#Negated in and instanceof operators S1

#Negated in and instanceof operators

#Status

Stage: 1
Champion: Pablo Gorostiaga Belio (@gorosgobe)

#Author

Pablo Gorostiaga Belio (@gorosgobe)

#Proposal

Presentations

• September 2023

#Motivation

JavaScript's in and instanceof operators have broadly the following behaviour:

a in obj; // returns true if property a is in obj or in its prototype chain, false otherwise
a instanceof C; // returns true if C.prototype is in a's prototype chain, false otherwise

To negate the result of these expressions, we can wrap them with the logical NOT (!) operator:

!(a in obj);
!(a instanceof C);

Negating an in/instanceof expression in this way suffers from a few problems:

#Error-proneness¹

The logical not operator, !, has to be applied to the whole expression to produce the intended result. Incorrect parenthesising of the sub-expression (which can be a part of an arbitrarily long expression) and/or applying the ! operator on the wrong operand can lead to errors that are hard to debug, i.e.:

For in:

if (!a in obj) { 
  // will not execute, unless obj has a 'true' or 'false' key
  // `in` accepts strings or symbols as the LHS parameter, and otherwise coerces all other values to a string
  // see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String#string_coercion
}

// correct usage
if (!(a in obj)) {
  // ...
}

For instanceof:

if (!a instanceof C) { 
  // will not execute, unless C provides a @@hasInstance method that returns true for booleans
}

// correct usage
if (!(a instanceof C)) {
  // ...
}

This type of error is fairly common. For in, this Sourcegraph query reveals that there are many instances of this issue (over ~2.1k instances when I ran it) across repos with thousands of stars on GitHub. While there are some false positives (from comments, for example), I highlight some notable examples below:

Similarly, for instanceof, this Sourcegraph query shows that there are also many instances of this bug (~19k occurrences when I ran it). As before, repos with thousands of stars are affected. Some examples follow below:

Within Bloomberg, we encourage the use of eslint and TypeScript, each of which have an error for these cases. However, because we allow teams to make some of their own decisions about tooling, bugs creeped through: in one large set of internal projects, we found that roughly an eighth of in/instanceof usages were negated in and instanceof expressions. More than 1% of negated in uses had this bug. This also affected negated instanceof, where more than 6% of uses had the bug. Our internal results are aligned with the data from the external sourcegraph queries: there is clearly a higher incidence of the bug on negated instanceof expressions compared to negated in expressions. While we are now fixing this internally, overall these results illustrate that this is a common problem due to the lack of ergonomics around negated in and instanceof expressions.

#Generates confusion

The negation of these expressions is not aligned with operators which have a negated version, such as ===/!==. This generates confusion among developers and leads to highly upvoted and viewed questions such as Is there a “not in” operator in JavaScript for checking object properties? and Javascript !instanceof If Statement.

#Readability

To negate the result of an in/instanceof expression, we introduce an additional grouping operator (denoted by two parentheses). In addition, the not is at the beginning of the expression, unlike how this would be read in natural English. Together, both of these factors result in less readable code.

#Worse developer experience

It is common to use in/instanceof as a guard in conditionals. Inverting these conditionals to reduce indentation in code, as this is correlated with code complexity, can lead to improved code readability and quality. With the existing operators, inverting the expression in the conditional requires the expression to be both wrapped with parentheses and negated.

#Solution

!in, a negated version of in, where

a !in obj;

is equivalent to

!(a in obj);

!instanceof, a negated version of instanceof, where

a !instanceof obj;

is equivalent to

!(a instanceof obj);

• Safer: No longer need to introduce additional grouping, and the negation is applied directly to the operator, as opposed to applying it next to the LHS operand in the expression.
• Improved readability: No longer requires extra grouping to negate the result of the expression. This is aligned with other operators such as !==. Reads more naturally and is more intuitive.
• Better developer experience: Again, easier to change when refactoring code - a single ! needs to be added to negate the expression.

#In other languages:

Python:

if item not in items:
  pass

if ref1 is not ref2:
  pass

Kotlin:

if (a !in arr) {}

if (a !is SomeClass) {}

C#:

if (a is not null) {}

Elixir:

a not in [1, 2, 3]

#Related Proposals

#Pattern matching

The pattern matching proposal proposes a new relational expression like a in b or a instanceof b, using a new operator is: https://github.com/tc39/proposal-pattern-matching#is-expression

In the same line as in and instanceof, we could extend the proposal to include a negated is operator such as !is.