Shortcuts on this page
r m x toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1from __future__ import annotations
3import errno
4import logging
5import queue
6import sys
7import typing
8import warnings
9import weakref
10from socket import timeout as SocketTimeout
11from types import TracebackType
13from ._base_connection import _TYPE_BODY
14from ._collections import HTTPHeaderDict
15from ._request_methods import RequestMethods
16from .connection import (
17 BaseSSLError,
18 BrokenPipeError,
19 DummyConnection,
20 HTTPConnection,
21 HTTPException,
22 HTTPSConnection,
23 ProxyConfig,
24 _wrap_proxy_error,
25)
26from .connection import port_by_scheme as port_by_scheme
27from .exceptions import (
28 ClosedPoolError,
29 EmptyPoolError,
30 FullPoolError,
31 HostChangedError,
32 InsecureRequestWarning,
33 LocationValueError,
34 MaxRetryError,
35 NewConnectionError,
36 ProtocolError,
37 ProxyError,
38 ReadTimeoutError,
39 SSLError,
40 TimeoutError,
41)
42from .response import BaseHTTPResponse
43from .util.connection import is_connection_dropped
44from .util.proxy import connection_requires_http_tunnel
45from .util.request import _TYPE_BODY_POSITION, set_file_position
46from .util.retry import Retry
47from .util.ssl_match_hostname import CertificateError
48from .util.timeout import _DEFAULT_TIMEOUT, _TYPE_DEFAULT, Timeout
49from .util.url import Url, _encode_target
50from .util.url import _normalize_host as normalize_host
51from .util.url import parse_url
52from .util.util import to_str
54if typing.TYPE_CHECKING:
55 import ssl
57 from typing_extensions import Self
59 from ._base_connection import BaseHTTPConnection, BaseHTTPSConnection
61log = logging.getLogger(__name__)
63_TYPE_TIMEOUT = typing.Union[Timeout, float, _TYPE_DEFAULT, None]
66# Pool objects
67class ConnectionPool:
68 """
69 Base class for all connection pools, such as
70 :class:`.HTTPConnectionPool` and :class:`.HTTPSConnectionPool`.
72 .. note::
73 ConnectionPool.urlopen() does not normalize or percent-encode target URIs
74 which is useful if your target server doesn't support percent-encoded
75 target URIs.
76 """
78 scheme: str | None = None
79 QueueCls = queue.LifoQueue
81 def __init__(self, host: str, port: int | None = None) -> None:
82 if not host:
83 raise LocationValueError("No host specified.")
85 self.host = _normalize_host(host, scheme=self.scheme)
86 self.port = port
88 # This property uses 'normalize_host()' (not '_normalize_host()')
89 # to avoid removing square braces around IPv6 addresses.
90 # This value is sent to `HTTPConnection.set_tunnel()` if called
91 # because square braces are required for HTTP CONNECT tunneling.
92 self._tunnel_host = normalize_host(host, scheme=self.scheme).lower()
94 def __str__(self) -> str:
95 return f"{type(self).__name__}(host={self.host!r}, port={self.port!r})"
97 def __enter__(self) -> Self:
98 return self
100 def __exit__(
101 self,
102 exc_type: type[BaseException] | None,
103 exc_val: BaseException | None,
104 exc_tb: TracebackType | None,
105 ) -> typing.Literal[False]:
106 self.close()
107 # Return False to re-raise any potential exceptions
108 return False
110 def close(self) -> None:
111 """
112 Close all pooled connections and disable the pool.
113 """
116# This is taken from http://hg.python.org/cpython/file/7aaba721ebc0/Lib/socket.py#l252
117_blocking_errnos = {errno.EAGAIN, errno.EWOULDBLOCK}
120class HTTPConnectionPool(ConnectionPool, RequestMethods):
121 """
122 Thread-safe connection pool for one host.
124 :param host:
125 Host used for this HTTP Connection (e.g. "localhost"), passed into
126 :class:`http.client.HTTPConnection`.
128 :param port:
129 Port used for this HTTP Connection (None is equivalent to 80), passed
130 into :class:`http.client.HTTPConnection`.
132 :param timeout:
133 Socket timeout in seconds for each individual connection. This can
134 be a float or integer, which sets the timeout for the HTTP request,
135 or an instance of :class:`urllib3.util.Timeout` which gives you more
136 fine-grained control over request timeouts. After the constructor has
137 been parsed, this is always a `urllib3.util.Timeout` object.
139 :param maxsize:
140 Number of connections to save that can be reused. More than 1 is useful
141 in multithreaded situations. If ``block`` is set to False, more
142 connections will be created but they will not be saved once they've
143 been used.
145 :param block:
146 If set to True, no more than ``maxsize`` connections will be used at
147 a time. When no free connections are available, the call will block
148 until a connection has been released. This is a useful side effect for
149 particular multithreaded situations where one does not want to use more
150 than maxsize connections per host to prevent flooding.
152 :param headers:
153 Headers to include with all requests, unless other headers are given
154 explicitly.
