Fix HTTPX support (part 2)

docs/installation.rst

@@ -22,6 +22,7 @@ The following HTTP libraries are supported:
 -  ``urllib2``
 -  ``urllib3``
 -  ``httpx``
+-  ``httpcore``
 
 Speed
 -----

setup.py

@@ -48,6 +48,7 @@ def find_version(*file_paths):
     "tests": [
         "aiohttp",
         "boto3",
+        "httpcore",
         "httplib2",
         "httpx",
         "pytest-aiohttp",

vcr/patch.py

@@ -92,12 +92,12 @@
 
 
 try:
-    import httpx
+    import httpcore
 except ImportError:  # pragma: no cover
     pass
 else:
-    _HttpxSyncClient_send_single_request = httpx.Client._send_single_request
-    _HttpxAsyncClient_send_single_request = httpx.AsyncClient._send_single_request
+    _HttpcoreConnectionPool_handle_request = httpcore.ConnectionPool.handle_request
+    _HttpcoreAsyncConnectionPool_handle_async_request = httpcore.AsyncConnectionPool.handle_async_request
 
 
 class CassettePatcherBuilder:
@@ -121,7 +121,7 @@ def build(self):
             self._httplib2(),
             self._tornado(),
             self._aiohttp(),
-            self._httpx(),
+            self._httpcore(),
             self._build_patchers_from_mock_triples(self._cassette.custom_patches),
         )
 
@@ -304,19 +304,22 @@ def _aiohttp(self):
             yield client.ClientSession, "_request", new_request
 
     @_build_patchers_from_mock_triples_decorator
-    def _httpx(self):
+    def _httpcore(self):
         try:
-            import httpx
+            import httpcore
         except ImportError:  # pragma: no cover
             return
         else:
-            from .stubs.httpx_stubs import async_vcr_send, sync_vcr_send
+            from .stubs.httpcore_stubs import vcr_handle_async_request, vcr_handle_request
 
-            new_async_client_send = async_vcr_send(self._cassette, _HttpxAsyncClient_send_single_request)
-            yield httpx.AsyncClient, "_send_single_request", new_async_client_send
+            new_handle_async_request = vcr_handle_async_request(
+                self._cassette,
+                _HttpcoreAsyncConnectionPool_handle_async_request,
+            )
+            yield httpcore.AsyncConnectionPool, "handle_async_request", new_handle_async_request
 
-            new_sync_client_send = sync_vcr_send(self._cassette, _HttpxSyncClient_send_single_request)
-            yield httpx.Client, "_send_single_request", new_sync_client_send
+            new_handle_request = vcr_handle_request(self._cassette, _HttpcoreConnectionPool_handle_request)
+            yield httpcore.ConnectionPool, "handle_request", new_handle_request
 
     def _urllib3_patchers(self, cpool, conn, stubs):
         http_connection_remover = ConnectionRemover(

