No Retry in HTTPResponse.read() method.

It appears that the retry support in urllib3 doesn’t cover the HTTPResponse.read() method. This means that if you pass preload_content=False to urlopen() as requests does, that any errors that occur while read()ing the body will not automatically be retried. Doing this transparently would be difficult to do since urllib3 doesn’t know what the calling library has done with the parts of the body it’s already received.

What would be nice though is to provide the Retry() object on the response object and provide instructions for how the caller of the library can use it to implement retries themselves. Perhaps this could look something like:

response = pool.urlopen(...)
while not response.closed and not response.retries.is_exhausted():
    try:
        data = []
        data.append(response.read(4096))
    except response.retries.exceptions as exc:
        # This would raise an error if there are no more retries
        response = response.retry()

Thoughts?

[Lukasa]

Nope

[gaborbernat]

any progress on this?