blob: 4a65125712c760ced18f2a9771b75ecc914ec95b [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
Antoine Pitrouff1bbba2010-12-18 18:04:38 +0000206# maximal line length when calling readline().
207_MAXLINE = 65536
208
Barry Warsaw820c1202008-06-12 04:06:45 +0000209class HTTPMessage(email.message.Message):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000210 # XXX The only usage of this method is in
211 # http.server.CGIHTTPRequestHandler. Maybe move the code there so
212 # that it doesn't need to be part of the public API. The API has
213 # never been defined so this could cause backwards compatibility
214 # issues.
215
Barry Warsaw820c1202008-06-12 04:06:45 +0000216 def getallmatchingheaders(self, name):
217 """Find all header lines matching a given header name.
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000218
Barry Warsaw820c1202008-06-12 04:06:45 +0000219 Look through the list of headers and find all lines matching a given
220 header name (and their continuation lines). A list of the lines is
221 returned, without interpretation. If the header does not occur, an
222 empty list is returned. If the header occurs multiple times, all
223 occurrences are returned. Case is not important in the header name.
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000224
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000225 """
Barry Warsaw820c1202008-06-12 04:06:45 +0000226 name = name.lower() + ':'
227 n = len(name)
228 lst = []
229 hit = 0
230 for line in self.keys():
231 if line[:n].lower() == name:
232 hit = 1
233 elif not line[:1].isspace():
234 hit = 0
235 if hit:
236 lst.append(line)
237 return lst
Jeremy Hylton6d0a4c72002-07-07 16:51:37 +0000238
Jeremy Hylton98eb6c22009-03-27 18:31:36 +0000239def parse_headers(fp, _class=HTTPMessage):
Barry Warsaw820c1202008-06-12 04:06:45 +0000240 """Parses only RFC2822 headers from a file pointer.
241
242 email Parser wants to see strings rather than bytes.
243 But a TextIOWrapper around self.rfile would buffer too many bytes
244 from the stream, bytes which we later need to read as bytes.
245 So we read the correct bytes here, as bytes, for email Parser
246 to parse.
247
248 """
Barry Warsaw820c1202008-06-12 04:06:45 +0000249 headers = []
250 while True:
Antoine Pitrouff1bbba2010-12-18 18:04:38 +0000251 line = fp.readline(_MAXLINE + 1)
252 if len(line) > _MAXLINE:
253 raise LineTooLong("header line")
Barry Warsaw820c1202008-06-12 04:06:45 +0000254 headers.append(line)
255 if line in (b'\r\n', b'\n', b''):
256 break
257 hstring = b''.join(headers).decode('iso-8859-1')
Jeremy Hylton98eb6c22009-03-27 18:31:36 +0000258 return email.parser.Parser(_class=_class).parsestr(hstring)
Greg Stein5e0fa402000-06-26 08:28:01 +0000259
Antoine Pitroub353c122009-02-11 00:39:14 +0000260class HTTPResponse(io.RawIOBase):
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000261
262 # strict: If true, raise BadStatusLine if the status line can't be
263 # parsed as a valid HTTP/1.0 or 1.1 status line. By default it is
Skip Montanaro186bec22002-07-25 16:10:38 +0000264 # false because it prevents clients from talking to HTTP/0.9
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000265 # servers. Note that a response with a sufficiently corrupted
266 # status line will look like an HTTP/0.9 response.
267
268 # See RFC 2616 sec 19.6 and RFC 1945 sec 6 for details.
269
Jeremy Hylton811fc142007-08-03 13:30:02 +0000270 # The bytes from the socket object are iso-8859-1 strings.
271 # See RFC 2616 sec 2.2 which notes an exception for MIME-encoded
272 # text following RFC 2047. The basic status line parsing only
273 # accepts iso-8859-1.
274
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000275 def __init__(self, sock, debuglevel=0, strict=0, method=None, url=None):
276 # If the response includes a content-length header, we need to
277 # make sure that the client doesn't read more than the
Jeremy Hylton39b198d2007-08-04 19:22:00 +0000278 # specified number of bytes. If it does, it will block until
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000279 # the server times out and closes the connection. This will
280 # happen if a self.fp.read() is done (without a size) whether
281 # self.fp is buffered or not. So, no self.fp.read() by
282 # clients unless they know what they are doing.
Benjamin Petersonf72d9fb2009-02-08 00:29:20 +0000283 self.fp = sock.makefile("rb")
Jeremy Hylton30f86742000-09-18 22:50:38 +0000284 self.debuglevel = debuglevel
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000285 self.strict = strict
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000286 self._method = method
Greg Stein5e0fa402000-06-26 08:28:01 +0000287
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000288 # The HTTPResponse object is returned via urllib. The clients
289 # of http and urllib expect different attributes for the
290 # headers. headers is used here and supports urllib. msg is
291 # provided as a backwards compatibility layer for http
292 # clients.
