Uploading files from the browser directly to S3 is needed in many applications. Fine-grained authorization is handled by the server, and the browser only handles file upload. This library does that.
Unfortunately S3 does not allow uploading files larger than 5GB in one chunk, and all the examples in AWS docs either support one chunk, or support multipart uploads only on the server.
As we don't want to proxy the upload traffic to a server (which negates the whole purpose of using S3), we need an S3 multipart upload solution from the browser.
The code by s3-multipart-upload-browser does some of that, but many of its features don't work. Most notably, it does not support AWS API V3.
This uploader supports upload prefixes (uploading only in a certain directory), fine-grained feedback, parallelized chunks, cancellation of the upload (and removal of the parts on S3), automated parts cleanup, drag & drop, progress bar and chunk size management.
- The browser Javascript code relies on jQuery. It also requires a browser with Blob, File and XHR2 support (most do since 2012).
- The server-side code is in PHP, but is straightforward enough to port to any language and is less than 200 LOC.
- If using with PHP, it needs AWS credentials and the AWS PHP SDK V3+ as well.
That's all!
Just put all files in a web directory, set the AWS credentials in index.php or in keys.php file, set your bucket's CORS config (can be done via a function in the code too), and you're ready.
I strongly recommend setting your S3 bucket for auto-removal of unfinished multipart upload parts. Otherwise, any incomplete upload will leave useless files on your bucket, for which you will be charged.
I also suggest using the largest supported chunk size (5GB) to make the XHR connections minimal. Uploading a 200GB file with 5GB chunks needs 40 parts, but using 100MB chunks requires 2000. Authorizing parts with AWS is both slow and pricy.
This work is released under MIT license.