156 :param retries:
157 Retry configuration to use by default with requests in this pool.
159 :param _proxy:
160 Parsed proxy URL, should not be used directly, instead, see
161 :class:`urllib3.ProxyManager`
163 :param _proxy_headers:
164 A dictionary with proxy headers, should not be used directly,
165 instead, see :class:`urllib3.ProxyManager`
167 :param \\**conn_kw:
168 Additional parameters are used to create fresh :class:`urllib3.connection.HTTPConnection`,
169 :class:`urllib3.connection.HTTPSConnection` instances.
170 """
172 scheme = "http"
173 ConnectionCls: type[BaseHTTPConnection] | type[BaseHTTPSConnection] = HTTPConnection
175 def __init__(
176 self,
177 host: str,
178 port: int | None = None,
179 timeout: _TYPE_TIMEOUT | None = _DEFAULT_TIMEOUT,
180 maxsize: int = 1,
181 block: bool = False,
182 headers: typing.Mapping[str, str] | None = None,
183 retries: Retry | bool | int | None = None,
184 _proxy: Url | None = None,
185 _proxy_headers: typing.Mapping[str, str] | None = None,
186 _proxy_config: ProxyConfig | None = None,
187 **conn_kw: typing.Any,
188 ):
189 ConnectionPool.__init__(self, host, port)
190 RequestMethods.__init__(self, headers)
192 if not isinstance(timeout, Timeout):
193 timeout = Timeout.from_float(timeout)
195 if retries is None:
196 retries = Retry.DEFAULT
198 self.timeout = timeout
199 self.retries = retries
201 self.pool: queue.LifoQueue[typing.Any] | None = self.QueueCls(maxsize)
202 self.block = block
204 self.proxy = _proxy
205 self.proxy_headers = _proxy_headers or {}
206 self.proxy_config = _proxy_config
208 # Fill the queue up so that doing get() on it will block properly
209 for _ in range(maxsize):
210 self.pool.put(None)
212 # These are mostly for testing and debugging purposes.
213 self.num_connections = 0
214 self.num_requests = 0
215 self.conn_kw = conn_kw
217 if self.proxy:
218 # Enable Nagle's algorithm for proxies, to avoid packet fragmentation.
219 # Defaulting `socket_options` to an empty list avoids it defaulting to
220 # ``HTTPConnection.default_socket_options``.
221 self.conn_kw.setdefault("socket_options", [])
223 self.conn_kw["proxy"] = self.proxy
224 self.conn_kw["proxy_config"] = self.proxy_config
226 # Do not pass 'self' as callback to 'finalize'.
227 # Then the 'finalize' would keep an endless living (leak) to self.
228 # By just passing a reference to the pool allows the garbage collector
229 # to free self if nobody else has a reference to it.
230 pool = self.pool
232 # Close all the HTTPConnections in the pool before the
233 # HTTPConnectionPool object is garbage collected.
234 weakref.finalize(self, _close_pool_connections, pool)
236 def _new_conn(self) -> BaseHTTPConnection:
237 """
238 Return a fresh :class:`HTTPConnection`.
239 """
240 self.num_connections += 1
241 log.debug(
242 "Starting new HTTP connection (%d): %s:%s",
243 self.num_connections,
244 self.host,
245 self.port or "80",
246 )
248 conn = self.ConnectionCls(
249 host=self.host,
250 port=self.port,
251 timeout=self.timeout.connect_timeout,
252 **self.conn_kw,
253 )
254 return conn
256 def _get_conn(self, timeout: float | None = None) -> BaseHTTPConnection:
257 """
258 Get a connection. Will return a pooled connection if one is available.
260 If no connections are available and :prop:`.block` is ``False``, then a
261 fresh connection is returned.
263 :param timeout:
264 Seconds to wait before giving up and raising
265 :class:`urllib3.exceptions.EmptyPoolError` if the pool is empty and
266 :prop:`.block` is ``True``.
267 """
268 conn = None
270 if self.pool is None:
271 raise ClosedPoolError(self, "Pool is closed.")
273 try:
274 conn = self.pool.get(block=self.block, timeout=timeout)
276 except AttributeError: # self.pool is None
277 raise ClosedPoolError(self, "Pool is closed.") from None # Defensive:
279 except queue.Empty:
280 if self.block:
281 raise EmptyPoolError(
282 self,
283 "Pool is empty and a new connection can't be opened due to blocking mode.",
284 ) from None
285 pass # Oh well, we'll create a new connection then
287 # If this is a persistent connection, check if it got disconnected
288 if conn and is_connection_dropped(conn):
289 log.debug("Resetting dropped connection: %s", self.host)
290 conn.close()
292 return conn or self._new_conn()
294 def _put_conn(self, conn: BaseHTTPConnection | None) -> None:
295 """
296 Put a connection back into the pool.
298 :param conn:
299 Connection object for the current host and port as returned by
300 :meth:`._new_conn` or :meth:`._get_conn`.
302 If the pool is already full, the connection is closed and discarded
303 because we exceeded maxsize. If connections are discarded frequently,
304 then maxsize should be increased.