293
294 self.headers = self.msg = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000295
Greg Steindd6eefb2000-07-18 09:09:48 +0000296 # from the Status-Line of the response
Tim Peters07e99cb2001-01-14 23:47:14 +0000297 self.version = _UNKNOWN # HTTP-Version
298 self.status = _UNKNOWN # Status-Code
299 self.reason = _UNKNOWN # Reason-Phrase
Greg Stein5e0fa402000-06-26 08:28:01 +0000300
Tim Peters07e99cb2001-01-14 23:47:14 +0000301 self.chunked = _UNKNOWN # is "chunked" being used?
302 self.chunk_left = _UNKNOWN # bytes left to read in current chunk
303 self.length = _UNKNOWN # number of bytes left in response
304 self.will_close = _UNKNOWN # conn will close at end of response
Greg Stein5e0fa402000-06-26 08:28:01 +0000305
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000306 def _read_status(self):
Jeremy Hylton04319c72007-08-04 03:41:19 +0000307 # Initialize with Simple-Response defaults.
Jeremy Hylton811fc142007-08-03 13:30:02 +0000308 line = str(self.fp.readline(), "iso-8859-1")
Jeremy Hylton30f86742000-09-18 22:50:38 +0000309 if self.debuglevel > 0:
Guido van Rossumbe19ed72007-02-09 05:37:30 +0000310 print("reply:", repr(line))
Jeremy Hyltonb6769522003-06-29 17:55:05 +0000311 if not line:
312 # Presumably, the server closed the connection before
313 # sending a valid response.
314 raise BadStatusLine(line)
Greg Steindd6eefb2000-07-18 09:09:48 +0000315 try:
Guido van Rossum34735a62000-12-15 15:09:42 +0000316 [version, status, reason] = line.split(None, 2)
Greg Steindd6eefb2000-07-18 09:09:48 +0000317 except ValueError:
318 try:
Guido van Rossum34735a62000-12-15 15:09:42 +0000319 [version, status] = line.split(None, 1)
Greg Steindd6eefb2000-07-18 09:09:48 +0000320 reason = ""
321 except ValueError:
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000322 # empty version will cause next test to fail and status
323 # will be treated as 0.9 response.
324 version = ""
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000325 if not version.startswith("HTTP/"):
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000326 if self.strict:
327 self.close()
328 raise BadStatusLine(line)
329 else:
Jeremy Hylton04319c72007-08-04 03:41:19 +0000330 # Assume it's a Simple-Response from an 0.9 server.
331 # We have to convert the first line back to raw bytes
332 # because self.fp.readline() needs to return bytes.
Guido van Rossum70d0dda2007-08-29 01:53:26 +0000333 self.fp = LineAndFileWrapper(bytes(line, "ascii"), self.fp)
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000334 return "HTTP/0.9", 200, ""
Greg Stein5e0fa402000-06-26 08:28:01 +0000335
Jeremy Hylton23d40472001-04-13 14:57:08 +0000336 # The status code is a three-digit number
337 try:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000338 status = int(status)
Jeremy Hylton23d40472001-04-13 14:57:08 +0000339 if status < 100 or status > 999:
340 raise BadStatusLine(line)
341 except ValueError:
342 raise BadStatusLine(line)
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000343 return version, status, reason
Greg Stein5e0fa402000-06-26 08:28:01 +0000344
Jeremy Hylton39c03802002-07-12 14:04:09 +0000345 def begin(self):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000346 if self.headers is not None:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000347 # we've already started reading the response
348 return
349
350 # read until we get a non-100 response
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000351 while True:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000352 version, status, reason = self._read_status()
Martin v. Löwis39a31782004-09-18 09:03:49 +0000353 if status != CONTINUE:
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000354 break
355 # skip the header from the 100 response
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000356 while True:
Antoine Pitrouff1bbba2010-12-18 18:04:38 +0000357 skip = self.fp.readline(_MAXLINE + 1)
358 if len(skip) > _MAXLINE:
359 raise LineTooLong("header line")
360 skip = skip.strip()
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000361 if not skip:
362 break
363 if self.debuglevel > 0:
Guido van Rossumbe19ed72007-02-09 05:37:30 +0000364 print("header:", skip)
Tim Petersc411dba2002-07-16 21:35:23 +0000365
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000366 self.code = self.status = status
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +0000367 self.reason = reason.strip()
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000368 if version == "HTTP/1.0":
