New: Add no-dupe-else-if rule (fixes #12469)

[mdjermanovic]

What is the purpose of this pull request? (put an "X" next to item)

[X] New rule #12469

Examples of incorrect code for this rule:

/*eslint no-dupe-else-if: "error"*/

if (isSomething(x)) {
    foo();
} else if (isSomething(x)) {
    bar();
}

if (a) {
    foo();
} else if (b) {
    bar();
} else if (c && d) {
    baz();
} else if (c && d) {
    quux();
} else {
    quuux();
}

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

What changes did you make? (Give an overview)

New rule no-dupe-else-if

Is there anything you'd like reviewers to focus on?

[mysticatea]

Looks very good to me, but I have a question around || expression.

[mysticatea]

I'm guessing that to check each condition of the || expression individually is useful if the test nodes are LogicalExpression with || operator. Thoughts?

For example:

if (a || b) {
    // ...
} else if (a) { // duplicated condition
    // ...
}

I'm imaging code like:

                    const nodeConds = getOrConditions(node.test)
                    const currentConds = getOrConditions(current.test)
                    if (nodeConds.every(a => currentConds.some(b => astUtils.equalTokens(a, b, sourceCode)))) {

[mdjermanovic]

I think it's a very good idea.

We could even compare with the union of || conditions from all parents in the chain?

For example:

if (a) {
    // ...
} else if (b) {
    // ...
} else if (a || b) { // duplicated condition: { a, b } is a subset of { a, b }
    // ...
}

That would also catch things like this:

if (a || b) {
    // ...
} else if (c || d) {
    // ...
} else if (a || d) { // duplicated condition: { a, d } is a subset of { a, b, c, d }
    // ...
}

[mysticatea]

Nice!

[platinumazure]

Is this really necessary? This seems like it would add some complexity and go beyond the original proposal.

[mdjermanovic]

Looks like there are 3 options at the moment:

1. To check for equal ancestors only, as originally proposed in New rule proposal: no-dupe-else-if #12469
2. To check for a subset of an ancestor as in this comment
3. To check for a subset of the union of all ancestors as in this comment

I'm fine with all three 👍

Options 2. and 3. look very nice, but only 1. is accepted, so I don't how to proceed 😕

It doesn't look too complicated to do 2. or 3. Though, compared to the simple 1. it would be certainly much more code. Not sure how to weight how useful these enhancements would be.

[platinumazure]

I don't have any complexity concerns anymore based on the implementation shared earlier.

I don't understand why this only applies to || operators only, though. Shouldn't we look at && as well?

[kaicataldo]

I like this addition! Agreed with @platinumazure. It does seem like it could be useful to also check for and warn on:

if (a && b) {} 
else if (b && a) {}

[mdjermanovic]

👍 to account for && as well.

The logic for && in a chain is a bit different, so option 3 wouldn't apply well. For instance, the following is valid code:

if (a && b) {
    // ...
} else if (a) {
    // ...
} else if (b) {
    // ... 
}

if (a && b) {
    // ...
} else if (b && c) {
    // ...
} else if (a && c) {
    // ...
}

On the other hand, 'inverted' option 2 should work quite well for && and is easy to implement. If the condition split by && is a superset of one of the previous conditions, then it's a 'duplicate'. For example, all these are invalid:

if (a) {
    // ...
} else if (a && b) {
    // ...
}

if (a && b) {
    // ...
} else if (b && a) {
    // ...
}

if (a && b) {
    // ...
} else if (a && b && c) {
    // ...
}

In addition, it also looks easy to split the original condition by && and do the || check with each element of the resulting array, which would catch cases that might not even look like a bug at first glance:

if (a || b) {
    // ...
} else if (b && c) { // perhaps a not-so-obvious bug
    // ...  
}

if (a) {
    // ...
} else if (b) {
    // ...
} else if ((a || b) && c) {
   // ...
}

That shouldn't even add new lines, just .map and .some to the existing lines in this draft

[mdjermanovic]

This is done now, the rule should be able to catch some relatively complex errors ('more complex errors with || and &&' section in invalid tests).

Examples from the documentation:

/*eslint no-dupe-else-if: "error"*/

if (a || b) {
    foo();
} else if (a) {
    bar();
}

if (a) {
    foo();
} else if (b) {
    bar();
} else if (a || b) {
    baz();
}

if (a) {
    foo();
} else if (a && b) {
    bar();
}

if (a && b) {
    foo();
} else if (a && b && c) {
    bar();
}

if (a || b) {
    foo();
} else if (b && c) {
    bar();
}

if (a) {
    foo();
} else if (b && c) {
    bar();
} else if (d && (c && e && b || a)) {
    baz();
}

[mdjermanovic]

Tried the rule on ESLint codebase, reports only 1 bug at indent-legacy.js#L779 and it's indeed a bug because of !parentVarNode which is used in the previous condition with ||.

[aladdin-add]

LGTM, but a small suggestion, thanks!

[aladdin-add]

unrelated: is there a rule to warn something like: a || a; -- it seems a common mistake.

[mdjermanovic]

I don't think so. There was an idea to check this in #12097 (comment)

[mdjermanovic]

Off-topic, maybe we should reopen #12097 or open a new issue.

[mysticatea]

LGTM, thank you so much!

[kaicataldo]

One small comment, but otherwise this LGTM!

[kaicataldo]

LGTM, thank you!