306 If the pool is closed, then the connection will be closed and discarded.
307 """
308 if self.pool is not None:
309 try:
310 self.pool.put(conn, block=False)
311 return # Everything is dandy, done.
312 except AttributeError:
313 # self.pool is None.
314 pass
315 except queue.Full:
316 # Connection never got put back into the pool, close it.
317 if conn:
318 conn.close()
320 if self.block:
321 # This should never happen if you got the conn from self._get_conn
322 raise FullPoolError(
323 self,
324 "Pool reached maximum size and no more connections are allowed.",
325 ) from None
327 log.warning(
328 "Connection pool is full, discarding connection: %s. Connection pool size: %s",
329 self.host,
330 self.pool.qsize(),
331 )
333 # Connection never got put back into the pool, close it.
334 if conn:
335 conn.close()
337 def _validate_conn(self, conn: BaseHTTPConnection) -> None:
338 """
339 Called right before a request is made, after the socket is created.
340 """
342 def _prepare_proxy(self, conn: BaseHTTPConnection) -> None:
343 # Nothing to do for HTTP connections.
344 pass
346 def _get_timeout(self, timeout: _TYPE_TIMEOUT) -> Timeout:
347 """Helper that always returns a :class:`urllib3.util.Timeout`"""
348 if timeout is _DEFAULT_TIMEOUT:
349 return self.timeout.clone()
351 if isinstance(timeout, Timeout):
352 return timeout.clone()
353 else:
354 # User passed us an int/float. This is for backwards compatibility,
355 # can be removed later
356 return Timeout.from_float(timeout)
358 def _raise_timeout(
359 self,
360 err: BaseSSLError | OSError | SocketTimeout,
361 url: str,
362 timeout_value: _TYPE_TIMEOUT | None,
363 ) -> None:
364 """Is the error actually a timeout? Will raise a ReadTimeout or pass"""
366 if isinstance(err, SocketTimeout):
367 raise ReadTimeoutError(
368 self, url, f"Read timed out. (read timeout={timeout_value})"
369 ) from err
371 # See the above comment about EAGAIN in Python 3.
372 if hasattr(err, "errno") and err.errno in _blocking_errnos:
373 raise ReadTimeoutError(
374 self, url, f"Read timed out. (read timeout={timeout_value})"
375 ) from err
377 def _make_request(
378 self,
379 conn: BaseHTTPConnection,
380 method: str,
381 url: str,
382 body: _TYPE_BODY | None = None,
383 headers: typing.Mapping[str, str] | None = None,
384 retries: Retry | None = None,
385 timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT,
386 chunked: bool = False,
387 response_conn: BaseHTTPConnection | None = None,
388 preload_content: bool = True,
389 decode_content: bool = True,
390 enforce_content_length: bool = True,
391 ) -> BaseHTTPResponse:
392 """
393 Perform a request on a given urllib connection object taken from our
394 pool.
396 :param conn:
397 a connection from one of our connection pools
399 :param method:
400 HTTP request method (such as GET, POST, PUT, etc.)
402 :param url:
403 The URL to perform the request on.
405 :param body:
406 Data to send in the request body, either :class:`str`, :class:`bytes`,
407 an iterable of :class:`str`/:class:`bytes`, or a file-like object.
409 :param headers:
410 Dictionary of custom headers to send, such as User-Agent,
411 If-None-Match, etc. If None, pool headers are used. If provided,
412 these headers completely replace any pool-specific headers.
414 :param retries:
415 Configure the number of retries to allow before raising a
416 :class:`~urllib3.exceptions.MaxRetryError` exception.
418 Pass ``None`` to retry until you receive a response. Pass a
419 :class:`~urllib3.util.retry.Retry` object for fine-grained control
420 over different types of retries.
421 Pass an integer number to retry connection errors that many times,
422 but no other types of errors. Pass zero to never retry.
424 If ``False``, then retries are disabled and any exception is raised
425 immediately. Also, instead of raising a MaxRetryError on redirects,
426 the redirect response will be returned.
428 :type retries: :class:`~urllib3.util.retry.Retry`, False, or an int.
430 :param timeout:
431 If specified, overrides the default timeout for this one
432 request. It may be a float (in seconds) or an instance of
433 :class:`urllib3.util.Timeout`.
435 :param chunked:
436 If True, urllib3 will send the body using chunked transfer
437 encoding. Otherwise, urllib3 will send the body using the standard
438 content-length form. Defaults to False.
440 :param response_conn:
441 Set this to ``None`` if you will handle releasing the connection or
442 set the connection to have the response release it.
444 :param preload_content:
445 If True, the response's body will be preloaded during construction.
447 :param decode_content:
448 If True, will attempt to decode the body based on the
449 'content-encoding' header.
451 :param enforce_content_length:
452 Enforce content length checking. Body returned by server must match
453 value of Content-Length header, if present. Otherwise, raise error.
454 """
455 self.num_requests += 1
457 timeout_obj = self._get_timeout(timeout)
458 timeout_obj.start_connect()
459 conn.timeout = Timeout.resolve_default_timeout(timeout_obj.connect_timeout)
461 try:
462 # Trigger any extra validation we need to do.
