blob: 7c975608b2ae7f404c95a74b3737f3ddf3247e07 [file] [log] [blame]
Greg Stein5e0fa402000-06-26 08:28:01 +00001"""HTTP/1.1 client library
Guido van Rossum41999c11997-12-09 00:12:23 +00002
Greg Stein5e0fa402000-06-26 08:28:01 +00003<intro stuff goes here>
4<other stuff, too>
Guido van Rossum41999c11997-12-09 00:12:23 +00005
Thomas Wouters0e3f5912006-08-11 14:57:12 +00006HTTPConnection goes through a number of "states", which define when a client
Greg Stein5e0fa402000-06-26 08:28:01 +00007may legally make another request or fetch the response for a particular
8request. This diagram details these state transitions:
Guido van Rossum41999c11997-12-09 00:12:23 +00009
Greg Stein5e0fa402000-06-26 08:28:01 +000010 (null)
11 |
12 | HTTPConnection()
13 v
14 Idle
15 |
16 | putrequest()
17 v
18 Request-started
19 |
20 | ( putheader() )* endheaders()
21 v
22 Request-sent
23 |
24 | response = getresponse()
25 v
26 Unread-response [Response-headers-read]
27 |\____________________
Tim Peters5ceadc82001-01-13 19:16:21 +000028 | |
29 | response.read() | putrequest()
30 v v
31 Idle Req-started-unread-response
32 ______/|
33 / |
34 response.read() | | ( putheader() )* endheaders()
35 v v
36 Request-started Req-sent-unread-response
37 |
38 | response.read()
39 v
40 Request-sent
Greg Stein5e0fa402000-06-26 08:28:01 +000041
42This diagram presents the following rules:
43 -- a second request may not be started until {response-headers-read}
44 -- a response [object] cannot be retrieved until {request-sent}
45 -- there is no differentiation between an unread response body and a
46 partially read response body
47
48Note: this enforcement is applied by the HTTPConnection class. The
49 HTTPResponse class does not enforce this state machine, which
50 implies sophisticated clients may accelerate the request/response
51 pipeline. Caution should be taken, though: accelerating the states
52 beyond the above pattern may imply knowledge of the server's
53 connection-close behavior for certain requests. For example, it
54 is impossible to tell whether the server will close the connection
55 UNTIL the response headers have been read; this means that further
56 requests cannot be placed into the pipeline until it is known that
57 the server will NOT be closing the connection.
58
59Logical State __state __response
60------------- ------- ----------
61Idle _CS_IDLE None
62Request-started _CS_REQ_STARTED None
63Request-sent _CS_REQ_SENT None
64Unread-response _CS_IDLE <response_class>
65Req-started-unread-response _CS_REQ_STARTED <response_class>
66Req-sent-unread-response _CS_REQ_SENT <response_class>
Guido van Rossum41999c11997-12-09 00:12:23 +000067"""
Guido van Rossum23acc951994-02-21 16:36:04 +000068
Barry Warsaw820c1202008-06-12 04:06:45 +000069import email.parser
70import email.message
Jeremy Hylton636950f2009-03-28 04:34:21 +000071import io
72import os
73import socket
Jeremy Hylton1afc1692008-06-18 20:49:58 +000074from urllib.parse import urlsplit
Thomas Wouters89d996e2007-09-08 17:39:28 +000075import warnings
Guido van Rossum23acc951994-02-21 16:36:04 +000076
Thomas Wouters47b49bf2007-08-30 22:15:33 +000077__all__ = ["HTTPResponse", "HTTPConnection",
Skip Montanaro951a8842001-06-01 16:25:38 +000078 "HTTPException", "NotConnected", "UnknownProtocol",
Jeremy Hylton7c75c992002-06-28 23:38:14 +000079 "UnknownTransferEncoding", "UnimplementedFileMode",
80 "IncompleteRead", "InvalidURL", "ImproperConnectionState",
81 "CannotSendRequest", "CannotSendHeader", "ResponseNotReady",
Georg Brandl6aab16e2006-02-17 19:17:25 +000082 "BadStatusLine", "error", "responses"]
Skip Montanaro2dd42762001-01-23 15:35:05 +000083
Guido van Rossum23acc951994-02-21 16:36:04 +000084HTTP_PORT = 80
Guido van Rossum09c8b6c1999-12-07 21:37:17 +000085HTTPS_PORT = 443
86
Greg Stein5e0fa402000-06-26 08:28:01 +000087_UNKNOWN = 'UNKNOWN'
88
89# connection states
90_CS_IDLE = 'Idle'
91_CS_REQ_STARTED = 'Request-started'
92_CS_REQ_SENT = 'Request-sent'
93
Martin v. Löwis39a31782004-09-18 09:03:49 +000094# status codes
95# informational
96CONTINUE = 100
97SWITCHING_PROTOCOLS = 101
98PROCESSING = 102
99
100# successful
101OK = 200
102CREATED = 201
103ACCEPTED = 202
104NON_AUTHORITATIVE_INFORMATION = 203
105NO_CONTENT = 204
106RESET_CONTENT = 205
107PARTIAL_CONTENT = 206
108MULTI_STATUS = 207
109IM_USED = 226
110
111# redirection
112MULTIPLE_CHOICES = 300
113MOVED_PERMANENTLY = 301
114FOUND = 302
115SEE_OTHER = 303
116NOT_MODIFIED = 304
117USE_PROXY = 305
118TEMPORARY_REDIRECT = 307
119
120# client error
121BAD_REQUEST = 400
122UNAUTHORIZED = 401
123PAYMENT_REQUIRED = 402
124FORBIDDEN = 403
125NOT_FOUND = 404
126METHOD_NOT_ALLOWED = 405
127NOT_ACCEPTABLE = 406
128PROXY_AUTHENTICATION_REQUIRED = 407
129REQUEST_TIMEOUT = 408
130CONFLICT = 409
131GONE = 410
132LENGTH_REQUIRED = 411
133PRECONDITION_FAILED = 412
134REQUEST_ENTITY_TOO_LARGE = 413
135REQUEST_URI_TOO_LONG = 414
136UNSUPPORTED_MEDIA_TYPE = 415
137REQUESTED_RANGE_NOT_SATISFIABLE = 416
138EXPECTATION_FAILED = 417
139UNPROCESSABLE_ENTITY = 422
140LOCKED = 423
141FAILED_DEPENDENCY = 424
142UPGRADE_REQUIRED = 426
143
144# server error
145INTERNAL_SERVER_ERROR = 500
146NOT_IMPLEMENTED = 501
147BAD_GATEWAY = 502
148SERVICE_UNAVAILABLE = 503
149GATEWAY_TIMEOUT = 504
150HTTP_VERSION_NOT_SUPPORTED = 505
151INSUFFICIENT_STORAGE = 507
152NOT_EXTENDED = 510
153
Georg Brandl6aab16e2006-02-17 19:17:25 +0000154# Mapping status codes to official W3C names
155responses = {
156 100: 'Continue',
157 101: 'Switching Protocols',
158
159 200: 'OK',
160 201: 'Created',
161 202: 'Accepted',
162 203: 'Non-Authoritative Information',
163 204: 'No Content',
164 205: 'Reset Content',
165 206: 'Partial Content',
166
167 300: 'Multiple Choices',
168 301: 'Moved Permanently',
169 302: 'Found',
170 303: 'See Other',
171 304: 'Not Modified',
172 305: 'Use Proxy',
173 306: '(Unused)',
174 307: 'Temporary Redirect',
175
176 400: 'Bad Request',
177 401: 'Unauthorized',
178 402: 'Payment Required',
179 403: 'Forbidden',
180 404: 'Not Found',
181 405: 'Method Not Allowed',
182 406: 'Not Acceptable',
183 407: 'Proxy Authentication Required',
184 408: 'Request Timeout',
185 409: 'Conflict',
186 410: 'Gone',
187 411: 'Length Required',
188 412: 'Precondition Failed',
189 413: 'Request Entity Too Large',
190 414: 'Request-URI Too Long',
191 415: 'Unsupported Media Type',
192 416: 'Requested Range Not Satisfiable',
193 417: 'Expectation Failed',
194
195 500: 'Internal Server Error',
196 501: 'Not Implemented',
197 502: 'Bad Gateway',
198 503: 'Service Unavailable',
199 504: 'Gateway Timeout',
200 505: 'HTTP Version Not Supported',
201}
202
Georg Brandl80ba8e82005-09-29 20:16:07 +0000203# maximal amount of data to read at one time in _safe_read
204MAXAMOUNT = 1048576
205
Barry Warsaw820c1202008-06-12 04:06:45 +0000206class HTTPMessage(email.message.Message):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000207 # XXX The only usage of this method is in
208 # http.server.CGIHTTPRequestHandler. Maybe move the code there so
209 # that it doesn't need to be part of the public API. The API has
210 # never been defined so this could cause backwards compatibility
211 # issues.