Greg Steindd6eefb2000-07-18 09:09:48 +0000369 self.version = 10
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000370 elif version.startswith("HTTP/1."):
Tim Peters07e99cb2001-01-14 23:47:14 +0000371 self.version = 11 # use HTTP/1.1 code for HTTP/1.x where x>=1
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000372 elif version == "HTTP/0.9":
Jeremy Hylton110941a2000-10-12 19:58:36 +0000373 self.version = 9
Greg Steindd6eefb2000-07-18 09:09:48 +0000374 else:
375 raise UnknownProtocol(version)
Greg Stein5e0fa402000-06-26 08:28:01 +0000376
Jeremy Hylton110941a2000-10-12 19:58:36 +0000377 if self.version == 9:
Georg Brandl0aade9a2005-06-26 22:06:54 +0000378 self.length = None
Jeremy Hylton236156f2008-12-15 03:00:50 +0000379 self.chunked = False
380 self.will_close = True
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000381 self.headers = self.msg = email.message_from_string('')
Jeremy Hylton110941a2000-10-12 19:58:36 +0000382 return
383
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000384 self.headers = self.msg = parse_headers(self.fp)
Barry Warsaw820c1202008-06-12 04:06:45 +0000385
Jeremy Hylton30f86742000-09-18 22:50:38 +0000386 if self.debuglevel > 0:
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000387 for hdr in self.headers:
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000388 print("header:", hdr, end=" ")
Greg Stein5e0fa402000-06-26 08:28:01 +0000389
Greg Steindd6eefb2000-07-18 09:09:48 +0000390 # are we using the chunked-style of transfer encoding?
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000391 tr_enc = self.headers.get("transfer-encoding")
Jeremy Hyltond229b3a2002-09-03 19:24:24 +0000392 if tr_enc and tr_enc.lower() == "chunked":
Jeremy Hylton236156f2008-12-15 03:00:50 +0000393 self.chunked = True
Greg Steindd6eefb2000-07-18 09:09:48 +0000394 self.chunk_left = None
395 else:
Jeremy Hylton236156f2008-12-15 03:00:50 +0000396 self.chunked = False
Greg Stein5e0fa402000-06-26 08:28:01 +0000397
Greg Steindd6eefb2000-07-18 09:09:48 +0000398 # will the connection close at the end of the response?
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000399 self.will_close = self._check_close()
Greg Stein5e0fa402000-06-26 08:28:01 +0000400
Greg Steindd6eefb2000-07-18 09:09:48 +0000401 # do we have a Content-Length?
402 # NOTE: RFC 2616, S4.4, #3 says we ignore this if tr_enc is "chunked"
Jeremy Hylton0ee5eeb2007-08-04 03:25:17 +0000403 self.length = None
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000404 length = self.headers.get("content-length")
Barry Warsaw820c1202008-06-12 04:06:45 +0000405
406 # are we using the chunked-style of transfer encoding?
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000407 tr_enc = self.headers.get("transfer-encoding")
Greg Steindd6eefb2000-07-18 09:09:48 +0000408 if length and not self.chunked:
Jeremy Hylton30a81812000-09-14 20:34:27 +0000409 try:
410 self.length = int(length)
411 except ValueError:
Christian Heimesa612dc02008-02-24 13:08:18 +0000412 self.length = None
413 else:
414 if self.length < 0: # ignore nonsensical negative lengths
415 self.length = None
416 else:
417 self.length = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000418
Greg Steindd6eefb2000-07-18 09:09:48 +0000419 # does the body have a fixed length? (of zero)
Martin v. Löwis39a31782004-09-18 09:03:49 +0000420 if (status == NO_CONTENT or status == NOT_MODIFIED or
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000421 100 <= status < 200 or # 1xx codes
Jeremy Hyltone5d0e842007-08-03 13:45:24 +0000422 self._method == "HEAD"):
Greg Steindd6eefb2000-07-18 09:09:48 +0000423 self.length = 0
Greg Stein5e0fa402000-06-26 08:28:01 +0000424
Greg Steindd6eefb2000-07-18 09:09:48 +0000425 # if the connection remains open, and we aren't using chunked, and
426 # a content-length was not provided, then assume that the connection
427 # WILL close.
Jeremy Hylton0ee5eeb2007-08-04 03:25:17 +0000428 if (not self.will_close and
429 not self.chunked and
430 self.length is None):
Jeremy Hylton236156f2008-12-15 03:00:50 +0000431 self.will_close = True
Greg Stein5e0fa402000-06-26 08:28:01 +0000432
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000433 def _check_close(self):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000434 conn = self.headers.get("connection")
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000435 if self.version == 11:
436 # An HTTP/1.1 proxy is assumed to stay open unless
437 # explicitly closed.