463 try:
464 self._validate_conn(conn)
465 except (SocketTimeout, BaseSSLError) as e:
466 self._raise_timeout(err=e, url=url, timeout_value=conn.timeout)
467 raise
469 # _validate_conn() starts the connection to an HTTPS proxy
470 # so we need to wrap errors with 'ProxyError' here too.
471 except (
472 OSError,
473 NewConnectionError,
474 TimeoutError,
475 BaseSSLError,
476 CertificateError,
477 SSLError,
478 ) as e:
479 new_e: Exception = e
480 if isinstance(e, (BaseSSLError, CertificateError)):
481 new_e = SSLError(e)
482 # If the connection didn't successfully connect to it's proxy
483 # then there
484 if isinstance(
485 new_e, (OSError, NewConnectionError, TimeoutError, SSLError)
486 ) and (conn and conn.proxy and not conn.has_connected_to_proxy):
487 new_e = _wrap_proxy_error(new_e, conn.proxy.scheme)
488 raise new_e
490 # conn.request() calls http.client.*.request, not the method in
491 # urllib3.request. It also calls makefile (recv) on the socket.
492 try:
493 conn.request(
494 method,
495 url,
496 body=body,
497 headers=headers,
498 chunked=chunked,
499 preload_content=preload_content,
500 decode_content=decode_content,
501 enforce_content_length=enforce_content_length,
502 )
504 # We are swallowing BrokenPipeError (errno.EPIPE) since the server is
505 # legitimately able to close the connection after sending a valid response.
506 # With this behaviour, the received response is still readable.
507 except BrokenPipeError:
508 pass
509 except OSError as e:
510 # MacOS/Linux
511 # EPROTOTYPE and ECONNRESET are needed on macOS
512 # https://erickt.github.io/blog/2014/11/19/adventures-in-debugging-a-potential-osx-kernel-bug/
513 # Condition changed later to emit ECONNRESET instead of only EPROTOTYPE.
514 if e.errno != errno.EPROTOTYPE and e.errno != errno.ECONNRESET:
515 raise
517 # Reset the timeout for the recv() on the socket
518 read_timeout = timeout_obj.read_timeout
520 if not conn.is_closed:
521 # In Python 3 socket.py will catch EAGAIN and return None when you
522 # try and read into the file pointer created by http.client, which
523 # instead raises a BadStatusLine exception. Instead of catching
524 # the exception and assuming all BadStatusLine exceptions are read
525 # timeouts, check for a zero timeout before making the request.
526 if read_timeout == 0:
527 raise ReadTimeoutError(
528 self, url, f"Read timed out. (read timeout={read_timeout})"
529 )
530 conn.timeout = read_timeout
532 # Receive the response from the server
533 try:
534 response = conn.getresponse()
535 except (BaseSSLError, OSError) as e:
536 self._raise_timeout(err=e, url=url, timeout_value=read_timeout)
537 raise
539 # Set properties that are used by the pooling layer.
540 response.retries = retries
541 response._connection = response_conn # type: ignore[attr-defined]
542 response._pool = self # type: ignore[attr-defined]
544 log.debug(
545 '%s://%s:%s "%s %s %s" %s %s',
546 self.scheme,
547 self.host,
548 self.port,
549 method,
550 url,
551 response.version_string,
552 response.status,
553 response.length_remaining,
554 )
556 return response
558 def close(self) -> None:
559 """
560 Close all pooled connections and disable the pool.
561 """
562 if self.pool is None:
563 return
564 # Disable access to the pool
565 old_pool, self.pool = self.pool, None
567 # Close all the HTTPConnections in the pool.
568 _close_pool_connections(old_pool)
570 def is_same_host(self, url: str) -> bool:
571 """
572 Check if the given ``url`` is a member of the same host as this
573 connection pool.
574 """
575 if url.startswith("/"):
576 return True
578 # TODO: Add optional support for socket.gethostbyname checking.
579 scheme, _, host, port, *_ = parse_url(url)
580 scheme = scheme or "http"
581 if host is not None:
582 host = _normalize_host(host, scheme=scheme)
584 # Use explicit default port for comparison when none is given
585 if self.port and not port:
586 port = port_by_scheme.get(scheme)
587 elif not self.port and port == port_by_scheme.get(scheme):
588 port = None
590 return (scheme, host, port) == (self.scheme, self.host, self.port)
592 def urlopen( # type: ignore[override]
593 self,
594 method: str,
595 url: str,
596 body: _TYPE_BODY | None = None,
597 headers: typing.Mapping[str, str] | None = None,
598 retries: Retry | bool | int | None = None,
599 redirect: bool = True,
600 assert_same_host: bool = True,
601 timeout: _TYPE_TIMEOUT = _DEFAULT_TIMEOUT,
602 pool_timeout: int | None = None,
603 release_conn: bool | None = None,
604 chunked: bool = False,
605 body_pos: _TYPE_BODY_POSITION | None = None,
606 preload_content: bool = True,
607 decode_content: bool = True,
608 **response_kw: typing.Any,
609 ) -> BaseHTTPResponse:
610 """
611 Get a connection from the pool and perform an HTTP request. This is the
612 lowest level call for making a request, so you'll need to specify all
613 the raw details.
