Repository navigation
IO object used in requests data #744
Description
Describe the bug
I have an edge case where I had to use BytesIO as the data for requests, in production environment, the IO is consumed after the first request, and all subsequent requests read '' as a result. I wanted to reproduce this behavior/issue in tests (for regression tests). However, there seems to be no way to reproduce this behavior.
Additional context
No response
Version of responses
0.25.3
Steps to Reproduce
# reproduction_test.py
import io
import pytest_responses
import requests
def bad_io():
data = io.BytesIO(b"some binary data")
first = requests.post("https://httpbin.org/post", data=data)
last = requests.post("https://httpbin.org/post", data=data)
return first, last
def test_with_responses(responses):
responses.add(responses.POST, "https://httpbin.org/post")
bad_io()
assert responses.calls[0].request.body.getvalue() == b"some binary data"
assert responses.calls[1].request.body.getvalue() == b""
if __name__ == "__main__":
import pprint
first_repsonse, last_response = bad_io()
pprint.pprint(first_repsonse.json()["data"])
pprint.pprint(last_response.json()["data"])compare the difference in body/data between
python reproduction_test.py
vs
pytest reproduction_test.py
Expected Result
Test passes - the IO object is read/consumed after the first request
The second request body is ``
Actual Result
Test fails.
The second request got the same payload as the initial request.
Activity
markstory commentedon Dec 13, 2024
Currently responses will read and buffer any IO streams that are passed to it, which is why you're not able to reproduce the scenario you're looking to emulate. I don't have a good idea on how we could support both buffered IO objects and un-buffered ones.