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000438 conn = self.headers.get("connection")
Raymond Hettingerbac788a2004-05-04 09:21:43 +0000439 if conn and "close" in conn.lower():
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000440 return True
441 return False
442
Jeremy Hylton2c178252004-08-07 16:28:14 +0000443 # Some HTTP/1.0 implementations have support for persistent
444 # connections, using rules different than HTTP/1.1.
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000445
Christian Heimes895627f2007-12-08 17:28:33 +0000446 # For older HTTP, Keep-Alive indicates persistent connection.
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000447 if self.headers.get("keep-alive"):
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000448 return False
Tim Peters77c06fb2002-11-24 02:35:35 +0000449
Jeremy Hylton2c178252004-08-07 16:28:14 +0000450 # At least Akamai returns a "Connection: Keep-Alive" header,
451 # which was supposed to be sent by the client.
452 if conn and "keep-alive" in conn.lower():
453 return False
454
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000455 # Proxy-Connection is a netscape hack.
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000456 pconn = self.headers.get("proxy-connection")
Raymond Hettingerbac788a2004-05-04 09:21:43 +0000457 if pconn and "keep-alive" in pconn.lower():
Jeremy Hylton22b3a492002-11-13 17:27:43 +0000458 return False
459
460 # otherwise, assume it will close
461 return True
462
Greg Steindd6eefb2000-07-18 09:09:48 +0000463 def close(self):
464 if self.fp:
465 self.fp.close()
466 self.fp = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000467
Jeremy Hyltondf5f6b52007-08-08 17:36:33 +0000468 # These implementations are for the benefit of io.BufferedReader.
469
470 # XXX This class should probably be revised to act more like
471 # the "raw stream" that BufferedReader expects.
472
473 @property
474 def closed(self):
475 return self.isclosed()
476
477 def flush(self):
478 self.fp.flush()
479
Jeremy Hyltona7cff022009-04-01 02:35:56 +0000480 def readable(self):
481 return True
482
Jeremy Hyltondf5f6b52007-08-08 17:36:33 +0000483 # End of "raw stream" methods
484
Greg Steindd6eefb2000-07-18 09:09:48 +0000485 def isclosed(self):
486 # NOTE: it is possible that we will not ever call self.close(). This
487 # case occurs when will_close is TRUE, length is None, and we
488 # read up to the last byte, but NOT past it.
489 #
490 # IMPLIES: if will_close is FALSE, then self.close() will ALWAYS be
491 # called, meaning self.isclosed() is meaningful.
492 return self.fp is None
493
494 def read(self, amt=None):
495 if self.fp is None:
Guido van Rossum8ce8a782007-11-01 19:42:39 +0000496 return b""
Greg Steindd6eefb2000-07-18 09:09:48 +0000497
Senthil Kumaran26826382010-04-28 17:44:30 +0000498 if self._method == "HEAD":
Senthil Kumaran41142f52010-06-04 16:43:10 +0000499 self.close()
Senthil Kumaran26826382010-04-28 17:44:30 +0000500 return b""
501
Greg Steindd6eefb2000-07-18 09:09:48 +0000502 if self.chunked:
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000503 return self._read_chunked(amt)
Tim Peters230a60c2002-11-09 05:08:07 +0000504
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000505 if amt is None:
Greg Steindd6eefb2000-07-18 09:09:48 +0000506 # unbounded read
Jeremy Hyltondef9d2a2004-11-07 16:13:49 +0000507 if self.length is None:
Greg Steindd6eefb2000-07-18 09:09:48 +0000508 s = self.fp.read()
509 else:
510 s = self._safe_read(self.length)
Jeremy Hyltondef9d2a2004-11-07 16:13:49 +0000511 self.length = 0
Tim Peters07e99cb2001-01-14 23:47:14 +0000512 self.close() # we read everything
Greg Steindd6eefb2000-07-18 09:09:48 +0000513 return s
514
515 if self.length is not None:
516 if amt > self.length:
517 # clip the read to the "end of response"
518 amt = self.length
Greg Steindd6eefb2000-07-18 09:09:48 +0000519
520 # we do not use _safe_read() here because this may be a .will_close