615 .. note::
617 More commonly, it's appropriate to use a convenience method
618 such as :meth:`request`.
620 .. note::
622 `release_conn` will only behave as expected if
623 `preload_content=False` because we want to make
624 `preload_content=False` the default behaviour someday soon without
625 breaking backwards compatibility.
627 :param method:
628 HTTP request method (such as GET, POST, PUT, etc.)
630 :param url:
631 The URL to perform the request on.
633 :param body:
634 Data to send in the request body, either :class:`str`, :class:`bytes`,
635 an iterable of :class:`str`/:class:`bytes`, or a file-like object.
637 :param headers:
638 Dictionary of custom headers to send, such as User-Agent,
639 If-None-Match, etc. If None, pool headers are used. If provided,
640 these headers completely replace any pool-specific headers.
642 :param retries:
643 Configure the number of retries to allow before raising a
644 :class:`~urllib3.exceptions.MaxRetryError` exception.
646 If ``None`` (default) will retry 3 times, see ``Retry.DEFAULT``. Pass a
647 :class:`~urllib3.util.retry.Retry` object for fine-grained control
648 over different types of retries.
649 Pass an integer number to retry connection errors that many times,
650 but no other types of errors. Pass zero to never retry.
652 If ``False``, then retries are disabled and any exception is raised
653 immediately. Also, instead of raising a MaxRetryError on redirects,
654 the redirect response will be returned.
656 :type retries: :class:`~urllib3.util.retry.Retry`, False, or an int.
658 :param redirect:
659 If True, automatically handle redirects (status codes 301, 302,
660 303, 307, 308). Each redirect counts as a retry. Disabling retries
661 will disable redirect, too.
663 :param assert_same_host:
664 If ``True``, will make sure that the host of the pool requests is
665 consistent else will raise HostChangedError. When ``False``, you can
666 use the pool on an HTTP proxy and request foreign hosts.
668 :param timeout:
669 If specified, overrides the default timeout for this one
670 request. It may be a float (in seconds) or an instance of
671 :class:`urllib3.util.Timeout`.
673 :param pool_timeout:
674 If set and the pool is set to block=True, then this method will
675 block for ``pool_timeout`` seconds and raise EmptyPoolError if no
676 connection is available within the time period.
678 :param bool preload_content:
679 If True, the response's body will be preloaded into memory.
681 :param bool decode_content:
682 If True, will attempt to decode the body based on the
683 'content-encoding' header.
685 :param release_conn:
686 If False, then the urlopen call will not release the connection
687 back into the pool once a response is received (but will release if
688 you read the entire contents of the response such as when
689 `preload_content=True`). This is useful if you're not preloading
690 the response's content immediately. You will need to call
691 ``r.release_conn()`` on the response ``r`` to return the connection
692 back into the pool. If None, it takes the value of ``preload_content``
693 which defaults to ``True``.
695 :param bool chunked:
696 If True, urllib3 will send the body using chunked transfer
697 encoding. Otherwise, urllib3 will send the body using the standard
698 content-length form. Defaults to False.
700 :param int body_pos:
701 Position to seek to in file-like body in the event of a retry or
702 redirect. Typically this won't need to be set because urllib3 will
703 auto-populate the value when needed.
704 """
705 # Ensure that the URL we're connecting to is properly encoded
706 if url.startswith("/"):
707 # URLs starting with / are inherently schemeless.
708 url = to_str(_encode_target(url))
709 destination_scheme = None
710 else:
711 parsed_url = parse_url(url)
712 destination_scheme = parsed_url.scheme
713 url = to_str(parsed_url.url)
715 if headers is None:
716 headers = self.headers
718 if not isinstance(retries, Retry):
719 retries = Retry.from_int(retries, redirect=redirect, default=self.retries)
721 if release_conn is None:
722 release_conn = preload_content
724 # Check host
725 if assert_same_host and not self.is_same_host(url):
726 raise HostChangedError(self, url, retries)
728 conn = None
730 # Track whether `conn` needs to be released before
731 # returning/raising/recursing. Update this variable if necessary, and
732 # leave `release_conn` constant throughout the function. That way, if
733 # the function recurses, the original value of `release_conn` will be
734 # passed down into the recursive call, and its value will be respected.
735 #
736 # See issue #651 [1] for details.
737 #
738 # [1] <https://github.com/urllib3/urllib3/issues/651>
739 release_this_conn = release_conn
741 http_tunnel_required = connection_requires_http_tunnel(
742 self.proxy, self.proxy_config, destination_scheme
743 )
745 # Merge the proxy headers. Only done when not using HTTP CONNECT. We
746 # have to copy the headers dict so we can safely change it without those
747 # changes being reflected in anyone else's copy.
