Ensure body is consumed only once

Fixes: kevin1024#846
Signed-off-by: Mathieu Parent <math.parent@gmail.com>

tests/unit/test_stubs.py

@@ -1,9 +1,12 @@
 import contextlib
+import http.client as httplib
+from io import BytesIO
+from tempfile import NamedTemporaryFile
 from unittest import mock
 
 from pytest import mark
 
-from vcr import mode
+from vcr import mode, use_cassette
 from vcr.cassette import Cassette
 from vcr.stubs import VCRHTTPSConnection
 
@@ -21,3 +24,52 @@ def testing_connect(*args):
             vcr_connection.cassette = Cassette("test", record_mode=mode.ALL)
             vcr_connection.real_connection.connect()
             assert vcr_connection.real_connection.sock is not None
+
+    def test_body_consumed_once_stream(self, tmpdir, httpbin):
+        self._test_body_consumed_once(
+            tmpdir,
+            httpbin,
+            BytesIO(b"1234567890"),
+            BytesIO(b"9876543210"),
+            BytesIO(b"9876543210"),
+        )
+
+    def test_body_consumed_once_iterator(self, tmpdir, httpbin):
+        self._test_body_consumed_once(
+            tmpdir,
+            httpbin,
+            iter([b"1234567890"]),
+            iter([b"9876543210"]),
+            iter([b"9876543210"]),
+        )
+
+    # data2 and data3 should serve the same data, potentially as iterators
+    def _test_body_consumed_once(
+        self,
+        tmpdir,
+        httpbin,
+        data1,
+        data2,
+        data3,
+    ):
+        with NamedTemporaryFile(dir=tmpdir, suffix=".yml") as f:
+            testpath = f.name
+            # NOTE: ``use_cassette`` is not okay with the file existing
+            #       already.  So we using ``.close()`` to not only
+            #       close but also delete the empty file, before we start.
+            f.close()
+            host, port = httpbin.host, httpbin.port
+            match_on = ["method", "uri", "body"]
+            with use_cassette(testpath, match_on=match_on):
+                conn1 = httplib.HTTPConnection(host, port)
+                conn1.request("POST", "/anything", body=data1)
+                conn1.getresponse()
+                conn2 = httplib.HTTPConnection(host, port)
+                conn2.request("POST", "/anything", body=data2)
+                conn2.getresponse()
+            with use_cassette(testpath, match_on=match_on) as cass:
+                conn3 = httplib.HTTPConnection(host, port)
+                conn3.request("POST", "/anything", body=data3)
+                conn3.getresponse()
+            assert cass.play_counts[0] == 0
+            assert cass.play_counts[1] == 1

tests/unit/test_util.py

@@ -0,0 +1,33 @@
+from io import BytesIO, StringIO
+
+import pytest
+
+from vcr import request
+from vcr.util import read_body
+
+
+@pytest.mark.parametrize(
+    "input_, expected_output",
+    [
+        (BytesIO(b"Stream"), b"Stream"),
+        (StringIO("Stream"), b"Stream"),
+        (iter(["StringIter"]), b"StringIter"),
+        (iter(["String", "Iter"]), b"StringIter"),
+        (iter([b"BytesIter"]), b"BytesIter"),
+        (iter([b"Bytes", b"Iter"]), b"BytesIter"),
+        (iter([70, 111, 111]), b"Foo"),
+        (iter([]), b""),
+        ("String", b"String"),
+        (b"Bytes", b"Bytes"),
+    ],
+)
+def test_read_body(input_, expected_output):
+    r = request.Request("POST", "http://host.com/", input_, {})
+    assert read_body(r) == expected_output
+
+
+def test_unsupported_read_body():
+    r = request.Request("POST", "http://host.com/", iter([[]]), {})
+    with pytest.raises(ValueError) as excinfo:
+        assert read_body(r)
+    assert excinfo.value.args == ("Body type <class 'list'> not supported",)

vcr/request.py

@@ -3,7 +3,7 @@
 from io import BytesIO
 from urllib.parse import parse_qsl, urlparse
 
-from .util import CaseInsensitiveDict
+from .util import CaseInsensitiveDict, _is_nonsequence_iterator
 
 log = logging.getLogger(__name__)
 
@@ -17,8 +17,11 @@ def __init__(self, method, uri, body, headers):
         self.method = method
         self.uri = uri
         self._was_file = hasattr(body, "read")
+        self._was_iter = _is_nonsequence_iterator(body)
         if self._was_file:
             self.body = body.read()
+        elif self._was_iter:
+            self.body = list(body)
         else:
             self.body = body
         self.headers = headers
@@ -36,7 +39,11 @@ def headers(self, value):
 
     @property
     def body(self):
-        return BytesIO(self._body) if self._was_file else self._body
+        if self._was_file:
+            return BytesIO(self._body)
+        if self._was_iter:
+            return iter(self._body)
+        return self._body
 
     @body.setter
     def body(self, value):

vcr/util.py

@@ -89,9 +89,28 @@ def composed(incoming):
     return composed
 
 
+def _is_nonsequence_iterator(obj):
+    return hasattr(obj, "__iter__") and not isinstance(
+        obj,
+        (bytearray, bytes, dict, list, str),
+    )
+
+
 def read_body(request):
     if hasattr(request.body, "read"):
         return request.body.read()
+    if _is_nonsequence_iterator(request.body):
+        body = list(request.body)
+        if body:
+            if isinstance(body[0], str):
+                return "".join(body).encode("utf-8")
+            elif isinstance(body[0], (bytes, bytearray)):
+                return b"".join(body)
+            elif isinstance(body[0], int):
+                return bytes(body)
+            else:
+                raise ValueError(f"Body type {type(body[0])} not supported")
+        return b""
     return request.body