212
Barry Warsaw820c1202008-06-12 04:06:45 +0000213 def getallmatchingheaders(self, name):
214 """Find all header lines matching a given header name.
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000215
Barry Warsaw820c1202008-06-12 04:06:45 +0000216 Look through the list of headers and find all lines matching a given
217 header name (and their continuation lines). A list of the lines is
218 returned, without interpretation. If the header does not occur, an
219 empty list is returned. If the header occurs multiple times, all
220 occurrences are returned. Case is not important in the header name.
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000221
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000222 """
Barry Warsaw820c1202008-06-12 04:06:45 +0000223 name = name.lower() + ':'
224 n = len(name)
225 lst = []
226 hit = 0
227 for line in self.keys():
228 if line[:n].lower() == name:
229 hit = 1
230 elif not line[:1].isspace():
231 hit = 0
232 if hit:
233 lst.append(line)
234 return lst
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000235
Jeremy Hylton98eb6c22009-03-27 18:31:36 +0000236def parse_headers(fp, _class=HTTPMessage):
Barry Warsaw820c1202008-06-12 04:06:45 +0000237 """Parses only RFC2822 headers from a file pointer.
238
239 email Parser wants to see strings rather than bytes.
240 But a TextIOWrapper around self.rfile would buffer too many bytes
241 from the stream, bytes which we later need to read as bytes.
242 So we read the correct bytes here, as bytes, for email Parser
243 to parse.
244
245 """
Barry Warsaw820c1202008-06-12 04:06:45 +0000246 headers = []
247 while True:
248 line = fp.readline()
249 headers.append(line)
250 if line in (b'\r\n', b'\n', b''):
251 break
252 hstring = b''.join(headers).decode('iso-8859-1')
Jeremy Hylton98eb6c22009-03-27 18:31:36 +0000253 return email.parser.Parser(_class=_class).parsestr(hstring)
Greg Stein5e0fa402000-06-26 08:28:01 +0000254
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000255
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000256_strict_sentinel = object()
257
258class HTTPResponse(io.RawIOBase):
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000259
260 # See RFC 2616 sec 19.6 and RFC 1945 sec 6 for details.
261
Jeremy Hylton811fc142007-08-03 13:30:02 +0000262 # The bytes from the socket object are iso-8859-1 strings.
263 # See RFC 2616 sec 2.2 which notes an exception for MIME-encoded
264 # text following RFC 2047. The basic status line parsing only
265 # accepts iso-8859-1.
266
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000267 def __init__(self, sock, debuglevel=0, strict=_strict_sentinel, method=None, url=None):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000268 # If the response includes a content-length header, we need to
269 # make sure that the client doesn't read more than the
Jeremy Hylton39b198d2007-08-04 19:22:00 +0000270 # specified number of bytes. If it does, it will block until
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000271 # the server times out and closes the connection. This will
272 # happen if a self.fp.read() is done (without a size) whether
273 # self.fp is buffered or not. So, no self.fp.read() by
274 # clients unless they know what they are doing.
Benjamin Petersonf72d9fb2009-02-08 00:29:20 +0000275 self.fp = sock.makefile("rb")
Jeremy Hylton30f86742000-09-18 22:50:38 +0000276 self.debuglevel = debuglevel
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000277 if strict is not _strict_sentinel:
278 warnings.warn("the 'strict' argument isn't supported anymore; "
279 "http.client now always assumes HTTP/1.x compliant servers.",
280 DeprecationWarning, 2)
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000281 self._method = method
Greg Stein5e0fa402000-06-26 08:28:01 +0000282
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000283 # The HTTPResponse object is returned via urllib. The clients
284 # of http and urllib expect different attributes for the
285 # headers. headers is used here and supports urllib. msg is
286 # provided as a backwards compatibility layer for http
287 # clients.
288
289 self.headers = self.msg = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000290
Greg Steindd6eefb2000-07-18 09:09:48 +0000291 # from the Status-Line of the response
Tim Peters07e99cb2001-01-14 23:47:14 +0000292 self.version = _UNKNOWN # HTTP-Version
293 self.status = _UNKNOWN # Status-Code
294 self.reason = _UNKNOWN # Reason-Phrase
Greg Stein5e0fa402000-06-26 08:28:01 +0000295
Tim Peters07e99cb2001-01-14 23:47:14 +0000296 self.chunked = _UNKNOWN # is "chunked" being used?
297 self.chunk_left = _UNKNOWN # bytes left to read in current chunk
298 self.length = _UNKNOWN # number of bytes left in response
299 self.will_close = _UNKNOWN # conn will close at end of response
Greg Stein5e0fa402000-06-26 08:28:01 +0000300
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000301 def _read_status(self):
Jeremy Hylton811fc142007-08-03 13:30:02 +0000302 line = str(self.fp.readline(), "iso-8859-1")
Jeremy Hylton30f86742000-09-18 22:50:38 +0000303 if self.debuglevel > 0:
Guido van Rossumbe19ed72007-02-09 05:37:30 +0000304 print("reply:", repr(line))
Jeremy Hyltonb6769522003-06-29 17:55:05 +0000305 if not line:
306 # Presumably, the server closed the connection before
307 # sending a valid response.
308 raise BadStatusLine(line)
Greg Steindd6eefb2000-07-18 09:09:48 +0000309 try:
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000310 version, status, reason = line.split(None, 2)
Greg Steindd6eefb2000-07-18 09:09:48 +0000311 except ValueError:
312 try:
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000313 version, status = line.split(None, 1)
Greg Steindd6eefb2000-07-18 09:09:48 +0000314 reason = ""
315 except ValueError:
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000316 # empty version will cause next test to fail.
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000317 version = ""
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000318 if not version.startswith("HTTP/"):
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000319 self.close()
320 raise BadStatusLine(line)
Greg Stein5e0fa402000-06-26 08:28:01 +0000321
Jeremy Hylton23d40472001-04-13 14:57:08 +0000322 # The status code is a three-digit number
323 try:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000324 status = int(status)
Jeremy Hylton23d40472001-04-13 14:57:08 +0000325 if status < 100 or status > 999:
326 raise BadStatusLine(line)
327 except ValueError:
328 raise BadStatusLine(line)
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000329 return version, status, reason
Greg Stein5e0fa402000-06-26 08:28:01 +0000330
Jeremy Hylton39c03802002-07-12 14:04:09 +0000331 def begin(self):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000332 if self.headers is not None:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000333 # we've already started reading the response
334 return
335
336 # read until we get a non-100 response
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000337 while True:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000338 version, status, reason = self._read_status()
Martin v. Löwis39a31782004-09-18 09:03:49 +0000339 if status != CONTINUE:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000340 break
341 # skip the header from the 100 response
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000342 while True:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000343 skip = self.fp.readline().strip()
344 if not skip:
345 break
346 if self.debuglevel > 0:
Guido van Rossumbe19ed72007-02-09 05:37:30 +0000347 print("header:", skip)
Tim Petersc411dba2002-07-16 21:35:23 +0000348
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000349 self.code = self.status = status
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000350 self.reason = reason.strip()
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000351 if version in ("HTTP/1.0", "HTTP/0.9"):
352 # Some servers might still return "0.9", treat it as 1.0 anyway
Greg Steindd6eefb2000-07-18 09:09:48 +0000353 self.version = 10
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000354 elif version.startswith("HTTP/1."):
Tim Peters07e99cb2001-01-14 23:47:14 +0000355 self.version = 11 # use HTTP/1.1 code for HTTP/1.x where x>=1
Greg Steindd6eefb2000-07-18 09:09:48 +0000356 else:
357 raise UnknownProtocol(version)
Greg Stein5e0fa402000-06-26 08:28:01 +0000358
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000359 self.headers = self.msg = parse_headers(self.fp)
Barry Warsaw820c1202008-06-12 04:06:45 +0000360
Jeremy Hylton30f86742000-09-18 22:50:38 +0000361 if self.debuglevel > 0:
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000362 for hdr in self.headers:
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000363 print("header:", hdr, end=" ")
Greg Stein5e0fa402000-06-26 08:28:01 +0000364
Greg Steindd6eefb2000-07-18 09:09:48 +0000365 # are we using the chunked-style of transfer encoding?
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000366 tr_enc = self.headers.get("transfer-encoding")
Jeremy Hyltond229b3a2002-09-03 19:24:24 +0000367 if tr_enc and tr_enc.lower() == "chunked":
Jeremy Hylton236156f2008-12-15 03:00:50 +0000368 self.chunked = True
Greg Steindd6eefb2000-07-18 09:09:48 +0000369 self.chunk_left = None
370 else:
Jeremy Hylton236156f2008-12-15 03:00:50 +0000371 self.chunked = False
Greg Stein5e0fa402000-06-26 08:28:01 +0000372
Greg Steindd6eefb2000-07-18 09:09:48 +0000373 # will the connection close at the end of the response?
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000374 self.will_close = self._check_close()
Greg Stein5e0fa402000-06-26 08:28:01 +0000375
Greg Steindd6eefb2000-07-18 09:09:48 +0000376 # do we have a Content-Length?
377 # NOTE: RFC 2616, S4.4, #3 says we ignore this if tr_enc is "chunked"
Jeremy Hylton0ee5eeb2007-08-04 03:25:17 +0000378 self.length = None
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000379 length = self.headers.get("content-length")
Barry Warsaw820c1202008-06-12 04:06:45 +0000380
381 # are we using the chunked-style of transfer encoding?
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000382 tr_enc = self.headers.get("transfer-encoding")
Greg Steindd6eefb2000-07-18 09:09:48 +0000383 if length and not self.chunked:
Jeremy Hylton30a81812000-09-14 20:34:27 +0000384 try:
385 self.length = int(length)
386 except ValueError:
Christian Heimesa612dc02008-02-24 13:08:18 +0000387 self.length = None
388 else:
389 if self.length < 0: # ignore nonsensical negative lengths
390 self.length = None
391 else:
392 self.length = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000393
Greg Steindd6eefb2000-07-18 09:09:48 +0000394 # does the body have a fixed length? (of zero)
Martin v. Löwis39a31782004-09-18 09:03:49 +0000395 if (status == NO_CONTENT or status == NOT_MODIFIED or
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000396 100 <= status < 200 or # 1xx codes
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000397 self._method == "HEAD"):
Greg Steindd6eefb2000-07-18 09:09:48 +0000398 self.length = 0
Greg Stein5e0fa402000-06-26 08:28:01 +0000399
Greg Steindd6eefb2000-07-18 09:09:48 +0000400 # if the connection remains open, and we aren't using chunked, and
401 # a content-length was not provided, then assume that the connection
402 # WILL close.
Jeremy Hylton0ee5eeb2007-08-04 03:25:17 +0000403 if (not self.will_close and
404 not self.chunked and
405 self.length is None):
Jeremy Hylton236156f2008-12-15 03:00:50 +0000406 self.will_close = True
Greg Stein5e0fa402000-06-26 08:28:01 +0000407
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000408 def _check_close(self):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000409 conn = self.headers.get("connection")
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000410 if self.version == 11:
411 # An HTTP/1.1 proxy is assumed to stay open unless
412 # explicitly closed.
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000413 conn = self.headers.get("connection")
Raymond Hettingerbac788a2004-05-04 09:21:43 +0000414 if conn and "close" in conn.lower():
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000415 return True
416 return False
417
Jeremy Hylton2c178252004-08-07 16:28:14 +0000418 # Some HTTP/1.0 implementations have support for persistent
419 # connections, using rules different than HTTP/1.1.
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000420
Christian Heimes895627f2007-12-08 17:28:33 +0000421 # For older HTTP, Keep-Alive indicates persistent connection.
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000422 if self.headers.get("keep-alive"):
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000423 return False
Tim Peters77c06fb2002-11-24 02:35:35 +0000424
Jeremy Hylton2c178252004-08-07 16:28:14 +0000425 # At least Akamai returns a "Connection: Keep-Alive" header,
426 # which was supposed to be sent by the client.
427 if conn and "keep-alive" in conn.lower():
428 return False
429
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000430 # Proxy-Connection is a netscape hack.
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000431 pconn = self.headers.get("proxy-connection")
Raymond Hettingerbac788a2004-05-04 09:21:43 +0000432 if pconn and "keep-alive" in pconn.lower():
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000433 return False
434
435 # otherwise, assume it will close
436 return True
437
Greg Steindd6eefb2000-07-18 09:09:48 +0000438 def close(self):
439 if self.fp:
440 self.fp.close()
441 self.fp = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000442
Jeremy Hyltondf5f6b52007-08-08 17:36:33 +0000443 # These implementations are for the benefit of io.BufferedReader.
444
445 # XXX This class should probably be revised to act more like
446 # the "raw stream" that BufferedReader expects.
447
448 @property
449 def closed(self):
450 return self.isclosed()
451
452 def flush(self):
453 self.fp.flush()
454
Jeremy Hyltona7cff022009-04-01 02:35:56 +0000455 def readable(self):
456 return True
457
Jeremy Hyltondf5f6b52007-08-08 17:36:33 +0000458 # End of "raw stream" methods
459
Greg Steindd6eefb2000-07-18 09:09:48 +0000460 def isclosed(self):
461 # NOTE: it is possible that we will not ever call self.close(). This
462 # case occurs when will_close is TRUE, length is None, and we
463 # read up to the last byte, but NOT past it.
464 #
465 # IMPLIES: if will_close is FALSE, then self.close() will ALWAYS be
466 # called, meaning self.isclosed() is meaningful.