vcr/stubs/httpcore_stubs.py

@@ -0,0 +1,215 @@
+import asyncio
+import functools
+import logging
+from collections import defaultdict
+from collections.abc import AsyncIterable, Iterable
+
+from httpcore import Response
+from httpcore._models import ByteStream
+
+from vcr.errors import CannotOverwriteExistingCassetteException
+from vcr.filters import decode_response
+from vcr.request import Request as VcrRequest
+from vcr.serializers.compat import convert_body_to_bytes
+
+_logger = logging.getLogger(__name__)
+
+
+async def _convert_byte_stream(stream):
+    if isinstance(stream, Iterable):
+        return list(stream)
+
+    if isinstance(stream, AsyncIterable):
+        return [part async for part in stream]
+
+    raise TypeError(
+        f"_convert_byte_stream: stream must be Iterable or AsyncIterable, got {type(stream).__name__}",
+    )
+
+
+def _serialize_headers(real_response):
+    """
+    Some headers can appear multiple times, like "Set-Cookie".
+    Therefore serialize every header key to a list of values.
+    """
+
+    headers = defaultdict(list)
+
+    for name, value in real_response.headers:
+        headers[name.decode("ascii")].append(value.decode("ascii"))
+
+    return dict(headers)
+
+
+async def _serialize_response(real_response):
+    # The reason_phrase may not exist
+    try:
+        reason_phrase = real_response.extensions["reason_phrase"].decode("ascii")
+    except KeyError:
+        reason_phrase = None
+
+    # Reading the response stream consumes the iterator, so we need to restore it afterwards
+    content = b"".join(await _convert_byte_stream(real_response.stream))
+    real_response.stream = ByteStream(content)
+
+    return {
+        "status": {"code": real_response.status, "message": reason_phrase},
+        "headers": _serialize_headers(real_response),
+        "body": {"string": content},
+    }
+
+
+def _deserialize_headers(headers):
+    """
+    httpcore accepts headers as list of tuples of header key and value.
+    """
+
+    return [
+        (name.encode("ascii"), value.encode("ascii")) for name, values in headers.items() for value in values
+    ]
+
+
+def _deserialize_response(vcr_response):
+    # Cassette format generated for HTTPX requests by older versions of
+    # vcrpy. We restructure the content to resemble what a regular
+    # cassette looks like.
+    if "status_code" in vcr_response:
+        vcr_response = decode_response(
+            convert_body_to_bytes(
+                {
+                    "headers": vcr_response["headers"],
+                    "body": {"string": vcr_response["content"]},
+                    "status": {"code": vcr_response["status_code"]},
+                },
+            ),
+        )
+        extensions = None
+    else:
+        extensions = (
+            {"reason_phrase": vcr_response["status"]["message"].encode("ascii")}
+            if vcr_response["status"]["message"]
+            else None
+        )
+
+    return Response(
+        vcr_response["status"]["code"],
+        headers=_deserialize_headers(vcr_response["headers"]),
+        content=vcr_response["body"]["string"],
+        extensions=extensions,
+    )
+
+
+async def _make_vcr_request(real_request):
+    # Reading the request stream consumes the iterator, so we need to restore it afterwards
+    body = b"".join(await _convert_byte_stream(real_request.stream))
+    real_request.stream = ByteStream(body)
+
+    uri = bytes(real_request.url).decode("ascii")
+
+    # As per HTTPX: If there are multiple headers with the same key, then we concatenate them with commas
+    headers = defaultdict(list)
+
+    for name, value in real_request.headers:
+        headers[name.decode("ascii")].append(value.decode("ascii"))
+
+    headers = {name: ", ".join(values) for name, values in headers.items()}
+
+    return VcrRequest(real_request.method.decode("ascii"), uri, body, headers)
+
+
+async def _vcr_request(cassette, real_request):
+    vcr_request = await _make_vcr_request(real_request)
+
+    if cassette.can_play_response_for(vcr_request):
+        return vcr_request, _play_responses(cassette, vcr_request)
+
+    if cassette.write_protected and cassette.filter_request(vcr_request):
+        raise CannotOverwriteExistingCassetteException(
+            cassette=cassette,
+            failed_request=vcr_request,
+        )
+
+    _logger.info("%s not in cassette, sending to real server", vcr_request)
+
+    return vcr_request, None
+
+
+async def _record_responses(cassette, vcr_request, real_response):
+    cassette.append(vcr_request, await _serialize_response(real_response))
+
+
+def _play_responses(cassette, vcr_request):
+    vcr_response = cassette.play_response(vcr_request)
+    real_response = _deserialize_response(vcr_response)
+
+    return real_response
+
+
+async def _vcr_handle_async_request(
+    cassette,
+    real_handle_async_request,
+    self,
+    real_request,
+):
+    vcr_request, vcr_response = await _vcr_request(cassette, real_request)
+
+    if vcr_response:
+        return vcr_response
+
+    real_response = await real_handle_async_request(self, real_request)
+    await _record_responses(cassette, vcr_request, real_response)
+
+    return real_response
+
+
+def vcr_handle_async_request(cassette, real_handle_async_request):
+    @functools.wraps(real_handle_async_request)
+    def _inner_handle_async_request(self, real_request):
+        return _vcr_handle_async_request(
+            cassette,
+            real_handle_async_request,
+            self,
+            real_request,
+        )
+
+    return _inner_handle_async_request
+
+
+def _run_async_function(sync_func, *args, **kwargs):
+    """
+    Safely run an asynchronous function from a synchronous context.
+    Handles both cases:
+    - An event loop is already running.
+    - No event loop exists yet.
+    """
+    try:
+        asyncio.get_running_loop()
+    except RuntimeError:
+        return asyncio.run(sync_func(*args, **kwargs))
+    else:
+        # If inside a running loop, create a task and wait for it
+        return asyncio.ensure_future(sync_func(*args, **kwargs))
+
+
+def _vcr_handle_request(cassette, real_handle_request, self, real_request):
+    vcr_request, vcr_response = _run_async_function(
+        _vcr_request,
+        cassette,
+        real_request,
+    )
+
+    if vcr_response:
+        return vcr_response
+
+    real_response = real_handle_request(self, real_request)
+    _run_async_function(_record_responses, cassette, vcr_request, real_response)
+
+    return real_response
+
+
+def vcr_handle_request(cassette, real_handle_request):
+    @functools.wraps(real_handle_request)
+    def _inner_handle_request(self, real_request):
+        return _vcr_handle_request(cassette, real_handle_request, self, real_request)
+
+    return _inner_handle_request