Open
Description
TextDecoder is a newer browser feature and supported in a few browsers. From my testing in the browser (reading 100,000 different messages with a few text fields) TextDecoder seems to be 700% faster than the current implementation. Trying it out locally, I edited this method: https://github.com/dcodeIO/protobuf.js/blob/master/src/reader.js#L320
/**
* Reads a string preceeded by its byte length as a varint.
* @returns {string} Value read
*/
Reader.prototype.string = function read_string() {
var bytes = this.bytes();
return utf8.read(bytes, 0, bytes.length);
};And replaced it with this
let decoder
if (global.window && global.window.TextDecoder) {
decoder = new TextDecoder('utf-8')
}
/**
* Reads a string preceeded by its byte length as a varint.
* @returns {string} Value read
*/
Reader.prototype.string = function read_string() {
var bytes = this.bytes();
if (decoder) {
return decoder.decode(bytes)
}
return utf8.read(bytes, 0, bytes.length);
};This worked well in the browser and in a node process, with higher performance than the regular implementation. This is an easy check and doesn't have much overhead.
You could also use TextEncoder when encoding text, however I haven't tried it out myself.