The logical AND and OR operators are the only lazy operators in JavaScript along with the ternary conditional operator. They are tested for short-circuit evaluation using the following rules:
false && anything === false
true || anything === true
This is the same way it is implemented in Haskell:
(&&) :: Bool -> Bool -> Bool
False && _ = False
True && x = x
(||) :: Bool -> Bool -> Bool
True || _ = True
False || x = x
However according to MDN logical operators in JavaScript are left associative. This is counter intuitive. In my humble opinion they should be right associative. Haskell does the right thing. Logical operators in Haskell are right associative:
infixr 3 &&
infixr 2 ||
Consider the following expression in Haskell:
False && True && True && True
Because &&
is right associative in Haskell the above expression is equivalent to:
False && (True && (True && True))
Hence it doesn't matter what the expression (True && (True && True))
evaluates to. Because of the first False
the entire expression is reduced to False
in a single step.
Now consider what would happen if &&
was left associative. The expression would be equivalent to:
((False && True) && True) && True
It would now take 3 reductions to evaluate the entire expression:
((False && True) && True) && True
(False && True) && True
False && True
False
As you can see it makes more sense for logical operators to be right associative. This brings me to my actual question:
Why are logical operators in JavaScript left associative? What does the ECMAScript specification have to say about this? Are logical operators in JavaScript actually right associative? Does the MDN docs have incorrect information about the associativity of logical operators?
Edit: According to the specification logical operators are left associative:
LogicalANDExpression = BitwiseORExpression
| LogicalANDExpression && BitwiseORExpression
LogicalORExpression = LogicalANDExpression
| LogicalORExpression || LogicalANDExpression
For any decent compiler the chosen associativity of these operators are pretty much irrelevant, and the outputted code will be the same regardless. Yes, the parse tree is different, but the emitted code doesn't need to be.
In all of the languages of the C family that I know of (to which Javascript also belongs), the logical operators are left associative. So the real question becomes, why do C-like languages define logical operators as left-associative? Since the chosen associativity is irrelevant (in terms of semantic, and in terms of efficiency), my suspicion is that the most "natural" (as in "what the majority of the other operators use") associativity was chosen, although I don't have any sources to back up my claims. Other possible explanation is that left associative operators take less stack space to parse using an LALR parser (which is not a big concern nowadays, but probably was back when C appeared).