The strings are not optimized on nodejs #85
Description
Hi,
The strings in josn5 are formed using character concatenation, this results in large amount of memory needed to store each in nodejs. v8 uses 'first' and 'second' pointer store concatenated strings. Since we are using character concatenation it ends up storing large tree instead of single string. One of the work around is to call string.charAt(0) on the resulting string so that v8 then converts the large tree into single string.
Here is the work around we are using.
function parse_json5(text, def) {
var json;
try {
json = json5.parse(text);
v8optimze(json);
} catch (ex) {
json = isUndefined(def) ? {} : def;
}
return json;
}
function v8optimze(jsonObj) {
var tempStr, key;
if (!jsonObj) {
return null;
}
switch (typeof jsonObj) {
case 'string':
if (jsonObj.length > 0) {
tempStr += jsonObj.charAt(0);
}
break;
case 'object':
if (Array.isArray(jsonObj)) {
for (key = 0; key < jsonObj.length; key++) {
v8optimze(jsonObj[key]);
}
} else {
for (key in jsonObj) {
if (jsonObj.hasOwnProperty(key)) {
v8optimze(key);
v8optimze(jsonObj[key]);
}
}
}
break;
}
}
The screenshot shows how a small string (25 bytes) is using 1284 bytes instead.
You can use node-heapdump (https://github.com/bnoordhuis/node-heapdump) to get heapdump in nodejs and analyze it in chrome dev tools.
Here is the code to get heapdump:
var json5 = require('json5');
var heapdump = require('heapdump');
global.test = json5.parse('{ foo: "hello world"}');
heapdump.writeSnapshot('/tmp/' + Date.now() + '.heapsnapshot', function (err, fil) {
console.log(err);
console.log('saved');
console.log(global.test);
});