Javascript string size limit: 256 MB for me - is i

Question

Curious about what was the maximum string length I could get in Javascript, I tested it myself, today, on my Firefox 43.0.1, running in Windows 7. I was able to construct a string with length 2^28 - 1, but when I tried to create a string with one more char, Firebug showed me the "allocation size overflow" error, meaning the string must be less than 256 MB.

Is this the same thing for all browsers, all computers, all operational systems, or it depends?

I created the following snippet to find out the limit:

(function() {
    strings = ["z"];
    try {
        while(true) {
            strings.push(strings[strings.length - 1] + strings[strings.length - 1]);
        }
    } catch(err) {
        var k = strings.length - 2;
        while(k >= 0) {
            try {
                strings.push(strings[strings.length - 1] + strings[k]);
                k--;
            } catch(err) {}
        }
        console.log("The maximum string length is " + strings[strings.length - 1].length);
    }
})();

If you are running a different browser/OS, I would like to see your results. My result was The maximum string length is 268435455.

P.S.: I searched around for an answer, but the most recent topic I found was from 2011, so I am looking for a more up-to-date information.

Answer 1

A bug report for the chromium tracker has this comment:

... When allocation fails, we create a 
Failure pointer encoding the amount requested, as well as some tag and 
type bits. This puts a limit on the maximally possible allocation 
request in 32-bit versions of 2^27-1. The maximal flat string length is 
~2^28 (512MB space), and the maximal string length is 2^29-1...

Note that this is from 2009, so I imagine that this still has consequences in current versions of V8 as the previous link is in regard to a NodeJS tool running into limits of toString().

Answer 2

Characters are stored on 16 bits

When you see that 256*2**20 characters are in a string, that does not mean that 256 megabytes of memory is allocated. JavaScript stores every character on two bytes (as it is utf16 encoded by the specification).

A word about ropes

Todays browsers (even IE) store strings in an advanced way, most often using a rope datastructure.

• Ropes do not need a coherent memory region to be allocated
• Can even deduplicate substrings, that means s+s does not necessarily use twice the memory as s
• Concatenation is very fast
• Element access is a bit slower

By examining some runs in IE and Chrome, I would say that both of them use some lazy evaluation for strings, and will try to expand them occasionally. After running the following snippet, none of the browsers used more memory than before. But if I tried to manipulate the stored window.LONGEST_STRING in the console, IE throw an out of memory error, and chrome froze for a short time, and consumed a lot of memory (>2 GB).

ps: On my laptop IE11 had a maximum string size of 4 GB, Chrome had 512 MB

Browser behaviour

IE11

Chrome47

A faster algorithm to determine max string size

var real_console_log = console.log;
console.log = function(x) {
  real_console_log.apply(console, arguments);
  var d = document,b=d.body,p=d.createElement('pre');
  p.style.margin = "0";
  p.appendChild(d.createTextNode(''+x));
  b.appendChild(p);
  window.scrollTo(0, b.scrollHeight);
};


function alloc(x) {
    if (x < 1) return '';
    var halfi = Math.floor(x/2);
    var half = alloc(halfi);
    return 2*halfi < x ? half + half + 'a' : half + half;
}

function test(x) {
    try {
        return alloc(x);
    } catch (e) {
        return null;
    }
}

function binsearch(predicateGreaterThan, min, max) {
    while (max > min) {
        var mid = Math.floor((max + min) / 2);
        var val = predicateGreaterThan(mid);
        if (val) {
            min = mid + 1;
        } else {
            max = mid;
        }
    }
    return max;
}

var maxStrLen = binsearch(test, 10, Math.pow(2, 52)) - 1;
console.log('Max string length is:');
console.log(maxStrLen + ' characters');
console.log(2*maxStrLen + ' bytes');
console.log(2*maxStrLen/1024/1024 + ' megabytes');
console.log('');
console.log('Store longest string');
window.LONGEST_STRING = alloc(maxStrLen);

console.log('Try to read first char');
console.log(window.LONGEST_STRING.charAt(0));
console.log('Try to read last char');
console.log(window.LONGEST_STRING.charAt(maxStrLen - 1));
console.log('Try to read length');
console.log(window.LONGEST_STRING.length);

Answer 3

Internal implementations can use either UCS2 or UTF16. As @hege_hegedus suggested, atleast Firefox uses the Rope structure (https://dxr.mozilla.org/mozilla-central/search?q=%2Btype-ref%3ARopeBuilder). The codes give me below results:

CHROME VERSION 39.0.2171.95 OS VERSION Linux: 3.13.0-43-generic

Firefox 34.0

Chrome output (from @@hege_hegedus code): Max string length is: 268435440 characters 536870880 bytes 511.9999694824219 megabytes Store longest string Try to read first char a Try to read last char a Try to read length 268435440

Firefox output (from OP code): "The maximum string length is 268435455"

Archived at http://gpupowered.org/string_js.txt