521 # connection, and the user is reading more bytes than will be provided
522 # (for example, reading in 1k chunks)
523 s = self.fp.read(amt)
Jeremy Hyltondef9d2a2004-11-07 16:13:49 +0000524 if self.length is not None:
525 self.length -= len(s)
Guido van Rossum8ce8a782007-11-01 19:42:39 +0000526 if not self.length:
527 self.close()
Greg Steindd6eefb2000-07-18 09:09:48 +0000528 return s
529
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000530 def _read_chunked(self, amt):
531 assert self.chunked != _UNKNOWN
532 chunk_left = self.chunk_left
Chris Withers5a86acb2009-09-04 17:15:46 +0000533 value = []
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000534 while True:
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000535 if chunk_left is None:
Antoine Pitrouff1bbba2010-12-18 18:04:38 +0000536 line = self.fp.readline(_MAXLINE + 1)
537 if len(line) > _MAXLINE:
538 raise LineTooLong("chunk size")
Georg Brandl95ba4692008-01-26 09:45:58 +0000539 i = line.find(b";")
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000540 if i >= 0:
541 line = line[:i] # strip chunk-extensions
Christian Heimesa612dc02008-02-24 13:08:18 +0000542 try:
543 chunk_left = int(line, 16)
544 except ValueError:
545 # close the connection as protocol synchronisation is
546 # probably lost
547 self.close()
Chris Withers5a86acb2009-09-04 17:15:46 +0000548 raise IncompleteRead(b''.join(value))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000549 if chunk_left == 0:
550 break
551 if amt is None:
Chris Withers5a86acb2009-09-04 17:15:46 +0000552 value.append(self._safe_read(chunk_left))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000553 elif amt < chunk_left:
Chris Withers5a86acb2009-09-04 17:15:46 +0000554 value.append(self._safe_read(amt))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000555 self.chunk_left = chunk_left - amt
Chris Withers5a86acb2009-09-04 17:15:46 +0000556 return b''.join(value)
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000557 elif amt == chunk_left:
Chris Withers5a86acb2009-09-04 17:15:46 +0000558 value.append(self._safe_read(amt))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000559 self._safe_read(2) # toss the CRLF at the end of the chunk
560 self.chunk_left = None
Chris Withers5a86acb2009-09-04 17:15:46 +0000561 return b''.join(value)
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000562 else:
Chris Withers5a86acb2009-09-04 17:15:46 +0000563 value.append(self._safe_read(chunk_left))
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000564 amt -= chunk_left
565
566 # we read the whole chunk, get another
567 self._safe_read(2) # toss the CRLF at the end of the chunk
568 chunk_left = None
569
570 # read and discard trailer up to the CRLF terminator
571 ### note: we shouldn't have any trailers!
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000572 while True:
Antoine Pitrouff1bbba2010-12-18 18:04:38 +0000573 line = self.fp.readline(_MAXLINE + 1)
574 if len(line) > _MAXLINE:
575 raise LineTooLong("trailer line")
Christian Heimes0bd4e112008-02-12 22:59:25 +0000576 if not line:
577 # a vanishingly small number of sites EOF without
578 # sending the trailer
579 break
Guido van Rossum8ce8a782007-11-01 19:42:39 +0000580 if line == b"\r\n":
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000581 break
582
583 # we read everything; close the "file"
Jeremy Hyltond4c472c2002-09-03 20:49:06 +0000584 self.close()
585
Chris Withers5a86acb2009-09-04 17:15:46 +0000586 return b''.join(value)
Tim Peters230a60c2002-11-09 05:08:07 +0000587
Greg Steindd6eefb2000-07-18 09:09:48 +0000588 def _safe_read(self, amt):
589 """Read the number of bytes requested, compensating for partial reads.
590
591 Normally, we have a blocking socket, but a read() can be interrupted
592 by a signal (resulting in a partial read).
593
594 Note that we cannot distinguish between EOF and an interrupt when zero
595 bytes have been read. IncompleteRead() will be raised in this
596 situation.
