjermanovic]

just a thought: for consecutive if, whether or not it should check, e.g.

if(foo === 0) {...}
if(foo === 0) {...} // to warn, or not?

I think it's good to cover it, or behide an option as an enhancement.

I think it's more suitable for a separate rule because this isn't necessarily an error (it could be just a bad practice to repeat conditions instead of having all code in a single block, or it might be even a correct code if foo can be assigned a different value in the first block), unlike certainly unreachable code that's being reported by the rule (assuming there are no side effects in conditions).

[kaicataldo]

just a thought: for consecutive if, whether or not it should check, e.g.

if(foo === 0) {...}
if(foo === 0) {...} // to warn, or not?

I think it's good to cover it, or behide an option as an enhancement.

Agreed with @mdjermanovic. I don't think this rule should catch this, since this rule shouldn't be disallowing code like the following:

function checkForB(a) { return a.b; }
function checkForC(a) { return a.c; }

const d = {
  c: true
};

let e = checkForB(d);
if (e !== undefined) {/* Do something */}
e = checkForC(d);
if (e !== undefined) {/* Do something */}

or even if we limit it to consecutive statements, potentially:

function checkForB(a) { return a.b; }
function checkForC(a) { return a.c; }

const d = {
  b: true
};

let e = checkForB(d);
if (e !== undefined) {
  if (typeof e !== 'number') {
    e = checkForC(d);
  } 
}
if (e !== undefined) {/* Do something */}

[aladdin-add]

emm.. after thinking for a while, I think it's not fit for this rule:

• its rulename: no-dupe-else-if :)
• it can catch some mistakes, but also has a few false positives.