748 if not http_tunnel_required:
749 headers = headers.copy() # type: ignore[attr-defined]
750 headers.update(self.proxy_headers) # type: ignore[union-attr]
752 # Must keep the exception bound to a separate variable or else Python 3
753 # complains about UnboundLocalError.
754 err = None
756 # Keep track of whether we cleanly exited the except block. This
757 # ensures we do proper cleanup in finally.
758 clean_exit = False
760 # Rewind body position, if needed. Record current position
761 # for future rewinds in the event of a redirect/retry.
762 body_pos = set_file_position(body, body_pos)
764 try:
765 # Request a connection from the queue.
766 timeout_obj = self._get_timeout(timeout)
767 conn = self._get_conn(timeout=pool_timeout)
769 conn.timeout = timeout_obj.connect_timeout # type: ignore[assignment]
771 # Is this a closed/new connection that requires CONNECT tunnelling?
772 if self.proxy is not None and http_tunnel_required and conn.is_closed:
773 try:
774 self._prepare_proxy(conn)
775 except (BaseSSLError, OSError, SocketTimeout) as e:
776 self._raise_timeout(
777 err=e, url=self.proxy.url, timeout_value=conn.timeout
778 )
779 raise
781 # If we're going to release the connection in ``finally:``, then
782 # the response doesn't need to know about the connection. Otherwise
783 # it will also try to release it and we'll have a double-release
784 # mess.
785 response_conn = conn if not release_conn else None
787 # Make the request on the HTTPConnection object
788 response = self._make_request(
789 conn,
790 method,
791 url,
792 timeout=timeout_obj,
793 body=body,
794 headers=headers,
795 chunked=chunked,
796 retries=retries,
797 response_conn=response_conn,
798 preload_content=preload_content,
799 decode_content=decode_content,
800 **response_kw,
801 )
803 # Everything went great!
804 clean_exit = True
806 except EmptyPoolError:
807 # Didn't get a connection from the pool, no need to clean up
808 clean_exit = True
809 release_this_conn = False
810 raise
812 except (
813 TimeoutError,
814 HTTPException,
815 OSError,
816 ProtocolError,
817 BaseSSLError,
818 SSLError,
819 CertificateError,
820 ProxyError,
821 ) as e:
822 # Discard the connection for these exceptions. It will be
823 # replaced during the next _get_conn() call.
824 clean_exit = False
825 new_e: Exception = e
826 if isinstance(e, (BaseSSLError, CertificateError)):
827 new_e = SSLError(e)
828 if isinstance(
829 new_e,
830 (
831 OSError,
832 NewConnectionError,
833 TimeoutError,
834 SSLError,
835 HTTPException,
836 ),
837 ) and (conn and conn.proxy and not conn.has_connected_to_proxy):
838 new_e = _wrap_proxy_error(new_e, conn.proxy.scheme)
839 elif isinstance(new_e, (OSError, HTTPException)):
840 new_e = ProtocolError("Connection aborted.", new_e)
842 retries = retries.increment(
843 method, url, error=new_e, _pool=self, _stacktrace=sys.exc_info()[2]
844 )
845 retries.sleep()
847 # Keep track of the error for the retry warning.
848 err = e
850 finally:
851 if not clean_exit:
852 # We hit some kind of exception, handled or otherwise. We need
853 # to throw the connection away unless explicitly told not to.
854 # Close the connection, set the variable to None, and make sure
855 # we put the None back in the pool to avoid leaking it.
856 if conn:
857 conn.close()
858 conn = None
859 release_this_conn = True
861 if release_this_conn:
862 # Put the connection back to be reused. If the connection is
863 # expired then it will be None, which will get replaced with a
864 # fresh connection during _get_conn.
865 self._put_conn(conn)
867 if not conn:
868 # Try again
869 log.warning(
870 "Retrying (%r) after connection broken by '%r': %s", retries, err, url
871 )
872 return self.urlopen(
873 method,
874 url,
875 body,
876 headers,
877 retries,
878 redirect,
879 assert_same_host,
880 timeout=timeout,
881 pool_timeout=pool_timeout,
882 release_conn=release_conn,
883 chunked=chunked,
884 body_pos=body_pos,
885 preload_content=preload_content,
886 decode_content=decode_content,
887 **response_kw,
888 )
890 # Handle redirect?
891 redirect_location = redirect and response.get_redirect_location()
892 if redirect_location:
893 if response.status == 303:
894 # Change the method according to RFC 9110, Section 15.4.4.
895 method = "GET"
896 # And lose the body not to transfer anything sensitive.
897 body = None
898 headers = HTTPHeaderDict(headers)._prepare_for_method_change()
900 # Strip headers marked as unsafe to forward to the redirected location.
901 # Check remove_headers_on_redirect to avoid a potential network call within
902 # self.is_same_host() which may use socket.gethostbyname() in the future.
