Uploading file from file object with PyCurl

Question

I'm attempting to upload a file like this:

import pycurl

c = pycurl.Curl()

values = [
     ("name", "tom"),
     ("image", (pycurl.FORM_FILE, "tom.png"))
]

c.setopt(c.URL, "http://upload.com/submit")
c.setopt(c.HTTPPOST, values)
c.perform()
c.close()
This works fine. However, this only works if the file is local. If I was to fetch the image such that:

import urllib2
resp = urllib2.urlopen("http://upload.com/people/tom.png")
How would I pass resp.fp as a file object instead of writing it to a file and passing the filename? Is this possible?

Answer 1

It might be possible in perfect situations to basically connect the two streams, but it wouldn't be a very robust solution. There are a bunch of ugly boundary conditions:

The response socket might still be receiving data, and/or be stalled, thus causing you to starve out and break the POST (because PycURL is not expecting to have to wait for data beyond the current end of the "file").
The response might reset, and then you don't have the complete file, but you've already POSTed a bunch of data - what to do in this case?
The file you're fetching with urllib might be chunked-encoded, so you need to perform some operations on the MIME headers for reassembly - you can't just blindly forward the data.
You don't necessarily know how big the file you're getting is, so it's hard to provide the proper content-length on the POST, so then you have to write chunked.
Probably a bunch of other problems I can't think of off the top of my head...
You'll be much better off writing the file to disk temporarily and then POSTing it once you know you have the whole thing.

If you did want to do this, the best way would probably be to implement your own file-like object which would manage the bridge between the two connections (could properly buffer, handle decoding, etc.).