597
598 This function should be used when <amt> bytes "should" be present for
599 reading. If the bytes are truly not available (due to EOF), then the
600 IncompleteRead exception can be used to detect the problem.
601 """
Georg Brandl80ba8e82005-09-29 20:16:07 +0000602 s = []
Greg Steindd6eefb2000-07-18 09:09:48 +0000603 while amt > 0:
Georg Brandl80ba8e82005-09-29 20:16:07 +0000604 chunk = self.fp.read(min(amt, MAXAMOUNT))
Greg Steindd6eefb2000-07-18 09:09:48 +0000605 if not chunk:
Benjamin Peterson6accb982009-03-02 22:50:25 +0000606 raise IncompleteRead(b''.join(s), amt)
Georg Brandl80ba8e82005-09-29 20:16:07 +0000607 s.append(chunk)
Raymond Hettingerb2e0b922003-02-26 22:45:18 +0000608 amt -= len(chunk)
Guido van Rossuma00f1232007-09-12 19:43:09 +0000609 return b"".join(s)
Greg Steindd6eefb2000-07-18 09:09:48 +0000610
Antoine Pitroub353c122009-02-11 00:39:14 +0000611 def fileno(self):
612 return self.fp.fileno()
613
Greg Steindd6eefb2000-07-18 09:09:48 +0000614 def getheader(self, name, default=None):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000615 if self.headers is None:
Greg Steindd6eefb2000-07-18 09:09:48 +0000616 raise ResponseNotReady()
Senthil Kumaran33578402010-08-02 12:01:21 +0000617 headers = self.headers.get_all(name) or default
618 if isinstance(headers, str) or not hasattr(headers, '__iter__'):
619 return headers
620 else:
621 return ', '.join(headers)
Greg Stein5e0fa402000-06-26 08:28:01 +0000622
Martin v. Löwisdeacce22004-08-18 12:46:26 +0000623 def getheaders(self):
624 """Return list of (header, value) tuples."""
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000625 if self.headers is None:
Martin v. Löwisdeacce22004-08-18 12:46:26 +0000626 raise ResponseNotReady()
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000627 return list(self.headers.items())
Martin v. Löwisdeacce22004-08-18 12:46:26 +0000628
Antoine Pitroub353c122009-02-11 00:39:14 +0000629 # We override IOBase.__iter__ so that it doesn't check for closed-ness
630
631 def __iter__(self):
632 return self
633
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000634 # For compatibility with old-style urllib responses.
635
636 def info(self):
637 return self.headers
638
639 def geturl(self):
640 return self.url
641
642 def getcode(self):
643 return self.status
Greg Stein5e0fa402000-06-26 08:28:01 +0000644
645class HTTPConnection:
646
Greg Steindd6eefb2000-07-18 09:09:48 +0000647 _http_vsn = 11
648 _http_vsn_str = 'HTTP/1.1'
Greg Stein5e0fa402000-06-26 08:28:01 +0000649
Greg Steindd6eefb2000-07-18 09:09:48 +0000650 response_class = HTTPResponse
651 default_port = HTTP_PORT
652 auto_open = 1
Jeremy Hylton30f86742000-09-18 22:50:38 +0000653 debuglevel = 0
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000654 strict = 0
Greg Stein5e0fa402000-06-26 08:28:01 +0000655
Georg Brandlf78e02b2008-06-10 17:40:04 +0000656 def __init__(self, host, port=None, strict=None,
657 timeout=socket._GLOBAL_DEFAULT_TIMEOUT):
Guido van Rossumd8faa362007-04-27 19:54:29 +0000658 self.timeout = timeout
Greg Steindd6eefb2000-07-18 09:09:48 +0000659 self.sock = None
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000660 self._buffer = []
Greg Steindd6eefb2000-07-18 09:09:48 +0000661 self.__response = None
662 self.__state = _CS_IDLE
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000663 self._method = None
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000664 self._tunnel_host = None
665 self._tunnel_port = None
Georg Brandl27743102011-02-25 10:18:11 +0000666 self._tunnel_headers = {}
Tim Petersc411dba2002-07-16 21:35:23 +0000667
Greg Steindd6eefb2000-07-18 09:09:48 +0000668 self._set_hostport(host, port)
Jeremy Hyltond46aa372002-07-06 18:48:07 +0000669 if strict is not None:
670 self.strict = strict
Greg Stein5e0fa402000-06-26 08:28:01 +0000671
Senthil Kumaran4b9fbeb2009-12-20 07:18:22 +0000672 def _set_tunnel(self, host, port=None, headers=None):
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000673 self._tunnel_host = host
674 self._tunnel_port = port
Senthil Kumaran4b9fbeb2009-12-20 07:18:22 +0000675 if headers:
676 self._tunnel_headers = headers
677 else:
678 self._tunnel_headers.clear()
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000679
Greg Steindd6eefb2000-07-18 09:09:48 +0000680 def _set_hostport(self, host, port):
681 if port is None:
Skip Montanaro10e6e0e2004-09-14 16:32:02 +0000682 i = host.rfind(':')
Skip Montanarocae14d22004-09-14 17:55:21 +0000683 j = host.rfind(']') # ipv6 addresses have [...]
