Upload image available at public URL to S3 using boto

I'm working in a Python web environment and I can simply upload a file from the filesystem to S3 using boto's key.set_contents_from_filename(path/to/file). However, I'd like to upload an image that is already on the web (say https://pbs.twimg.com/media/A9h_htACIAAaCf6.jpg:large).

Should I somehow download the image to the filesystem, and then upload it to S3 using boto as usual, then delete the image?

What would be ideal is if there is a way to get boto's key.set_contents_from_file or some other command that would accept a URL and nicely stream the image to S3 without having to explicitly download a file copy to my server.

def upload(url):
    try:
        conn = boto.connect_s3(settings.AWS_ACCESS_KEY_ID, settings.AWS_SECRET_ACCESS_KEY)
        bucket_name = settings.AWS_STORAGE_BUCKET_NAME
        bucket = conn.get_bucket(bucket_name)
        k = Key(bucket)
        k.key = "test"
        k.set_contents_from_file(url)
        k.make_public()
                return "Success?"
    except Exception, e:
            return e

Using set_contents_from_file, as above, I get a "string object has no attribute 'tell'" error. Using set_contents_from_filename with the url, I get a No such file or directory error . The boto storage documentation leaves off at uploading local files and does not mention uploading files stored remotely.

Here is how I did it with requests, the key being to set stream=True when initially making the request, and uploading to s3 using the upload.fileobj() method:

import requests
import boto3

url = "https://upload.wikimedia.org/wikipedia/en/a/a9/Example.jpg"
r = requests.get(url, stream=True)

session = boto3.Session()
s3 = session.resource('s3')

bucket_name = 'your-bucket-name'
key = 'your-key-name' # key is the name of file on your bucket

bucket = s3.Bucket(bucket_name)
bucket.upload_fileobj(r.raw, key)

Ok, from @garnaat, it doesn't sound like S3 currently allows uploads by url. I managed to upload remote images to S3 by reading them into memory only. This works.

def upload(url):
    try:
        conn = boto.connect_s3(settings.AWS_ACCESS_KEY_ID, settings.AWS_SECRET_ACCESS_KEY)
        bucket_name = settings.AWS_STORAGE_BUCKET_NAME
        bucket = conn.get_bucket(bucket_name)
        k = Key(bucket)
        k.key = url.split('/')[::-1][0]    # In my situation, ids at the end are unique
        file_object = urllib2.urlopen(url)           # 'Like' a file object
        fp = StringIO.StringIO(file_object.read())   # Wrap object    
        k.set_contents_from_file(fp)
        return "Success"
    except Exception, e:
        return e

Also thanks to How can I create a GzipFile instance from the “file-like object” that urllib.urlopen() returns?

For a 2017-relevant answer to this question which uses the official 'boto3' package (instead of the old 'boto' package from the original answer):

Python 3.5

If you're on a clean Python install, pip install both packages first:

pip install boto3

pip install requests

import boto3
import requests

# Uses the creds in ~/.aws/credentials
s3 = boto3.resource('s3')
bucket_name_to_upload_image_to = 'photos'
s3_image_filename = 'test_s3_image.png'
internet_image_url = 'https://docs.python.org/3.7/_static/py.png'


# Do this as a quick and easy check to make sure your S3 access is OK
for bucket in s3.buckets.all():
    if bucket.name == bucket_name_to_upload_image_to:
        print('Good to go. Found the bucket to upload the image into.')
        good_to_go = True

if not good_to_go:
    print('Not seeing your s3 bucket, might want to double check permissions in IAM')

# Given an Internet-accessible URL, download the image and upload it to S3,
# without needing to persist the image to disk locally
req_for_image = requests.get(internet_image_url, stream=True)
file_object_from_req = req_for_image.raw
req_data = file_object_from_req.read()

# Do the actual upload to s3
s3.Bucket(bucket_name_to_upload_image_to).put_object(Key=s3_image_filename, Body=req_data)

Unfortunately, there really isn't any way to do this. At least not at the moment. We could add a method to boto, say set_contents_from_url, but that method would still have to download the file to the local machine and then upload it. It might still be a convenient method but it wouldn't save you anything.

In order to do what you really want to do, we would need to have some capability on the S3 service itself that would allow us to pass it the URL and have it store the URL to a bucket for us. That sounds like a pretty useful feature. You might want to post that to the S3 forums.