903 if retries.remove_headers_on_redirect and not self.is_same_host(
904 redirect_location
905 ):
906 new_headers = headers.copy() # type: ignore[union-attr]
907 for header in headers:
908 if header.lower() in retries.remove_headers_on_redirect:
909 new_headers.pop(header, None)
910 headers = new_headers
912 try:
913 retries = retries.increment(method, url, response=response, _pool=self)
914 except MaxRetryError:
915 if retries.raise_on_redirect:
916 response.drain_conn()
917 raise
918 return response
920 response.drain_conn()
921 retries.sleep_for_retry(response)
922 log.debug("Redirecting %s -> %s", url, redirect_location)
923 return self.urlopen(
924 method,
925 redirect_location,
926 body,
927 headers,
928 retries=retries,
929 redirect=redirect,
930 assert_same_host=assert_same_host,
931 timeout=timeout,
932 pool_timeout=pool_timeout,
933 release_conn=release_conn,
934 chunked=chunked,
935 body_pos=body_pos,
936 preload_content=preload_content,
937 decode_content=decode_content,
938 **response_kw,
939 )
941 # Check if we should retry the HTTP response.
942 has_retry_after = bool(response.headers.get("Retry-After"))
943 if retries.is_retry(method, response.status, has_retry_after):
944 try:
945 retries = retries.increment(method, url, response=response, _pool=self)
946 except MaxRetryError:
947 if retries.raise_on_status:
948 response.drain_conn()
949 raise
950 return response
952 response.drain_conn()
953 retries.sleep(response)
954 log.debug("Retry: %s", url)
955 return self.urlopen(
956 method,
957 url,
958 body,
959 headers,
960 retries=retries,
961 redirect=redirect,
962 assert_same_host=assert_same_host,
963 timeout=timeout,
964 pool_timeout=pool_timeout,
965 release_conn=release_conn,
966 chunked=chunked,
967 body_pos=body_pos,
968 preload_content=preload_content,
969 decode_content=decode_content,
970 **response_kw,
971 )
973 return response
976class HTTPSConnectionPool(HTTPConnectionPool):
977 """
978 Same as :class:`.HTTPConnectionPool`, but HTTPS.
980 :class:`.HTTPSConnection` uses one of ``assert_fingerprint``,
981 ``assert_hostname`` and ``host`` in this order to verify connections.
982 If ``assert_hostname`` is False, no verification is done.
984 The ``key_file``, ``cert_file``, ``cert_reqs``, ``ca_certs``,
985 ``ca_cert_dir``, ``ssl_version``, ``key_password`` are only used if :mod:`ssl`
986 is available and are fed into :meth:`urllib3.util.ssl_wrap_socket` to upgrade
987 the connection socket into an SSL socket.
988 """
990 scheme = "https"
991 ConnectionCls: type[BaseHTTPSConnection] = HTTPSConnection
993 def __init__(
994 self,
995 host: str,
996 port: int | None = None,
997 timeout: _TYPE_TIMEOUT | None = _DEFAULT_TIMEOUT,
998 maxsize: int = 1,
999 block: bool = False,
1000 headers: typing.Mapping[str, str] | None = None,
1001 retries: Retry | bool | int | None = None,
1002 _proxy: Url | None = None,
1003 _proxy_headers: typing.Mapping[str, str] | None = None,
1004 key_file: str | None = None,
1005 cert_file: str | None = None,
1006 cert_reqs: int | str | None = None,
1007 key_password: str | None = None,
1008 ca_certs: str | None = None,
1009 ssl_version: int | str | None = None,
1010 ssl_minimum_version: ssl.TLSVersion | None = None,
1011 ssl_maximum_version: ssl.TLSVersion | None = None,
1012 assert_hostname: str | typing.Literal[False] | None = None,
1013 assert_fingerprint: str | None = None,
1014 ca_cert_dir: str | None = None,
1015 **conn_kw: typing.Any,
1016 ) -> None:
1017 super().__init__(
1018 host,
1019 port,
1020 timeout,
1021 maxsize,
1022 block,
1023 headers,
1024 retries,
1025 _proxy,
1026 _proxy_headers,
1027 **conn_kw,
1028 )
1030 self.key_file = key_file
1031 self.cert_file = cert_file
1032 self.cert_reqs = cert_reqs
1033 self.key_password = key_password
1034 self.ca_certs = ca_certs
1035 self.ca_cert_dir = ca_cert_dir
1036 self.ssl_version = ssl_version
1037 self.ssl_minimum_version = ssl_minimum_version
1038 self.ssl_maximum_version = ssl_maximum_version
1039 self.assert_hostname = assert_hostname
1040 self.assert_fingerprint = assert_fingerprint
1042 def _prepare_proxy(self, conn: HTTPSConnection) -> None: # type: ignore[override]
1043 """Establishes a tunnel connection through HTTP CONNECT."""