684 if i > j:
Skip Montanaro9d389972002-03-24 16:53:50 +0000685 try:
686 port = int(host[i+1:])
687 except ValueError:
Jeremy Hyltonfbd79942002-07-02 20:19:08 +0000688 raise InvalidURL("nonnumeric port: '%s'" % host[i+1:])
Greg Steindd6eefb2000-07-18 09:09:48 +0000689 host = host[:i]
690 else:
691 port = self.default_port
Raymond Hettinger4d037912004-10-14 15:23:38 +0000692 if host and host[0] == '[' and host[-1] == ']':
Brett Cannon0a1af4a2004-09-15 23:26:23 +0000693 host = host[1:-1]
Greg Steindd6eefb2000-07-18 09:09:48 +0000694 self.host = host
695 self.port = port
Greg Stein5e0fa402000-06-26 08:28:01 +0000696
Jeremy Hylton30f86742000-09-18 22:50:38 +0000697 def set_debuglevel(self, level):
698 self.debuglevel = level
699
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000700 def _tunnel(self):
701 self._set_hostport(self._tunnel_host, self._tunnel_port)
Georg Brandl27743102011-02-25 10:18:11 +0000702 connect_str = "CONNECT %s:%d HTTP/1.0\r\n" % (self.host, self.port)
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000703 connect_bytes = connect_str.encode("ascii")
704 self.send(connect_bytes)
Georg Brandl27743102011-02-25 10:18:11 +0000705 for header, value in self._tunnel_headers.items():
Senthil Kumaran4b9fbeb2009-12-20 07:18:22 +0000706 header_str = "%s: %s\r\n" % (header, value)
707 header_bytes = header_str.encode("ascii")
708 self.send(header_bytes)
Georg Brandl27743102011-02-25 10:18:11 +0000709 self.send(b'\r\n')
Senthil Kumaran4b9fbeb2009-12-20 07:18:22 +0000710
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000711 response = self.response_class(self.sock, strict = self.strict,
Senthil Kumaran4b9fbeb2009-12-20 07:18:22 +0000712 method = self._method)
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000713 (version, code, message) = response._read_status()
Senthil Kumaran4b9fbeb2009-12-20 07:18:22 +0000714
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000715 if code != 200:
716 self.close()
717 raise socket.error("Tunnel connection failed: %d %s" % (code,
718 message.strip()))
719 while True:
Antoine Pitrouff1bbba2010-12-18 18:04:38 +0000720 line = response.fp.readline(_MAXLINE + 1)
721 if len(line) > _MAXLINE:
722 raise LineTooLong("header line")
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000723 if line == b'\r\n':
724 break
725
Greg Steindd6eefb2000-07-18 09:09:48 +0000726 def connect(self):
727 """Connect to the host and port specified in __init__."""
Guido van Rossumd8faa362007-04-27 19:54:29 +0000728 self.sock = socket.create_connection((self.host,self.port),
729 self.timeout)
Senthil Kumaran0ac1f832009-07-26 12:39:47 +0000730 if self._tunnel_host:
731 self._tunnel()
Greg Stein5e0fa402000-06-26 08:28:01 +0000732
Greg Steindd6eefb2000-07-18 09:09:48 +0000733 def close(self):
734 """Close the connection to the HTTP server."""
735 if self.sock:
Tim Peters07e99cb2001-01-14 23:47:14 +0000736 self.sock.close() # close it manually... there may be other refs
Greg Steindd6eefb2000-07-18 09:09:48 +0000737 self.sock = None
738 if self.__response:
739 self.__response.close()
740 self.__response = None
741 self.__state = _CS_IDLE
Greg Stein5e0fa402000-06-26 08:28:01 +0000742
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000743 def send(self, data):
744 """Send `data' to the server."""
Greg Steindd6eefb2000-07-18 09:09:48 +0000745 if self.sock is None:
746 if self.auto_open:
747 self.connect()
748 else:
749 raise NotConnected()
Greg Stein5e0fa402000-06-26 08:28:01 +0000750
Jeremy Hyltone3252ec2002-07-16 21:41:43 +0000751 # send the data to the server. if we get a broken pipe, then close
Greg Steindd6eefb2000-07-18 09:09:48 +0000752 # the socket. we want to reconnect when somebody tries to send again.
753 #
754 # NOTE: we DO propagate the error, though, because we cannot simply
755 # ignore the error... the caller will know if they can retry.
Jeremy Hylton30f86742000-09-18 22:50:38 +0000756 if self.debuglevel > 0:
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000757 print("send:", repr(data))
Jeremy Hylton636950f2009-03-28 04:34:21 +0000758 blocksize = 8192
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000759 if hasattr(data, "read") :
Jeremy Hylton636950f2009-03-28 04:34:21 +0000760 if self.debuglevel > 0:
761 print("sendIng a read()able")
762 encode = False
Antoine Pitrou595ad322009-09-29 19:02:24 +0000763 try:
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000764 mode = data.mode
Antoine Pitrou595ad322009-09-29 19:02:24 +0000765 except AttributeError:
766 # io.BytesIO and other file-like objects don't have a `mode`
767 # attribute.
