Add support for the no_proxy env var mechanism in the HTTP client

PR #4445 by @scirelli.
Fixes #4431.

Co-authored-by: Steve Cirelli <[email protected]>
Co-authored-by: Sviatoslav Sydorenko <[email protected]>
(cherry picked from commit 1a4126a)

CHANGES/4431.bugfix

@@ -0,0 +1 @@
+Fixed HTTP client requests to honor ``no_proxy`` environment variables.

CONTRIBUTORS.txt

@@ -254,6 +254,7 @@ Stanislav Prokop
 Stefan Tjarks
 Stepan Pletnev
 Stephan Jaensch
+Stephen Cirelli
 Stephen Granade
 Steven Seguin
 Sunghyun Hwang

aiohttp/client.py

@@ -8,6 +8,7 @@
 import sys
 import traceback
 import warnings
+from contextlib import suppress
 from types import SimpleNamespace, TracebackType
 from typing import (
     Any,
@@ -76,8 +77,8 @@
     BasicAuth,
     TimeoutHandle,
     ceil_timeout,
+    get_env_proxy_for_url,
     get_running_loop,
-    proxies_from_env,
     sentinel,
     strip_auth_from_url,
 )
@@ -483,11 +484,8 @@ async def _request(
                     if proxy is not None:
                         proxy = URL(proxy)
                     elif self._trust_env:
-                        for scheme, proxy_info in proxies_from_env().items():
-                            if scheme == url.scheme:
-                                proxy = proxy_info.proxy
-                                proxy_auth = proxy_info.proxy_auth
-                                break
+                        with suppress(LookupError):
+                            proxy, proxy_auth = get_env_proxy_for_url(url)
 
                     req = self._request_class(
                         method,

aiohttp/helpers.py

@@ -4,6 +4,7 @@
 import base64
 import binascii
 import cgi
+import dataclasses
 import datetime
 import functools
 import inspect
@@ -41,7 +42,7 @@
     cast,
 )
 from urllib.parse import quote
-from urllib.request import getproxies
+from urllib.request import getproxies, proxy_bypass
 
 import async_timeout
 import attr
@@ -298,6 +299,21 @@ def isasyncgenfunction(obj: Any) -> bool:
 
 
 @attr.s(auto_attribs=True, frozen=True, slots=True)
+def get_env_proxy_for_url(url: URL) -> Tuple[URL, Optional[BasicAuth]]:
+    """Get a permitted proxy for the given URL from the env."""
+    if url.host is not None and proxy_bypass(url.host):
+        raise LookupError(f"Proxying is disallowed for `{url.host!r}`")
+
+    proxies_in_env = proxies_from_env()
+    try:
+        proxy_info = proxies_in_env[url.scheme]
+    except KeyError:
+        raise LookupError(f"No proxies found for `{url!s}` in the env")
+    else:
+        return proxy_info.proxy, proxy_info.proxy_auth
+
+
+@dataclasses.dataclass(frozen=True)
 class MimeType:
     type: str
     subtype: str

tests/test_helpers.py

@@ -6,6 +6,7 @@
 import tempfile
 from math import isclose, modf
 from unittest import mock
+from urllib.request import getproxies_environment
 
 import pytest
 from multidict import MultiDict
@@ -497,6 +498,96 @@ async def test_get_running_loop_ok(loop) -> None:
     assert helpers.get_running_loop() is loop
 
 
+# --------------------- get_env_proxy_for_url ------------------------------
+
+
+@pytest.fixture
+def proxy_env_vars(monkeypatch, request):
+    for schema in getproxies_environment().keys():
+        monkeypatch.delenv(f"{schema}_proxy", False)
+
+    for proxy_type, proxy_list in request.param.items():
+        monkeypatch.setenv(proxy_type, proxy_list)
+
+    return request.param
+
+
+@pytest.mark.parametrize(
+    ("proxy_env_vars", "url_input", "expected_err_msg"),
+    (
+        (
+            {"no_proxy": "aiohttp.io"},
+            "http://aiohttp.io/path",
+            r"Proxying is disallowed for `'aiohttp.io'`",
+        ),
+        (
+            {"no_proxy": "aiohttp.io,proxy.com"},
+            "http://aiohttp.io/path",
+            r"Proxying is disallowed for `'aiohttp.io'`",
+        ),
+        (
+            {"http_proxy": "http://example.com"},
+            "https://aiohttp.io/path",
+            r"No proxies found for `https://aiohttp.io/path` in the env",
+        ),
+        (
+            {"https_proxy": "https://example.com"},
+            "http://aiohttp.io/path",
+            r"No proxies found for `http://aiohttp.io/path` in the env",
+        ),
+        (
+            {},
+            "https://aiohttp.io/path",
+            r"No proxies found for `https://aiohttp.io/path` in the env",
+        ),
+        (
+            {"https_proxy": "https://example.com"},
+            "",
+            r"No proxies found for `` in the env",
+        ),
+    ),
+    indirect=["proxy_env_vars"],
+    ids=(
+        "url_matches_the_no_proxy_list",
+        "url_matches_the_no_proxy_list_multiple",
+        "url_scheme_does_not_match_http_proxy_list",
+        "url_scheme_does_not_match_https_proxy_list",
+        "no_proxies_are_set",
+        "url_is_empty",
+    ),
+)
+@pytest.mark.usefixtures("proxy_env_vars")
+def test_get_env_proxy_for_url_negative(url_input, expected_err_msg) -> None:
+    url = URL(url_input)
+    with pytest.raises(LookupError, match=expected_err_msg):
+        helpers.get_env_proxy_for_url(url)
+
+
+@pytest.mark.parametrize(
+    ("proxy_env_vars", "url_input"),
+    (
+        ({"http_proxy": "http://example.com"}, "http://aiohttp.io/path"),
+        ({"https_proxy": "http://example.com"}, "https://aiohttp.io/path"),
+        (
+            {"http_proxy": "http://example.com,http://proxy.org"},
+            "http://aiohttp.io/path",
+        ),
+    ),
+    indirect=["proxy_env_vars"],
+    ids=(
+        "url_scheme_match_http_proxy_list",
+        "url_scheme_match_https_proxy_list",
+        "url_scheme_match_http_proxy_list_multiple",
+    ),
+)
+def test_get_env_proxy_for_url(proxy_env_vars, url_input) -> None:
+    url = URL(url_input)
+    proxy, proxy_auth = helpers.get_env_proxy_for_url(url)
+    proxy_list = proxy_env_vars[url.scheme + "_proxy"]
+    assert proxy == URL(proxy_list)
+    assert proxy_auth is None
+
+
 # ------------- set_result / set_exception ----------------------