2012-07-24

Apply and arrays: three tricks

This blog post describes three tricks for working with arrays via apply.

The apply method

apply is a method that all functions have. Its signature is
    func.apply(thisValue, [arg1, arg2, ...])
Ignoring thisValue, the above invocation is equivalent to:
    func(arg1, arg2, ...)
So, apply lets us unwrap an array “into” the arguments of a function call. Let’s look at three tricks that apply allows us to perform.

Trick 1: hand an array to a function that does not accept arrays

JavaScript does not have a function that returns the maximum of an array of numbers. However, there is the function Math.max that works with an arbitrary number of number-valued arguments. Thanks to apply, we can use that function for our purposes:
    > Math.max.apply(null, [10, -1, 5])
    10

Trick 2: eliminate holes in arrays

Holes in arrays

Quick reminder: In JavaScript, an array is a map from numbers to values. So there is a difference between a missing element (a hole) and an element having the value undefined [1]. The former is skipped by the Array.prototype iteration methods (forEach, map, etc.), the latter isn’t:
    > ["a",,"b"].forEach(function (x) { console.log(x) })
    a
    b
    
    > ["a",undefined,"b"].forEach(function (x) { console.log(x) })
    a
    undefined
    b
You can also check for the presence of holes via the in operator.
    > 1 in ["a",,"b"]
    false
    > 1 in ["a", undefined, "b"]
    true
But if you read a hole, the result is undefined for both holes and undefined elements.
    > ["a",,"b"][1]
    undefined
    > ["a", undefined, "b"][1]
    undefined

Eliminating holes

With apply and Array (which can be used as either a function or a constructor), you can turn holes into undefined elements:
    > Array.apply(null, ["a",,"b"])
    [ 'a', undefined, 'b' ]
The above works, because apply does not ignore holes, it passes them as undefined arguments to the function:
    > function returnArgs() { return [].slice.call(arguments) }
    > returnArgs.apply(null, ["a",,"b"])
    [ 'a', undefined, 'b' ]
Alas, you are faced with a quirk here. If Array receives a single numeric argument, an empty array is created whose length is the number [2]:
    > Array.apply(null, [ 3 ])
    [ , ,  ]
Therefore, if you need this functionality, it is better to write your own function:
    function fillHoles(arr) {
        var result = [];
        for(var i=0; i < arr.length; i++) {
            result[i] = arr[i];
        }
        return result;
    }
Example:
    > fillHoles(["a",,"b"])
    [ 'a', undefined, 'b' ]
Underscore gives you the _.compact function which removes all falsy values, including holes:
    > _.compact(["a",,"b"])
    [ 'a', 'b' ]
    > _.compact(["a", undefined, "b"])
    [ 'a', 'b' ]
    > _.compact(["a", false, "b"])
    [ 'a', 'b' ]

Trick 3: flatten an array

Task: turn an array of arrays with elements into just an array of elements. Again we use the unwrapping ability of apply to make concat do this work for us:
    > Array.prototype.concat.apply([], [["a"], ["b"]])
    [ 'a', 'b' ]
Non-array elements are added as is:
    > Array.prototype.concat.apply([], [["a"], "b"])
    [ 'a', 'b' ]
apply’s thisValue must be [] here, because concat is a method, not a non-method function. Only one level of flattening is performed:
    > Array.prototype.concat.apply([], [[["a"]], ["b"]])
    [ [ 'a' ], 'b' ]

If you want your code to be self-descriptive, you should consider alternatives, including implementing your own properly named function. Underscore has _.flatten which handles any level of nesting:

    > _.flatten([[["a"]], ["b"]])
    [ 'a', 'b' ]

References

  1. JavaScript: sparse arrays vs. dense arrays
  2. ECMAScript.next: Array.from() and Array.of()

No comments: