Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> You can also use the unary + operator to convert values to numbers:

    + "42"; // 42
> [...] However the "+" operator simply converts the string to NaN if there is any invalid character in it.

Being a bit pedantic here, why not recommending the Number function which may be less obscure for beginners?

    Number("42"); // 42


If you use a double NOT ~~"foo" you will always get a finite int32 value... which is useful when you can use 0 as a "default" representation. For that matter, all bitwise operations will coerce a value to an int32 representation. Which is something to be aware of, if you want larger integer values (< 2^53 - 1 and > -2^53 + 1).


Number() also converts to NaN if it's not a number though I would still avoid the + operator here because it is code smell (looks like you forgot your variable to concat).

Number automatically accepts exponent notation and auto-converts to hex (ignoring octal). parseInt() auto-converts hex and octal unless a radix is specified, but it also ignores the remaining non-number characters at the end of the string.

There isn't a single method that does whatever you want.


the String function is also a nice way to cast a value to a string (using .toString() is unsafe because it may not exist, such as with undefined and null): String(42) === "42".

An other thing that could maybe be covered is this specifity:

    typeof "42" === "string"
    typeof String(42) === "string"
    typeof new String(42) === "object"
Native constructors are ... fun beasts, to say the least.


Another way to cast to a Number is the bitwise-or operator. It has the useful property of always yielding a number.

    '42' | 0;      // 42
    NaN | 0;       // 0
    null | 0;      // 0
    undefined | 0; // 0
    false | 0;     // 0
    true | 0;      // 1


Bitwise operations don't produce Numbers (f64), but signed integers (i32).


> produce...signed integers

Well, except for right logical shifts.


It always yields an integer, so it's useful if you know you want an integer.


Correct me if I'm wrong, but I thought that the unary operator and the Number constructor yield equivalent results?


Since calling constructor functions without `new` is generally an error, I'd rather use `parseFloat`.


> Since calling constructor functions without `new` is generally an error

The behavior of Number, Boolean, String, and Array is well-defined, it's safe to call them without new. In fact, in the case of String/Boolean/Number, calling them with new will often do something you don't expect. (Calling them with new gives you a Number/String/Boolean object, not primitive, which can cause trouble when you try to compare them with ===, unless you remember to use their `valueOf` method)

Also, as noted below, it's possible for objects to not have a `toString` method, so attempting to call it to get the object's value as a string could blow up. So it's actually safer to coerce to string by adding an empty string or passing to String().


> The behavior of Number, Boolean, String, and Array is well-defined, it's safe to call them without new.

ES6 is packed with weird but well-defined things.

The problem is that calling a constructor function (a PascalCase'd function) without `new` looks like an error because it generally is an error. To make matters worse, without closely examining that function, you cannot tell if it's an error.

I do know that `Number()` happens to be one of those constructor functions which not only work without `new`, it also happens to behave differently when `new` is missing. It does not return an object. It returns a primitive.

Someone who doesn't know about this unusual secondary function will waste some time if they spot this apparent mistake.

Now, to defuse this time-wasting trap, you could either add a comment... or just do the sensible thing and write it in a way which does not require a comment.


That's fair, I just find them super useful in various situations, for example filtering over an array with Boolean to remove falsey values.


Yes, that's a good suggestion. Boolean works as well, for coercing to boolean of course.


Because Number(...) uses more bytes on the wire.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: