Odd Issue Using BytesIO With Boto3

I’ve been trying to use more Python lately and I ran across an odd issue that I had to solve. I don’t know if it’s something that I’m doing incorrectly, a bug, or just a quirk.

Hey this is Jamie from the future. I wasn’t doing it correctly. Anytime you want to read an IO stream object in Python it’s always best practice to call seek and set it to the beginning or just use the getvalue function.

The Problem

I needed to convert a list of objects to a CSV, and then upload that CSV to an S3 bucket. This was a small CSV file so I wanted to do it all in-memory to save having to perform any disk IO. The Boto3 library has a upload_fileobj function that can take a file-like object that is in binary mode. The Python csv module can support any object that supports the methods required for what you need to do. Since I wanted to write, I could use many of the classes in the IO package. I decided I would use StringIO with the csv.writer to get an in-memory stream of my CSV file. I then can convert that text stream into a binary stream to pass to upload_fileobj function.

The Code

This is an example of the initial solution that I developed.

stream = StringIO()
writer = csv.writer(stream)
writer.writerow(['Test1', 'Test2'])
writer.writerow(['TestA', 'TestB'])

client = boto3.client('s3')
client.upload_fileobj(BytesIO(stream.read().encode()), 'upload_bucket', 'csv_key'))

One key thing to point out is that getting the value of the StringIO stream and then encoding it to bytes is the easiest way to convert it to a ByteIO object. After that, the in-memory CSV is pushed to my S3 bucket. When I went to verify that it all worked, I noticed that the file was empty. It took me way too long to figure it and that’s why I decided to write a post. I was only getting the value of the stream from the current index which is where the writer had stopped writing. After a little research, I realized that I would need to set seek on the stream to set it back to the start of the stream.

Here is the improved code:

stream = StringIO()
writer = csv.writer(stream)
writer.writerow(['Test1', 'Test2'])
writer.writerow(['TestA', 'TestB'])
stream.seek(0)

client = boto3.client('s3')
client.upload_fileobj(BytesIO(stream.read().encode()), 'upload_bucket', 'csv_key'))

After that tweak, I ran my test again and when I verified my uploaded file, it had all the content.

Realizations While Writing This Post

After starting this blog post, I realized that if using the read function on any IO object, you are required to set the stream position back to the start of the stream when using a writer of any sort as it’s ready to receive the next write. Then I realized that I probably shouldn’t be using read at all, I can just call getvalue and it will get all the contents of the stream. Here is the final example that is a little cleaner.

stream = StringIO()
writer = csv.writer(stream)
writer.writerow(['Test1', 'Test2'])
writer.writerow(['TestA', 'TestB'])

client = boto3.client('s3')
client.upload_fileobj(BytesIO(stream.getvalue().encode()), 'upload_bucket', 'csv_key'))

Wrapping Up

It’s always a journey when you start using a programming language more for tasks you aren’t as familiar with. I learned a lot of intricacies that I take for granted in the languages that I know better. I hope someone else finds this helpful or amusing.

Thanks for reading,

Jamie

If you enjoy the content, then consider buying me a coffee.