467 return self.fp is None
468
469 def read(self, amt=None):
470 if self.fp is None:
Guido van Rossum8ce8a782007-11-01 19:42:39 +0000471 return b""
Greg Steindd6eefb2000-07-18 09:09:48 +0000472
Senthil Kumaran71fb6c82010-04-28 17:39:48 +0000473 if self._method == "HEAD":
Senthil Kumaran447f82f2010-06-04 16:38:00 +0000474 self.close()
Senthil Kumaran71fb6c82010-04-28 17:39:48 +0000475 return b""
476
Greg Steindd6eefb2000-07-18 09:09:48 +0000477 if self.chunked:
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000478 return self._read_chunked(amt)
Tim Peters230a60c2002-11-09 05:08:07 +0000479
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000480 if amt is None:
Greg Steindd6eefb2000-07-18 09:09:48 +0000481 # unbounded read
Jeremy Hyltondef9d2a2004-11-07 16:13:49 +0000482 if self.length is None:
Greg Steindd6eefb2000-07-18 09:09:48 +0000483 s = self.fp.read()
484 else:
485 s = self._safe_read(self.length)
Jeremy Hyltondef9d2a2004-11-07 16:13:49 +0000486 self.length = 0
Tim Peters07e99cb2001-01-14 23:47:14 +0000487 self.close() # we read everything
Greg Steindd6eefb2000-07-18 09:09:48 +0000488 return s
489
490 if self.length is not None:
491 if amt > self.length:
492 # clip the read to the "end of response"
493 amt = self.length
Greg Steindd6eefb2000-07-18 09:09:48 +0000494
495 # we do not use _safe_read() here because this may be a .will_close
496 # connection, and the user is reading more bytes than will be provided
497 # (for example, reading in 1k chunks)
498 s = self.fp.read(amt)
Jeremy Hyltondef9d2a2004-11-07 16:13:49 +0000499 if self.length is not None:
500 self.length -= len(s)
Guido van Rossum8ce8a782007-11-01 19:42:39 +0000501 if not self.length:
502 self.close()
Greg Steindd6eefb2000-07-18 09:09:48 +0000503 return s
504
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000505 def _read_chunked(self, amt):
506 assert self.chunked != _UNKNOWN
507 chunk_left = self.chunk_left
Chris Withers9c33cbf2009-09-04 17:04:16 +0000508 value = []
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000509 while True:
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000510 if chunk_left is None:
511 line = self.fp.readline()
Georg Brandl95ba4692008-01-26 09:45:58 +0000512 i = line.find(b";")
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000513 if i >= 0:
514 line = line[:i] # strip chunk-extensions
Christian Heimesa612dc02008-02-24 13:08:18 +0000515 try:
516 chunk_left = int(line, 16)
517 except ValueError:
518 # close the connection as protocol synchronisation is
519 # probably lost
520 self.close()
Chris Withers9c33cbf2009-09-04 17:04:16 +0000521 raise IncompleteRead(b''.join(value))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000522 if chunk_left == 0:
523 break
524 if amt is None:
Chris Withers9c33cbf2009-09-04 17:04:16 +0000525 value.append(self._safe_read(chunk_left))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000526 elif amt < chunk_left:
Chris Withers9c33cbf2009-09-04 17:04:16 +0000527 value.append(self._safe_read(amt))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000528 self.chunk_left = chunk_left - amt
Chris Withers9c33cbf2009-09-04 17:04:16 +0000529 return b''.join(value)
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000530 elif amt == chunk_left:
Chris Withers9c33cbf2009-09-04 17:04:16 +0000531 value.append(self._safe_read(amt))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000532 self._safe_read(2) # toss the CRLF at the end of the chunk
533 self.chunk_left = None
Chris Withers9c33cbf2009-09-04 17:04:16 +0000534 return b''.join(value)
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000535 else:
Chris Withers9c33cbf2009-09-04 17:04:16 +0000536 value.append(self._safe_read(chunk_left))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000537 amt -= chunk_left
538
539 # we read the whole chunk, get another
540 self._safe_read(2) # toss the CRLF at the end of the chunk
541 chunk_left = None
542
543 # read and discard trailer up to the CRLF terminator
544 ### note: we shouldn't have any trailers!
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000545 while True:
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000546 line = self.fp.readline()
Christian Heimes0bd4e112008-02-12 22:59:25 +0000547 if not line:
548 # a vanishingly small number of sites EOF without
549 # sending the trailer
550 break
Guido van Rossum8ce8a782007-11-01 19:42:39 +0000551 if line == b"\r\n":
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000552 break
553
554 # we read everything; close the "file"
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000555 self.close()
556
Chris Withers9c33cbf2009-09-04 17:04:16 +0000557 return b''.join(value)
Tim Peters230a60c2002-11-09 05:08:07 +0000558
Greg Steindd6eefb2000-07-18 09:09:48 +0000559 def _safe_read(self, amt):
560 """Read the number of bytes requested, compensating for partial reads.
561
562 Normally, we have a blocking socket, but a read() can be interrupted
563 by a signal (resulting in a partial read).
564
565 Note that we cannot distinguish between EOF and an interrupt when zero
566 bytes have been read. IncompleteRead() will be raised in this
567 situation.
568
569 This function should be used when <amt> bytes "should" be present for
570 reading. If the bytes are truly not available (due to EOF), then the
571 IncompleteRead exception can be used to detect the problem.
572 """
Georg Brandl80ba8e82005-09-29 20:16:07 +0000573 s = []
Greg Steindd6eefb2000-07-18 09:09:48 +0000574 while amt > 0:
Georg Brandl80ba8e82005-09-29 20:16:07 +0000575 chunk = self.fp.read(min(amt, MAXAMOUNT))
Greg Steindd6eefb2000-07-18 09:09:48 +0000576 if not chunk:
Benjamin Peterson6accb982009-03-02 22:50:25 +0000577 raise IncompleteRead(b''.join(s), amt)
Georg Brandl80ba8e82005-09-29 20:16:07 +0000578 s.append(chunk)
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000579 amt -= len(chunk)
Guido van Rossuma00f1232007-09-12 19:43:09 +0000580 return b"".join(s)
Greg Steindd6eefb2000-07-18 09:09:48 +0000581
Antoine Pitroub353c122009-02-11 00:39:14 +0000582 def fileno(self):
583 return self.fp.fileno()
584
Greg Steindd6eefb2000-07-18 09:09:48 +0000585 def getheader(self, name, default=None):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000586 if self.headers is None:
Greg Steindd6eefb2000-07-18 09:09:48 +0000587 raise ResponseNotReady()
Senthil Kumaran9f8dc442010-08-02 11:04:58 +0000588 headers = self.headers.get_all(name) or default
589 if isinstance(headers, str) or not hasattr(headers, '__iter__'):
590 return headers
591 else:
592 return ', '.join(headers)
Greg Stein5e0fa402000-06-26 08:28:01 +0000593
Martin v. Löwisdeacce22004-08-18 12:46:26 +0000594 def getheaders(self):
595 """Return list of (header, value) tuples."""
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000596 if self.headers is None:
Martin v. Löwisdeacce22004-08-18 12:46:26 +0000597 raise ResponseNotReady()
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000598 return list(self.headers.items())
Martin v. Löwisdeacce22004-08-18 12:46:26 +0000599
Antoine Pitroub353c122009-02-11 00:39:14 +0000600 # We override IOBase.__iter__ so that it doesn't check for closed-ness
601
602 def __iter__(self):
603 return self
604
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000605 # For compatibility with old-style urllib responses.
