mirror of
https://github.com/ascribe/onion.git
synced 2024-11-15 09:35:10 +01:00
61 lines
2.0 KiB
JavaScript
61 lines
2.0 KiB
JavaScript
'use strict';
|
|
|
|
import SparkMD5 from 'spark-md5';
|
|
|
|
/**
|
|
* Takes a string, creates a text file and returns the URL
|
|
*
|
|
* @param {string} text regular javascript string
|
|
* @return {string} regular javascript string
|
|
*/
|
|
export function makeTextFile(text) {
|
|
let data = new Blob([text], {type: 'text/plain'});
|
|
return window.URL.createObjectURL(data);
|
|
}
|
|
|
|
/**
|
|
* Takes a file Object, computes the MD5 hash and returns the URL of the textfile with the hash
|
|
*
|
|
* @param {File} file javascript File object
|
|
* @return {string} regular javascript string
|
|
*/
|
|
export function computeHashOfFile(file) {
|
|
let blobSlice = File.prototype.slice || File.prototype.mozSlice || File.prototype.webkitSlice,
|
|
chunkSize = 2097152, // Read in chunks of 2MB
|
|
chunks = Math.ceil(file.size / chunkSize),
|
|
currentChunk = 0,
|
|
spark = new SparkMD5.ArrayBuffer(),
|
|
fileReader = new FileReader();
|
|
|
|
let startTime = new Date();
|
|
fileReader.onload = function (e) {
|
|
//console.log('read chunk nr', currentChunk + 1, 'of', chunks);
|
|
spark.append(e.target.result); // Append array buffer
|
|
currentChunk++;
|
|
|
|
if (currentChunk < chunks) {
|
|
loadNext();
|
|
} else {
|
|
let fileHash = spark.end();
|
|
console.info('computed hash %s (took %d s)',
|
|
fileHash,
|
|
Math.round(((new Date() - startTime) / 1000) % 60)); // Compute hash
|
|
let hashFile = this.makeTextFile(fileHash);
|
|
console.info('hash: ', hashFile);
|
|
return hashFile;
|
|
}
|
|
}.bind(this);
|
|
|
|
fileReader.onerror = function () {
|
|
console.warn('oops, something went wrong.');
|
|
};
|
|
function loadNext() {
|
|
var start = currentChunk * chunkSize,
|
|
end = ((start + chunkSize) >= file.size) ? file.size : start + chunkSize;
|
|
|
|
fileReader.readAsArrayBuffer(blobSlice.call(file, start, end));
|
|
}
|
|
|
|
loadNext();
|
|
|
|
} |