mlhpdx/s3-upload-stream

Code library that provides a Stream implementation that makes working with uploads to S3 easier where the size of the content isn't known a priori. It holds only partial content in memory (works with large objects), and is compatible with code libraries that work with output streams.

s3-upload-stream's Issues

Write zero-length files.

Thank you for the example code. Wrapping the AWS SDK S3 Multipart Upload in a stream is quite useful.

However, in my use case, my inputs are compressed files that I must expand and write to S3. Sometimes the compressed file is empty (zero-length), other times it holds files over a gigabyte in size, usually the files are a few megabytes, but the content varies widely.

Can this be easily modified to write zero-length files to S3?

In my own debugging, it looks like the .Net 6 framework is calling the Write method on a separate thread, with only framework code in the call stack. And then, only if there is content to write - I guess a file header with no content doesn't pass the test.

Question about part length

Hi,

Thanks a lot for the example, this solved me a big issue.
I just have a question regarding the calculation of the part length in StartNewPart method:
_metadata.PartLength = Math.Min(MAX_PART_LENGTH, Math.Max(_metadata.PartLength, (_metadata.PartCount / 2 + 1) * MIN_PART_LENGTH));

s3-upload-stream/S3UploadStream/S3UploadStream.cs

Line 100 in c3f274f

_metadata.PartLength = Math.Min(MAX_PART_LENGTH, Math.Max(_metadata.PartLength, (_metadata.PartCount / 2 + 1) * MIN_PART_LENGTH));

What is the logic behind _metadata.PartCount / 2 + 1?