Source code
from __future__ import annotations
import os
import sys
import time
from collections import OrderedDict
from collections.abc import Generator, Mapping, MutableMapping
from datetime import timedelta
from typing import TYPE_CHECKING, Any, cast
from ._internal_utils import to_native_string
from ._types import is_prepared as _is_prepared
from .adapters import HTTPAdapter
from .auth import _basic_auth_str
from .compat import cookielib, urljoin, urlparse
from .cookies import (
RequestsCookieJar,
cookiejar_from_dict,
extract_cookies_to_jar,
merge_cookies,
)
from .exceptions import (
ChunkedEncodingError,
ContentDecodingError,
InvalidSchema,
TooManyRedirects,
)
from .hooks import default_hooks, dispatch_hook
from .models import ( # noqa: F401
DEFAULT_REDIRECT_LIMIT,
REDIRECT_STATI,
PreparedRequest,
Request,
Response,
)
from .status_codes import codes
from .structures import CaseInsensitiveDict
from .utils import ( # noqa: F401
DEFAULT_PORTS,
default_headers,
get_auth_from_url,
get_environ_proxies,
get_netrc_auth,
requote_uri,
resolve_proxies,
rewind_body,
should_bypass_proxies, # re-export for external consumers
to_key_val_list,
)
class SessionRedirectMixin:
max_redirects: int
trust_env: bool
cookies: RequestsCookieJar
def send(self, request: PreparedRequest, **kwargs: Any) -> Response: ...
def get_redirect_target(self, resp: Response) -> str | None:
"""Receives a Response. Returns a redirect URI or ``None``"""
# Due to the nature of how requests processes redirects this method will
# be called at least once upon the original response and at least twice
# on each subsequent redirect response (if any).
# If a custom mixin is used to handle this logic, it may be advantageous
# to cache the redirect location onto the response object as a private
# attribute.
if resp.is_redirect:
location = resp.headers["location"]
# Currently the underlying http module on py3 decode headers
# in latin1, but empirical evidence suggests that latin1 is very
# rarely used with non-ASCII characters in HTTP headers.
# It is more likely to get UTF8 header rather than latin1.
# This causes incorrect handling of UTF8 encoded location headers.
# To solve this, we re-encode the location in latin1.
location = location.encode("latin1")
return to_native_string(location, "utf8")
return None
def should_strip_auth(self, old_url: str, new_url: str) -> bool:
"""Decide whether Authorization header should be removed when redirecting"""
old_parsed = urlparse(old_url)
new_parsed = urlparse(new_url)
if old_parsed.hostname != new_parsed.hostname:
return True
# Special case: allow http -> https redirect when using the standard
# ports. This isn't specified by RFC 7235, but is kept to avoid
# breaking backwards compatibility with older versions of requests
# that allowed any redirects on the same host.
if (
old_parsed.scheme == "http"
and old_parsed.port in (80, None)
and new_parsed.scheme == "https"
and new_parsed.port in (443, None)
):
return False
# Handle default port usage corresponding to scheme.
changed_scheme = old_parsed.scheme != new_parsed.scheme
default_port = (DEFAULT_PORTS.get(old_parsed.scheme, None), None)
if (
not changed_scheme
and old_parsed.port in default_port
and new_parsed.port in default_port
):
return False
# Standard case: root URI must match
return (old_parsed.port != new_parsed.port) or changed_scheme
def resolve_redirects(
self,
resp: Response,
req: PreparedRequest,
stream: bool = False,
timeout: _t.TimeoutType = None,
verify: _t.VerifyType = True,
cert: _t.CertType = None,
proxies: dict[str, str] | None = None,
yield_requests: bool = False,
**adapter_kwargs: Any,
) -> Generator[Response, None, None]:
"""Receives a Response. Returns a generator of Responses or Requests."""
hist: list[Response] = [] # keep track of history
url = self.get_redirect_target(resp)
previous_fragment = urlparse(req.url).fragment
while url:
prepared_request = req.copy()
# Update history and keep track of redirects.
resp.history = hist[:]
hist.append(resp)
try:
resp.content # Consume socket so it can be released
except (ChunkedEncodingError, ContentDecodingError, RuntimeError):
resp.raw.read(decode_content=False)
if len(resp.history) >= self.max_redirects:
raise TooManyRedirects(
f"Exceeded {self.max_redirects} redirects.", response=resp
)
# Release the connection back into the pool.
resp.close()
# Handle redirection without scheme (see: RFC 1808 Section 4)
if url.startswith("//"):
parsed_rurl = urlparse(resp.url)
url = ":".join([to_native_string(parsed_rurl.scheme), url])
# Normalize url case and attach previous fragment if needed (RFC 7231 7.1.2)
parsed = urlparse(url)
if parsed.fragment == "" and previous_fragment:
parsed = parsed._replace(fragment=previous_fragment)
elif parsed.fragment:
previous_fragment = parsed.fragment
url = parsed.geturl()