1044 if self.proxy and self.proxy.scheme == "https":
1045 tunnel_scheme = "https"
1046 else:
1047 tunnel_scheme = "http"
1049 conn.set_tunnel(
1050 scheme=tunnel_scheme,
1051 host=self._tunnel_host,
1052 port=self.port,
1053 headers=self.proxy_headers,
1054 )
1055 conn.connect()
1057 def _new_conn(self) -> BaseHTTPSConnection:
1058 """
1059 Return a fresh :class:`urllib3.connection.HTTPConnection`.
1060 """
1061 self.num_connections += 1
1062 log.debug(
1063 "Starting new HTTPS connection (%d): %s:%s",
1064 self.num_connections,
1065 self.host,
1066 self.port or "443",
1067 )
1069 if not self.ConnectionCls or self.ConnectionCls is DummyConnection: # type: ignore[comparison-overlap]
1070 raise ImportError(
1071 "Can't connect to HTTPS URL because the SSL module is not available."
1072 )
1074 actual_host: str = self.host
1075 actual_port = self.port
1076 if self.proxy is not None and self.proxy.host is not None:
1077 actual_host = self.proxy.host
1078 actual_port = self.proxy.port
1080 return self.ConnectionCls(
1081 host=actual_host,
1082 port=actual_port,
1083 timeout=self.timeout.connect_timeout,
1084 cert_file=self.cert_file,
1085 key_file=self.key_file,
1086 key_password=self.key_password,
1087 cert_reqs=self.cert_reqs,
1088 ca_certs=self.ca_certs,
1089 ca_cert_dir=self.ca_cert_dir,
1090 assert_hostname=self.assert_hostname,
1091 assert_fingerprint=self.assert_fingerprint,
1092 ssl_version=self.ssl_version,
1093 ssl_minimum_version=self.ssl_minimum_version,
1094 ssl_maximum_version=self.ssl_maximum_version,
1095 **self.conn_kw,
1096 )
1098 def _validate_conn(self, conn: BaseHTTPConnection) -> None:
1099 """
1100 Called right before a request is made, after the socket is created.
1101 """
1102 super()._validate_conn(conn)
1104 # Force connect early to allow us to validate the connection.
1105 if conn.is_closed:
1106 conn.connect()
1108 # TODO revise this, see https://github.com/urllib3/urllib3/issues/2791
1109 if not conn.is_verified and not conn.proxy_is_verified:
1110 warnings.warn(
1111 (
1112 f"Unverified HTTPS request is being made to host '{conn.host}'. "
1113 "Adding certificate verification is strongly advised. See: "
1114 "https://urllib3.readthedocs.io/en/latest/advanced-usage.html"
1115 "#tls-warnings"
1116 ),
1117 InsecureRequestWarning,
1118 )
1121def connection_from_url(url: str, **kw: typing.Any) -> HTTPConnectionPool:
1122 """
1123 Given a url, return an :class:`.ConnectionPool` instance of its host.
1125 This is a shortcut for not having to parse out the scheme, host, and port
1126 of the url before creating an :class:`.ConnectionPool` instance.
1128 :param url:
1129 Absolute URL string that must include the scheme. Port is optional.
1131 :param \\**kw:
1132 Passes additional parameters to the constructor of the appropriate
1133 :class:`.ConnectionPool`. Useful for specifying things like
1134 timeout, maxsize, headers, etc.
1136 Example::
1138 >>> conn = connection_from_url('http://google.com/')
1139 >>> r = conn.request('GET', '/')
1140 """
1141 scheme, _, host, port, *_ = parse_url(url)
1142 scheme = scheme or "http"
1143 port = port or port_by_scheme.get(scheme, 80)
1144 if scheme == "https":
1145 return HTTPSConnectionPool(host, port=port, **kw) # type: ignore[arg-type]
1146 else:
1147 return HTTPConnectionPool(host, port=port, **kw) # type: ignore[arg-type]
1150@typing.overload
1151def _normalize_host(host: None, scheme: str | None) -> None: ...
1154@typing.overload
1155def _normalize_host(host: str, scheme: str | None) -> str: ...
1158def _normalize_host(host: str | None, scheme: str | None) -> str | None:
1159 """
1160 Normalize hosts for comparisons and use with sockets.
1161 """
1163 host = normalize_host(host, scheme)
1165 # httplib doesn't like it when we include brackets in IPv6 addresses
1166 # Specifically, if we include brackets but also pass the port then
1167 # httplib crazily doubles up the square brackets on the Host header.
1168 # Instead, we need to make sure we never pass ``None`` as the port.
1169 # However, for backward compatibility reasons we can't actually
1170 # *assert* that. See http://bugs.python.org/issue28539
1171 if host and host.startswith("[") and host.endswith("]"):
1172 host = host[1:-1]
1173 return host
1176def _url_from_pool(
1177 pool: HTTPConnectionPool | HTTPSConnectionPool, path: str | None = None
1178) -> str:
1179 """Returns the URL from a given connection pool. This is mainly used for testing and logging."""
1180 return Url(scheme=pool.scheme, host=pool.host, port=pool.port, path=path).url
1183def _close_pool_connections(pool: queue.LifoQueue[typing.Any]) -> None:
1184 """Drains a queue of connections and closes each one."""
1185 try:
1186 while True:
1187 conn = pool.get(block=False)
1188 if conn:
1189 conn.close()
1190 except queue.Empty:
1191 pass # Done.