service/s3/s3manager: Downloader ignores byte range

[sandoracs]

Please fill out the sections below to help us address your issue.

Version of AWS SDK for Go?

v1.8.12-6-gbaba9e7

Version of Go (go version)?

go version go1.8 darwin/amd64

What issue did you see?

Byte range is ignored by 'downloader.Download()'

Steps to reproduce

`downloader := s3manager.NewDownloader(sess)
f, _ := os.Create("/tmp/data")
defer f.Close()

size, _ = downloader.Download(f, &s3.GetObjectInput{
Bucket: aws.String("bucket"),
Key: aws.String("key"),
Range: aws.String("bytes=0-8191"),
})`

It will download the full object instead of the first 8MB.

[jasdel]

Thanks for creating this issue @sandoracs. I agree this looks to be an oversight with S3 Download Manager. I think it makes sense for the downloader to use this value.

Since the value was being ignored before we'd need to figure out how the SDK should handle this field if it is not formatted properly.

[jasdel]

In the meantime @sandoracs the S3.GetObject API call will work correctly with the Range parameter. This could be used as a workaround until the SDK's download manager issue is fixed.

[sandoracs]

Thank you!

[jasdel]

In taking a look at this issue, it looks to support this feature the S3 Download Manager would need a pretty significant refactor. To address the issue of taking a slice of an Object I think the downloader can be updated to short circuit multipart download functionality and fallback to using GetObject directly.

We can then see if there is demand for full byte-range support within the download manager and prioritize that as a feature request.

[jasdel]

Hi @sandoracs I've just merged in #1311 . This basic support for Byte-Range with the S3 manager. Setting this field will disable S3's concurrency support. We can take a look at adding further support for byte range with multipart get object download as a feature request.

[sandoracs]

Hi! Thank you very much! 👍