# Facilitate relative 'location' headers, as allowed by RFC 7231.
# (e.g. '/path/to/resource' instead of 'http://domain.tld/path/to/resource')
# Compliant with RFC3986, we percent encode the url.
if not parsed.netloc:
url = urljoin(resp.url, requote_uri(url))
else:
url = requote_uri(url)
prepared_request.url = to_native_string(url)
self.rebuild_method(prepared_request, resp)
# https://github.com/psf/requests/issues/1084
if resp.status_code not in (
codes.temporary_redirect,
codes.permanent_redirect,
):
# https://github.com/psf/requests/issues/3490
purged_headers = ("Content-Length", "Content-Type", "Transfer-Encoding")
for header in purged_headers:
prepared_request.headers.pop(header, None)
prepared_request.body = None
headers = prepared_request.headers
headers.pop("Cookie", None)
# Extract any cookies sent on the response to the cookiejar
# in the new request. Because we've mutated our copied prepared
# request, use the old one that we haven't yet touched.
cookie_jar = cast("CookieJar", prepared_request._cookies)
extract_cookies_to_jar(cookie_jar, req, resp.raw)
merge_cookies(cookie_jar, self.cookies)
prepared_request.prepare_cookies(cookie_jar)
# Rebuild auth and proxy information.
proxies = self.rebuild_proxies(prepared_request, proxies)
self.rebuild_auth(prepared_request, resp)
# A failed tell() sets `_body_position` to `object()`. This non-None
# value ensures `rewindable` will be True, allowing us to raise an
# UnrewindableBodyError, instead of hanging the connection.
rewindable = prepared_request._body_position is not None and (
"Content-Length" in headers or "Transfer-Encoding" in headers
)
# Attempt to rewind consumed file-like object.
if rewindable:
rewind_body(prepared_request)
# Override the original request.
req = prepared_request
if yield_requests:
yield req # type: ignore[misc] # Internal use only, returns PreparedRequest
else:
resp = self.send(
req,
stream=stream,
timeout=timeout,
verify=verify,
cert=cert,
proxies=proxies,
allow_redirects=False,
**adapter_kwargs,
)
extract_cookies_to_jar(self.cookies, prepared_request, resp.raw)
# extract redirect url, if any, for the next loop
url = self.get_redirect_target(resp)
yield resp
def rebuild_auth(
self, prepared_request: PreparedRequest, response: Response
) -> None:
"""When being redirected we may want to strip authentication from the
request to avoid leaking credentials. This method intelligently removes
and reapplies authentication where possible to avoid credential loss.
"""
original_request = response.request
assert _is_prepared(original_request)
assert _is_prepared(prepared_request)
headers = prepared_request.headers
original_url = original_request.url
url = prepared_request.url
if "Authorization" in headers and self.should_strip_auth(original_url, url):
# If we get redirected to a new host, we should strip out any
# authentication headers.
del headers["Authorization"]
# .netrc might have more auth for us on our new host.
new_auth = get_netrc_auth(url) if self.trust_env else None
if new_auth is not None:
prepared_request.prepare_auth(new_auth)
def rebuild_proxies(
self,
prepared_request: PreparedRequest,
proxies: dict[str, str] | None,
) -> dict[str, str]:
"""This method re-evaluates the proxy configuration by considering the
environment variables. If we are redirected to a URL covered by
NO_PROXY, we strip the proxy configuration. Otherwise, we set missing
proxy keys for this URL (in case they were stripped by a previous
redirect).
This method also replaces the Proxy-Authorization header where
necessary.
:rtype: dict
"""
assert _is_prepared(prepared_request)
headers = prepared_request.headers
scheme = urlparse(prepared_request.url).scheme
new_proxies = resolve_proxies(prepared_request, proxies, self.trust_env)
if "Proxy-Authorization" in headers:
del headers["Proxy-Authorization"]
try:
username, password = get_auth_from_url(new_proxies[scheme])
except KeyError:
username, password = None, None
# urllib3 handles proxy authorization for us in the standard adapter.
# Avoid appending this to TLS tunneled requests where it may be leaked.
if not scheme.startswith("https") and username and password:
headers["Proxy-Authorization"] = _basic_auth_str(username, password)
return new_proxies
def rebuild_method(
self, prepared_request: PreparedRequest, response: Response
) -> None:
"""When being redirected we may want to change the method of the request
based on certain specs or browser behavior.
"""
method = prepared_request.method
# https://tools.ietf.org/html/rfc7231#section-6.4.4
if response.status_code == codes.see_other and method != "HEAD":
method = "GET"
# Do what the browsers do, despite standards...
# First, turn 302s into GETs.
if response.status_code == codes.found and method != "HEAD":
method = "GET"
# Second, if a POST is responded to with a 301, turn it into a GET.
# This bizarre behaviour is explained in Issue 1704.
if response.status_code == codes.moved and method == "POST":
method = "GET"
prepared_request.method = method