606
607 def info(self):
608 return self.headers
609
610 def geturl(self):
611 return self.url
612
613 def getcode(self):
614 return self.status
Greg Stein5e0fa402000-06-26 08:28:01 +0000615
616class HTTPConnection:
617
Greg Steindd6eefb2000-07-18 09:09:48 +0000618 _http_vsn = 11
619 _http_vsn_str = 'HTTP/1.1'
Greg Stein5e0fa402000-06-26 08:28:01 +0000620
Greg Steindd6eefb2000-07-18 09:09:48 +0000621 response_class = HTTPResponse
622 default_port = HTTP_PORT
623 auto_open = 1
Jeremy Hylton30f86742000-09-18 22:50:38 +0000624 debuglevel = 0
Greg Stein5e0fa402000-06-26 08:28:01 +0000625
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000626 def __init__(self, host, port=None, strict=_strict_sentinel,
Gregory P. Smithb4066372010-01-03 03:28:29 +0000627 timeout=socket._GLOBAL_DEFAULT_TIMEOUT, source_address=None):
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000628 if strict is not _strict_sentinel:
629 warnings.warn("the 'strict' argument isn't supported anymore; "
630 "http.client now always assumes HTTP/1.x compliant servers.",
631 DeprecationWarning, 2)
Guido van Rossumd8faa362007-04-27 19:54:29 +0000632 self.timeout = timeout
Gregory P. Smithb4066372010-01-03 03:28:29 +0000633 self.source_address = source_address
Greg Steindd6eefb2000-07-18 09:09:48 +0000634 self.sock = None
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000635 self._buffer = []
Greg Steindd6eefb2000-07-18 09:09:48 +0000636 self.__response = None
637 self.__state = _CS_IDLE
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000638 self._method = None
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000639 self._tunnel_host = None
640 self._tunnel_port = None
Tim Petersc411dba2002-07-16 21:35:23 +0000641
Greg Steindd6eefb2000-07-18 09:09:48 +0000642 self._set_hostport(host, port)
Greg Stein5e0fa402000-06-26 08:28:01 +0000643
Senthil Kumaran47fff872009-12-20 07:10:31 +0000644 def set_tunnel(self, host, port=None, headers=None):
645 """ Sets up the host and the port for the HTTP CONNECT Tunnelling.
646
647 The headers argument should be a mapping of extra HTTP headers
648 to send with the CONNECT request.
649 """
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000650 self._tunnel_host = host
651 self._tunnel_port = port
Senthil Kumaran47fff872009-12-20 07:10:31 +0000652 if headers:
653 self._tunnel_headers = headers
654 else:
655 self._tunnel_headers.clear()
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000656
Greg Steindd6eefb2000-07-18 09:09:48 +0000657 def _set_hostport(self, host, port):
658 if port is None:
Skip Montanaro10e6e0e2004-09-14 16:32:02 +0000659 i = host.rfind(':')
Skip Montanarocae14d22004-09-14 17:55:21 +0000660 j = host.rfind(']') # ipv6 addresses have [...]
661 if i > j:
Skip Montanaro9d389972002-03-24 16:53:50 +0000662 try:
663 port = int(host[i+1:])
664 except ValueError:
Jeremy Hyltonfbd79942002-07-02 20:19:08 +0000665 raise InvalidURL("nonnumeric port: '%s'" % host[i+1:])
Greg Steindd6eefb2000-07-18 09:09:48 +0000666 host = host[:i]
667 else:
668 port = self.default_port
Raymond Hettinger4d037912004-10-14 15:23:38 +0000669 if host and host[0] == '[' and host[-1] == ']':
Brett Cannon0a1af4a2004-09-15 23:26:23 +0000670 host = host[1:-1]
Greg Steindd6eefb2000-07-18 09:09:48 +0000671 self.host = host
672 self.port = port
Greg Stein5e0fa402000-06-26 08:28:01 +0000673
Jeremy Hylton30f86742000-09-18 22:50:38 +0000674 def set_debuglevel(self, level):
675 self.debuglevel = level
676
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000677 def _tunnel(self):
678 self._set_hostport(self._tunnel_host, self._tunnel_port)
Senthil Kumaran47fff872009-12-20 07:10:31 +0000679 connect_str = "CONNECT %s:%d HTTP/1.0\r\n" %(self.host, self.port)
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000680 connect_bytes = connect_str.encode("ascii")
681 self.send(connect_bytes)
Senthil Kumaran47fff872009-12-20 07:10:31 +0000682 for header, value in self._tunnel_headers.iteritems():
683 header_str = "%s: %s\r\n" % (header, value)
684 header_bytes = header_str.encode("ascii")
685 self.send(header_bytes)
686
Antoine Pitrou988dbd72010-12-17 17:35:56 +0000687 response = self.response_class(self.sock, method = self._method)
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000688 (version, code, message) = response._read_status()
Senthil Kumaran47fff872009-12-20 07:10:31 +0000689
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000690 if code != 200:
691 self.close()
692 raise socket.error("Tunnel connection failed: %d %s" % (code,
693 message.strip()))
694 while True:
695 line = response.fp.readline()
696 if line == b'\r\n':
697 break
698
Greg Steindd6eefb2000-07-18 09:09:48 +0000699 def connect(self):
700 """Connect to the host and port specified in __init__."""
Guido van Rossumd8faa362007-04-27 19:54:29 +0000701 self.sock = socket.create_connection((self.host,self.port),
Gregory P. Smithb4066372010-01-03 03:28:29 +0000702 self.timeout, self.source_address)
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +0000703 if self._tunnel_host:
704 self._tunnel()
Greg Stein5e0fa402000-06-26 08:28:01 +0000705
Greg Steindd6eefb2000-07-18 09:09:48 +0000706 def close(self):
707 """Close the connection to the HTTP server."""
708 if self.sock:
Tim Peters07e99cb2001-01-14 23:47:14 +0000709 self.sock.close() # close it manually... there may be other refs
Greg Steindd6eefb2000-07-18 09:09:48 +0000710 self.sock = None
711 if self.__response:
712 self.__response.close()
713 self.__response = None
714 self.__state = _CS_IDLE
Greg Stein5e0fa402000-06-26 08:28:01 +0000715
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000716 def send(self, data):
717 """Send `data' to the server."""
Greg Steindd6eefb2000-07-18 09:09:48 +0000718 if self.sock is None:
719 if self.auto_open:
720 self.connect()
721 else:
722 raise NotConnected()
Greg Stein5e0fa402000-06-26 08:28:01 +0000723
Jeremy Hylton30f86742000-09-18 22:50:38 +0000724 if self.debuglevel > 0:
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000725 print("send:", repr(data))
Jeremy Hylton636950f2009-03-28 04:34:21 +0000726 blocksize = 8192
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000727 if hasattr(data, "read") :
Jeremy Hylton636950f2009-03-28 04:34:21 +0000728 if self.debuglevel > 0:
729 print("sendIng a read()able")
730 encode = False
Antoine Pitrouead1d622009-09-29 18:44:53 +0000731 try:
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000732 mode = data.mode
Antoine Pitrouead1d622009-09-29 18:44:53 +0000733 except AttributeError:
734 # io.BytesIO and other file-like objects don't have a `mode`
735 # attribute.
736 pass
737 else:
738 if "b" not in mode:
739 encode = True
740 if self.debuglevel > 0:
741 print("encoding file using iso-8859-1")
Jeremy Hylton636950f2009-03-28 04:34:21 +0000742 while 1:
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000743 datablock = data.read(blocksize)
744 if not datablock:
Jeremy Hylton636950f2009-03-28 04:34:21 +0000745 break
746 if encode:
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000747 datablock = datablock.encode("iso-8859-1")
748 self.sock.sendall(datablock)
Jeremy Hylton636950f2009-03-28 04:34:21 +0000749 else:
Senthil Kumaran5a2da3b2010-10-02 10:33:13 +0000750 self.sock.sendall(data)
Greg Stein5e0fa402000-06-26 08:28:01 +0000751
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000752 def _output(self, s):
753 """Add a line of output to the current request buffer.
