[Feature Request] Multipart copy support

[aselder]

I'd love to see this library support a high level operation similar to S3.upload that would do a copy in parts transparently to the end user. The Ruby's SDKs AWS::S3::ObjectMultipartCopier.

We already have the upload_part_copy primitive, just need to do the coordination of all the parts and calculate start and end ranges and kick off a bunch of the upload_part_copy commands similar to the defimplin https://github.com/ex-aws/ex_aws_s3/blob/v2.0.1/lib/ex_aws/s3/upload.ex.

I'll try to start a PR, but I'll probably need polishing help. My plan.

1. If user provides the size, use it; otherwise do a head_object to get the size.
2. Create a list of {start, stop} tuples, based on given chunk size (or the standard 5MB)
3. Use Task.async_streams to iterate through the ranges, calling upload_part_copy for each one.

[aselder]

Quick question for @benwilson512 or who ever else might know...

When doing a multipart copy, the Ruby library initiates an upload and passes the upload_id and a part_numberto the put upload part command. Looking at https://docs.aws.amazon.com/AmazonS3/latest/API/mpUploadUploadPartCopy.html this seems to be required, but upload_part_copy doesn't seem to support them.

[kerryjj]

I found the same issue and ended up creating a patched version of ex_aws that supports it.

Here's my patch for it:
blendmedia@2836ec9

Just had a look at the latest upload_part_copy and it still doesn't seem to have the upload_id and part_number even though they are required by the S3 API.

@benwilson512 Am I missing something here? Or would you like me to submit this patch as a PR to add this support?

[kerryjj]

I've been using this patch in production since Feb 2019. Looking at it again now because I'd quite like to move back to using the official version of ex_aws and this is the one thing that's stopping me.