Function Composition of JavaScript and Haskell

In JavaScript we can define function composition as using a function which takes two functions f and g, to produce a new function:

function o(f, g) {
  return function(x) {
   return f(g(x));
  }
 }

This seems obvious, but my question is, does the runtime interpreter/compiler actually calculate f(g(x))?

Say we have some large data like an array with many elements, and is the previously composed function o as the f(g(x)) by the prior calculation faster than uncomposed f(g(x)) ?

Perhaps, in this case, o(f, g) as merely a macro expression of f(g(x))?

And if it's a macro, perhaps, the performance doesn't make much difference?

It may depends on the runtime environment, I especially interested in V8 engine for Chrome/node.js.

Haskell as a language of lazy evaluation strategy theoretically compose functions, am I correct? Does GHC actually calculate to compose functions?

No, the function call o(f, g) will return exactly the anonymous function:

function (x) {
    return f(g(x));
}

Then, only at the time, when you call that anonymous function, g(x) will execute and then f will execute on the result of g(x). Every time you call that anonymous function both g and f execute, one after the other. So the using the composed function will be just a little slower than manually calling f(g(x)) everywhere in your code due to the slight overhead of the extra anonymous function.

Example:

function o(f, g){
    return function(x) {
        return f(g(x));
    }
}

function addTo5(x){
    return x + 5;
}

function multiplyBy(x){
    return function(y){
        return x*y;
    }
}

var composed = o(multiplyBy, addTo5);
composed(5)(3); // Slightly slower than multiplyBy(addTo5(5))(3)

jsperf

No calculation is performed until you call the composed function returned by o. This new function just keeps track of what f and g are, and invoke them when you pass x in, evaluating and returning the final result.