Tim Peters469cdad2002-08-08 20:19:19 +0000754
Jeremy Hyltone3252ec2002-07-16 21:41:43 +0000755 Assumes that the line does *not* end with \\r\\n.
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000756 """
757 self._buffer.append(s)
758
Benjamin Peterson1742e402008-11-30 22:15:29 +0000759 def _send_output(self, message_body=None):
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000760 """Send the currently buffered request and clear the buffer.
761
Jeremy Hyltone3252ec2002-07-16 21:41:43 +0000762 Appends an extra \\r\\n to the buffer.
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000763 A message_body may be specified, to be appended to the request.
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000764 """
Martin v. Löwisdd5a8602007-06-30 09:22:09 +0000765 self._buffer.extend((b"", b""))
766 msg = b"\r\n".join(self._buffer)
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000767 del self._buffer[:]
Benjamin Peterson1742e402008-11-30 22:15:29 +0000768 # If msg and message_body are sent in a single send() call,
769 # it will avoid performance problems caused by the interaction
770 # between delayed ack and the Nagle algorithim.
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000771 if isinstance(message_body, bytes):
Benjamin Peterson1742e402008-11-30 22:15:29 +0000772 msg += message_body
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000773 message_body = None
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000774 self.send(msg)
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000775 if message_body is not None:
Jeremy Hylton236654b2009-03-27 20:24:34 +0000776 # message_body was not a string (i.e. it is a file), and
777 # we must run the risk of Nagle.
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000778 self.send(message_body)
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000779
Martin v. Löwisaf7dc8d2003-11-19 19:51:55 +0000780 def putrequest(self, method, url, skip_host=0, skip_accept_encoding=0):
Greg Steindd6eefb2000-07-18 09:09:48 +0000781 """Send a request to the server.
Greg Stein5e0fa402000-06-26 08:28:01 +0000782
Greg Steindd6eefb2000-07-18 09:09:48 +0000783 `method' specifies an HTTP request method, e.g. 'GET'.
784 `url' specifies the object being requested, e.g. '/index.html'.
Martin v. Löwisaf7dc8d2003-11-19 19:51:55 +0000785 `skip_host' if True does not add automatically a 'Host:' header
786 `skip_accept_encoding' if True does not add automatically an
787 'Accept-Encoding:' header
Greg Steindd6eefb2000-07-18 09:09:48 +0000788 """
Greg Stein5e0fa402000-06-26 08:28:01 +0000789
Greg Stein616a58d2003-06-24 06:35:19 +0000790 # if a prior response has been completed, then forget about it.
Greg Steindd6eefb2000-07-18 09:09:48 +0000791 if self.__response and self.__response.isclosed():
792 self.__response = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000793
Tim Peters58eb11c2004-01-18 20:29:55 +0000794
Greg Steindd6eefb2000-07-18 09:09:48 +0000795 # in certain cases, we cannot issue another request on this connection.
796 # this occurs when:
797 # 1) we are in the process of sending a request. (_CS_REQ_STARTED)
798 # 2) a response to a previous request has signalled that it is going
799 # to close the connection upon completion.
800 # 3) the headers for the previous response have not been read, thus
801 # we cannot determine whether point (2) is true. (_CS_REQ_SENT)
802 #
803 # if there is no prior response, then we can request at will.
804 #
805 # if point (2) is true, then we will have passed the socket to the
806 # response (effectively meaning, "there is no prior response"), and
807 # will open a new one when a new request is made.
808 #
809 # Note: if a prior response exists, then we *can* start a new request.
810 # We are not allowed to begin fetching the response to this new
811 # request, however, until that prior response is complete.
812 #
813 if self.__state == _CS_IDLE:
814 self.__state = _CS_REQ_STARTED
815 else:
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000816 raise CannotSendRequest(self.__state)
Greg Stein5e0fa402000-06-26 08:28:01 +0000817
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000818 # Save the method we use, we need it later in the response phase
819 self._method = method
Greg Steindd6eefb2000-07-18 09:09:48 +0000820 if not url:
821 url = '/'
Martin v. Löwisdd5a8602007-06-30 09:22:09 +0000822 request = '%s %s %s' % (method, url, self._http_vsn_str)
Greg Stein5e0fa402000-06-26 08:28:01 +0000823
Martin v. Löwisdd5a8602007-06-30 09:22:09 +0000824 # Non-ASCII characters should have been eliminated earlier
825 self._output(request.encode('ascii'))
Greg Stein5e0fa402000-06-26 08:28:01 +0000826
Greg Steindd6eefb2000-07-18 09:09:48 +0000827 if self._http_vsn == 11:
828 # Issue some standard headers for better HTTP/1.1 compliance
Greg Stein5e0fa402000-06-26 08:28:01 +0000829
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000830 if not skip_host:
831 # this header is issued *only* for HTTP/1.1
832 # connections. more specifically, this means it is
833 # only issued when the client uses the new
834 # HTTPConnection() class. backwards-compat clients
835 # will be using HTTP/1.0 and those clients may be
836 # issuing this header themselves. we should NOT issue
837 # it twice; some web servers (such as Apache) barf
838 # when they see two Host: headers
Guido van Rossumf6922aa2001-01-14 21:03:01 +0000839
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000840 # If we need a non-standard port,include it in the
841 # header. If the request is going through a proxy,
842 # but the host of the actual URL, not the host of the
843 # proxy.
Jeremy Hylton8acf1e02002-03-08 19:35:51 +0000844
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000845 netloc = ''
846 if url.startswith('http'):
847 nil, netloc, nil, nil, nil = urlsplit(url)
848
849 if netloc:
Thomas Wouters477c8d52006-05-27 19:21:47 +0000850 try:
851 netloc_enc = netloc.encode("ascii")
852 except UnicodeEncodeError:
853 netloc_enc = netloc.encode("idna")
854 self.putheader('Host', netloc_enc)
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000855 else:
Thomas Wouters477c8d52006-05-27 19:21:47 +0000856 try:
857 host_enc = self.host.encode("ascii")
858 except UnicodeEncodeError:
859 host_enc = self.host.encode("idna")
Senthil Kumaran74ebd9e2010-11-13 12:27:49 +0000860
861 # As per RFC 273, IPv6 address should be wrapped with []
862 # when used as Host header
863
864 if self.host.find(':') >= 0:
865 host_enc = b'[' + host_enc + b']'
866
Georg Brandl86b2fb92008-07-16 03:43:04 +0000867 if self.port == self.default_port:
Thomas Wouters477c8d52006-05-27 19:21:47 +0000868 self.putheader('Host', host_enc)
869 else:
Guido van Rossum98297ee2007-11-06 21:34:58 +0000870 host_enc = host_enc.decode("ascii")
Thomas Wouters477c8d52006-05-27 19:21:47 +0000871 self.putheader('Host', "%s:%s" % (host_enc, self.port))
Greg Stein5e0fa402000-06-26 08:28:01 +0000872
Greg Steindd6eefb2000-07-18 09:09:48 +0000873 # note: we are assuming that clients will not attempt to set these
874 # headers since *this* library must deal with the
875 # consequences. this also means that when the supporting
876 # libraries are updated to recognize other forms, then this
877 # code should be changed (removed or updated).
