Question

In JavaScript you can convert a number to a string representation with a specific radix as follows:

(12345).toString(36) // "9ix"

…and you can convert it back to a regular number like this:

parseInt("9ix", 36) // 12345

36 is the highest radix you can specify. It apparently uses the characters 0-9 and a-z for the digits (36 total).

My question: what’s the fastest way to convert a number to a base 64 representation (for example, using A-Z, and - and _ for the extra 28 digits)?

---

Update: Four people have posted responses saying this question is duplicated, or that I’m looking for Base64. I’m not.

“Base64” is a way of encoding binary data in a simple ASCII character set, to make it safe for transfer over networks etc. (so that text-only systems won’t garble the binary).

That’s not what I’m asking about. I’m asking about converting numbers to a radix 64 string representation. (JavaScript’s toString(radix) does this automatically for any radix up to 36; I need a custom function to get radix 64.)

---

Update 2: Here are some input & output examples…

0   → "0"
1   → "1"
9   → "9"
10  → "a"
35  → "z"
61  → "Z"
62  → "-"
63  → "_"
64  → "10"
65  → "11"
128 → "20"
etc.

Answer 1

Here is a sketch for a solution for NUMBERS (not arrays of bytes 🙂

only for positive numbers, ignores fractional parts, and not really tested — just a sketch!

Base64 = {

    _Rixits :
//   0       8       16      24      32      40      48      56     63
//   v       v       v       v       v       v       v       v      v
    "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz+/",
    // You have the freedom, here, to choose the glyphs you want for 
    // representing your base-64 numbers. The ASCII encoding guys usually
    // choose a set of glyphs beginning with ABCD..., but, looking at
    // your update #2, I deduce that you want glyphs beginning with 
    // 0123..., which is a fine choice and aligns the first ten numbers
    // in base 64 with the first ten numbers in decimal.

    // This cannot handle negative numbers and only works on the 
    //     integer part, discarding the fractional part.
    // Doing better means deciding on whether you're just representing
    // the subset of javascript numbers of twos-complement 32-bit integers 
    // or going with base-64 representations for the bit pattern of the
    // underlying IEEE floating-point number, or representing the mantissae
    // and exponents separately, or some other possibility. For now, bail
    fromNumber : function(number) {
        if (isNaN(Number(number)) || number === null ||
            number === Number.POSITIVE_INFINITY)
            throw "The input is not valid";
        if (number < 0)
            throw "Can't represent negative numbers now";

        var rixit; // like 'digit', only in some non-decimal radix 
        var residual = Math.floor(number);
        var result = '';
        while (true) {
            rixit = residual % 64
            // console.log("rixit : " + rixit);
            // console.log("result before : " + result);
            result = this._Rixits.charAt(rixit) + result;
            // console.log("result after : " + result);
            // console.log("residual before : " + residual);
            residual = Math.floor(residual / 64);
            // console.log("residual after : " + residual);

            if (residual == 0)
                break;
            }
        return result;
    },

    toNumber : function(rixits) {
        var result = 0;
        // console.log("rixits : " + rixits);
        // console.log("rixits.split('') : " + rixits.split(''));
        rixits = rixits.split('');
        for (var e = 0; e < rixits.length; e++) {
            // console.log("_Rixits.indexOf(" + rixits[e] + ") : " + 
                // this._Rixits.indexOf(rixits[e]));
            // console.log("result before : " + result);
            result = (result * 64) + this._Rixits.indexOf(rixits[e]);
            // console.log("result after : " + result);
        }
        return result;
    }
}

UPDATE: Here’s some (very lightweight) testing of the above, for running in NodeJs where you have console.log.

function testBase64(x) {
    console.log("My number is " + x);
    var g = Base64.fromNumber(x);
    console.log("My base-64 representation is " + g);
    var h = Base64.toNumber(g);
    console.log("Returning from base-64, I get " + h);
    if (h !== Math.floor(x))
        throw "TEST FAILED";
}

testBase64(0);
try {
    testBase64(-1);
    }
catch (err) {
    console.log("caught >>>>>>  " + err);
    }
try {
    testBase64(undefined);
    }
catch (err) {
    console.log("caught >>>>>>  " + err);
    }
try {
    testBase64(null);
    }
catch (err) {
    console.log("caught >>>>>>  " + err);
    }
try {
    testBase64(Number.NaN);
    }
catch (err) {
    console.log("caught >>>>>>  " + err);
    }
try {
    testBase64(Number.POSITIVE_INFINITY);
    }
catch (err) {
    console.log("caught >>>>>>  " + err);
    }
try {
    testBase64(Number.NEGATIVE_INFINITY);
    }
catch (err) {
    console.log("caught >>>>>>  " + err);
    }

for(i=0; i<100; i++)
    testBase64(Math.random()*1e14);