Greg Stein5e0fa402000-06-26 08:28:01 +0000878
Greg Steindd6eefb2000-07-18 09:09:48 +0000879 # we only want a Content-Encoding of "identity" since we don't
880 # support encodings such as x-gzip or x-deflate.
Martin v. Löwisaf7dc8d2003-11-19 19:51:55 +0000881 if not skip_accept_encoding:
882 self.putheader('Accept-Encoding', 'identity')
Greg Stein5e0fa402000-06-26 08:28:01 +0000883
Greg Steindd6eefb2000-07-18 09:09:48 +0000884 # we can accept "chunked" Transfer-Encodings, but no others
885 # NOTE: no TE header implies *only* "chunked"
886 #self.putheader('TE', 'chunked')
Greg Stein5e0fa402000-06-26 08:28:01 +0000887
Greg Steindd6eefb2000-07-18 09:09:48 +0000888 # if TE is supplied in the header, then it must appear in a
889 # Connection header.
890 #self.putheader('Connection', 'TE')
Greg Stein5e0fa402000-06-26 08:28:01 +0000891
Greg Steindd6eefb2000-07-18 09:09:48 +0000892 else:
893 # For HTTP/1.0, the server will assume "not chunked"
894 pass
Greg Stein5e0fa402000-06-26 08:28:01 +0000895
Benjamin Petersonf608c612008-11-16 18:33:53 +0000896 def putheader(self, header, *values):
Greg Steindd6eefb2000-07-18 09:09:48 +0000897 """Send a request header line to the server.
Greg Stein5e0fa402000-06-26 08:28:01 +0000898
Greg Steindd6eefb2000-07-18 09:09:48 +0000899 For example: h.putheader('Accept', 'text/html')
900 """
901 if self.__state != _CS_REQ_STARTED:
902 raise CannotSendHeader()
Greg Stein5e0fa402000-06-26 08:28:01 +0000903
Guido van Rossum98297ee2007-11-06 21:34:58 +0000904 if hasattr(header, 'encode'):
905 header = header.encode('ascii')
Benjamin Petersonf608c612008-11-16 18:33:53 +0000906 values = list(values)
907 for i, one_value in enumerate(values):
908 if hasattr(one_value, 'encode'):
909 values[i] = one_value.encode('ascii')
Senthil Kumaran58d5dbf2010-10-03 18:22:42 +0000910 elif isinstance(one_value, int):
911 values[i] = str(one_value).encode('ascii')
Benjamin Petersonf608c612008-11-16 18:33:53 +0000912 value = b'\r\n\t'.join(values)
Guido van Rossum98297ee2007-11-06 21:34:58 +0000913 header = header + b': ' + value
914 self._output(header)
Greg Stein5e0fa402000-06-26 08:28:01 +0000915
Benjamin Peterson1742e402008-11-30 22:15:29 +0000916 def endheaders(self, message_body=None):
917 """Indicate that the last header line has been sent to the server.
Greg Stein5e0fa402000-06-26 08:28:01 +0000918
Benjamin Peterson1742e402008-11-30 22:15:29 +0000919 This method sends the request to the server. The optional
920 message_body argument can be used to pass message body
921 associated with the request. The message body will be sent in
922 the same packet as the message headers if possible. The
923 message_body should be a string.
924 """
Greg Steindd6eefb2000-07-18 09:09:48 +0000925 if self.__state == _CS_REQ_STARTED:
926 self.__state = _CS_REQ_SENT
927 else:
928 raise CannotSendHeader()
Benjamin Peterson1742e402008-11-30 22:15:29 +0000929 self._send_output(message_body)
Greg Stein5e0fa402000-06-26 08:28:01 +0000930
Greg Steindd6eefb2000-07-18 09:09:48 +0000931 def request(self, method, url, body=None, headers={}):
932 """Send a complete request to the server."""
Jeremy Hylton636950f2009-03-28 04:34:21 +0000933 self._send_request(method, url, body, headers)
Greg Stein5e0fa402000-06-26 08:28:01 +0000934
Benjamin Peterson1742e402008-11-30 22:15:29 +0000935 def _set_content_length(self, body):
936 # Set the content-length based on the body.
937 thelen = None
938 try:
939 thelen = str(len(body))
940 except TypeError as te:
941 # If this is a file-like object, try to
942 # fstat its file descriptor
Benjamin Peterson1742e402008-11-30 22:15:29 +0000943 try:
944 thelen = str(os.fstat(body.fileno()).st_size)
945 except (AttributeError, OSError):
946 # Don't send a length if this failed
947 if self.debuglevel > 0: print("Cannot stat!!")
948
949 if thelen is not None:
950 self.putheader('Content-Length', thelen)
951
Greg Steindd6eefb2000-07-18 09:09:48 +0000952 def _send_request(self, method, url, body, headers):
Jeremy Hylton636950f2009-03-28 04:34:21 +0000953 # Honor explicitly requested Host: and Accept-Encoding: headers.
Jeremy Hylton2c178252004-08-07 16:28:14 +0000954 header_names = dict.fromkeys([k.lower() for k in headers])
955 skips = {}
956 if 'host' in header_names:
957 skips['skip_host'] = 1
958 if 'accept-encoding' in header_names:
959 skips['skip_accept_encoding'] = 1
Greg Stein5e0fa402000-06-26 08:28:01 +0000960
Jeremy Hylton2c178252004-08-07 16:28:14 +0000961 self.putrequest(method, url, **skips)
962
963 if body and ('content-length' not in header_names):
Benjamin Peterson1742e402008-11-30 22:15:29 +0000964 self._set_content_length(body)
Guido van Rossumcc2b0162007-02-11 06:12:03 +0000965 for hdr, value in headers.items():
Greg Steindd6eefb2000-07-18 09:09:48 +0000966 self.putheader(hdr, value)
Jeremy Hyltonef9f48e2009-03-26 22:04:05 +0000967 if isinstance(body, str):
Jeremy Hylton236654b2009-03-27 20:24:34 +0000968 # RFC 2616 Section 3.7.1 says that text default has a
969 # default charset of iso-8859-1.
970 body = body.encode('iso-8859-1')
Jeremy Hyltonef9f48e2009-03-26 22:04:05 +0000971 self.endheaders(body)
Greg Stein5e0fa402000-06-26 08:28:01 +0000972
Greg Steindd6eefb2000-07-18 09:09:48 +0000973 def getresponse(self):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000974 """Get the response from the server.
975
976 If the HTTPConnection is in the correct state, returns an
977 instance of HTTPResponse or of whatever object is returned by
978 class the response_class variable.
979
980 If a request has not been sent or if a previous response has
981 not be handled, ResponseNotReady is raised. If the HTTP
982 response indicates that the connection should be closed, then
983 it will be closed before the response is returned. When the
984 connection is closed, the underlying socket is closed.
985 """
Greg Stein5e0fa402000-06-26 08:28:01 +0000986
Greg Stein616a58d2003-06-24 06:35:19 +0000987 # if a prior response has been completed, then forget about it.
Greg Steindd6eefb2000-07-18 09:09:48 +0000988 if self.__response and self.__response.isclosed():
989 self.__response = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000990
Greg Steindd6eefb2000-07-18 09:09:48 +0000991 # if a prior response exists, then it must be completed (otherwise, we
992 # cannot read this response's header to determine the connection-close
993 # behavior)
994 #
995 # note: if a prior response existed, but was connection-close, then the
996 # socket and response were made independent of this HTTPConnection
997 # object since a new request requires that we open a whole new
998 # connection
999 #
1000 # this means the prior response had one of two states:
1001 # 1) will_close: this connection was reset and the prior socket and
1002 # response operate independently
1003 # 2) persistent: the response was retained and we await its
1004 # isclosed() status to become true.
1005 #
1006 if self.__state != _CS_REQ_SENT or self.__response:
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +00001007 raise ResponseNotReady(self.__state)
Greg Stein5e0fa402000-06-26 08:28:01 +00001008
Jeremy Hylton30f86742000-09-18 22:50:38 +00001009 if self.debuglevel > 0:
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001010 response = self.response_class(self.sock, self.debuglevel,
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +00001011 method=self._method)
Jeremy Hylton30f86742000-09-18 22:50:38 +00001012 else:
Antoine Pitrou988dbd72010-12-17 17:35:56 +00001013 response = self.response_class(self.sock, method=self._method)
Greg Stein5e0fa402000-06-26 08:28:01 +00001014
Jeremy Hylton39c03802002-07-12 14:04:09 +00001015 response.begin()
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +00001016 assert response.will_close != _UNKNOWN
Greg Steindd6eefb2000-07-18 09:09:48 +00001017 self.__state = _CS_IDLE
Greg Stein5e0fa402000-06-26 08:28:01 +00001018
Greg Steindd6eefb2000-07-18 09:09:48 +00001019 if response.will_close:
Guido van Rossumd8faa362007-04-27 19:54:29 +00001020 # this effectively passes the connection to the response
1021 self.close()
Greg Steindd6eefb2000-07-18 09:09:48 +00001022 else:
1023 # remember this, so we can tell when it is complete
1024 self.__response = response
Greg Stein5e0fa402000-06-26 08:28:01 +00001025
Greg Steindd6eefb2000-07-18 09:09:48 +00001026 return response
Greg Stein5e0fa402000-06-26 08:28:01 +00001027
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001028try:
1029 import ssl
1030except ImportError:
1031 pass
1032else:
1033 class HTTPSConnection(HTTPConnection):
1034 "This class allows communication via SSL."
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001035
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001036 default_port = HTTPS_PORT
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001037
Antoine Pitrou803e6d62010-10-13 10:36:15 +00001038 # XXX Should key_file and cert_file be deprecated in favour of context?
1039
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001040 def __init__(self, host, port=None, key_file=None, cert_file=None,
Antoine Pitrou988dbd72010-12-17 17:35:56 +00001041 strict=_strict_sentinel, timeout=socket._GLOBAL_DEFAULT_TIMEOUT,
Antoine Pitrou803e6d62010-10-13 10:36:15 +00001042 source_address=None, *, context=None, check_hostname=None):
Gregory P. Smithb4066372010-01-03 03:28:29 +00001043 super(HTTPSConnection, self).__init__(host, port, strict, timeout,
1044 source_address)
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001045 self.key_file = key_file
1046 self.cert_file = cert_file
Antoine Pitrou803e6d62010-10-13 10:36:15 +00001047 if context is None:
1048 # Some reasonable defaults
1049 context = ssl.SSLContext(ssl.PROTOCOL_SSLv23)
1050 context.options |= ssl.OP_NO_SSLv2
1051 will_verify = context.verify_mode != ssl.CERT_NONE
1052 if check_hostname is None:
1053 check_hostname = will_verify
1054 elif check_hostname and not will_verify:
1055 raise ValueError("check_hostname needs a SSL context with "
1056 "either CERT_OPTIONAL or CERT_REQUIRED")
1057 if key_file or cert_file:
Antoine Pitrou80603992010-10-13 10:45:16 +00001058 context.load_cert_chain(cert_file, key_file)
Antoine Pitrou803e6d62010-10-13 10:36:15 +00001059 self._context = context
1060 self._check_hostname = check_hostname
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001061
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001062 def connect(self):
1063 "Connect to a host on a given (SSL) port."
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001064
Jeremy Hylton636950f2009-03-28 04:34:21 +00001065 sock = socket.create_connection((self.host, self.port),
Gregory P. Smithb4066372010-01-03 03:28:29 +00001066 self.timeout, self.source_address)
Senthil Kumaran97f0c6b2009-07-25 04:24:38 +00001067
1068 if self._tunnel_host:
1069 self.sock = sock
1070 self._tunnel()
1071
Antoine Pitroud5323212010-10-22 18:19:07 +00001072 server_hostname = self.host if ssl.HAS_SNI else None
1073 self.sock = self._context.wrap_socket(sock,
1074 server_hostname=server_hostname)
Antoine Pitrou803e6d62010-10-13 10:36:15 +00001075 try:
1076 if self._check_hostname:
1077 ssl.match_hostname(self.sock.getpeercert(), self.host)
1078 except Exception:
1079 self.sock.shutdown(socket.SHUT_RDWR)
1080 self.sock.close()
1081 raise
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001082
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001083 __all__.append("HTTPSConnection")
Greg Stein5e0fa402000-06-26 08:28:01 +00001084
Greg Stein5e0fa402000-06-26 08:28:01 +00001085class HTTPException(Exception):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001086 # Subclasses that define an __init__ must call Exception.__init__
1087 # or define self.args. Otherwise, str() will fail.
Greg Steindd6eefb2000-07-18 09:09:48 +00001088 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001089
1090class NotConnected(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001091 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001092
Skip Montanaro9d389972002-03-24 16:53:50 +00001093class InvalidURL(HTTPException):
1094 pass
1095
Greg Stein5e0fa402000-06-26 08:28:01 +00001096class UnknownProtocol(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001097 def __init__(self, version):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001098 self.args = version,
Greg Steindd6eefb2000-07-18 09:09:48 +00001099 self.version = version
Greg Stein5e0fa402000-06-26 08:28:01 +00001100
1101class UnknownTransferEncoding(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001102 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001103
Greg Stein5e0fa402000-06-26 08:28:01 +00001104class UnimplementedFileMode(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001105 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001106
1107class IncompleteRead(HTTPException):
Benjamin Peterson6accb982009-03-02 22:50:25 +00001108 def __init__(self, partial, expected=None):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001109 self.args = partial,
Greg Steindd6eefb2000-07-18 09:09:48 +00001110 self.partial = partial
Benjamin Peterson6accb982009-03-02 22:50:25 +00001111 self.expected = expected
1112 def __repr__(self):
1113 if self.expected is not None:
1114 e = ', %i more expected' % self.expected
1115 else:
1116 e = ''
1117 return 'IncompleteRead(%i bytes read%s)' % (len(self.partial), e)
1118 def __str__(self):
1119 return repr(self)
Greg Stein5e0fa402000-06-26 08:28:01 +00001120
1121class ImproperConnectionState(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001122 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001123
1124class CannotSendRequest(ImproperConnectionState):
Greg Steindd6eefb2000-07-18 09:09:48 +00001125 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001126
1127class CannotSendHeader(ImproperConnectionState):
Greg Steindd6eefb2000-07-18 09:09:48 +00001128 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001129
1130class ResponseNotReady(ImproperConnectionState):
Greg Steindd6eefb2000-07-18 09:09:48 +00001131 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001132
1133class BadStatusLine(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001134 def __init__(self, line):
Benjamin Peterson11dbfd42010-03-21 22:50:04 +00001135 if not line:
1136 line = repr(line)
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001137 self.args = line,
Greg Steindd6eefb2000-07-18 09:09:48 +00001138 self.line = line
Greg Stein5e0fa402000-06-26 08:28:01 +00001139
1140# for backwards compatibility
1141error = HTTPException