768 pass
769 else:
770 if "b" not in mode:
771 encode = True
772 if self.debuglevel > 0:
773 print("encoding file using iso-8859-1")
Jeremy Hylton636950f2009-03-28 04:34:21 +0000774 while 1:
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000775 datablock = data.read(blocksize)
776 if not datablock:
Jeremy Hylton636950f2009-03-28 04:34:21 +0000777 break
778 if encode:
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000779 datablock = datablock.encode("iso-8859-1")
780 self.sock.sendall(datablock)
Jeremy Hylton636950f2009-03-28 04:34:21 +0000781 else:
Senthil Kumaran996e13c2010-10-02 10:35:24 +0000782 self.sock.sendall(data)
Greg Stein5e0fa402000-06-26 08:28:01 +0000783
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000784 def _output(self, s):
785 """Add a line of output to the current request buffer.
Tim Peters469cdad2002-08-08 20:19:19 +0000786
Jeremy Hyltone3252ec2002-07-16 21:41:43 +0000787 Assumes that the line does *not* end with \\r\\n.
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000788 """
789 self._buffer.append(s)
790
Benjamin Peterson1742e402008-11-30 22:15:29 +0000791 def _send_output(self, message_body=None):
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000792 """Send the currently buffered request and clear the buffer.
793
Jeremy Hyltone3252ec2002-07-16 21:41:43 +0000794 Appends an extra \\r\\n to the buffer.
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000795 A message_body may be specified, to be appended to the request.
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000796 """
Martin v. Löwisdd5a8602007-06-30 09:22:09 +0000797 self._buffer.extend((b"", b""))
798 msg = b"\r\n".join(self._buffer)
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000799 del self._buffer[:]
Benjamin Peterson1742e402008-11-30 22:15:29 +0000800 # If msg and message_body are sent in a single send() call,
801 # it will avoid performance problems caused by the interaction
Ezio Melotti42da6632011-03-15 05:18:48 +0200802 # between delayed ack and the Nagle algorithm.
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000803 if isinstance(message_body, bytes):
Benjamin Peterson1742e402008-11-30 22:15:29 +0000804 msg += message_body
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000805 message_body = None
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000806 self.send(msg)
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000807 if message_body is not None:
Jeremy Hylton236654b2009-03-27 20:24:34 +0000808 # message_body was not a string (i.e. it is a file), and
809 # we must run the risk of Nagle.
Benjamin Peterson822b21c2009-01-18 00:04:57 +0000810 self.send(message_body)
Jeremy Hylton8531b1b2002-07-16 21:21:11 +0000811
Martin v. Löwisaf7dc8d2003-11-19 19:51:55 +0000812 def putrequest(self, method, url, skip_host=0, skip_accept_encoding=0):
Greg Steindd6eefb2000-07-18 09:09:48 +0000813 """Send a request to the server.
Greg Stein5e0fa402000-06-26 08:28:01 +0000814
Greg Steindd6eefb2000-07-18 09:09:48 +0000815 `method' specifies an HTTP request method, e.g. 'GET'.
816 `url' specifies the object being requested, e.g. '/index.html'.
Martin v. Löwisaf7dc8d2003-11-19 19:51:55 +0000817 `skip_host' if True does not add automatically a 'Host:' header
818 `skip_accept_encoding' if True does not add automatically an
819 'Accept-Encoding:' header
Greg Steindd6eefb2000-07-18 09:09:48 +0000820 """
Greg Stein5e0fa402000-06-26 08:28:01 +0000821
Greg Stein616a58d2003-06-24 06:35:19 +0000822 # if a prior response has been completed, then forget about it.
Greg Steindd6eefb2000-07-18 09:09:48 +0000823 if self.__response and self.__response.isclosed():
824 self.__response = None
Greg Stein5e0fa402000-06-26 08:28:01 +0000825
Tim Peters58eb11c2004-01-18 20:29:55 +0000826
Greg Steindd6eefb2000-07-18 09:09:48 +0000827 # in certain cases, we cannot issue another request on this connection.
828 # this occurs when:
829 # 1) we are in the process of sending a request. (_CS_REQ_STARTED)
830 # 2) a response to a previous request has signalled that it is going
831 # to close the connection upon completion.
832 # 3) the headers for the previous response have not been read, thus
833 # we cannot determine whether point (2) is true. (_CS_REQ_SENT)
834 #
835 # if there is no prior response, then we can request at will.
836 #
837 # if point (2) is true, then we will have passed the socket to the
838 # response (effectively meaning, "there is no prior response"), and
839 # will open a new one when a new request is made.
840 #
841 # Note: if a prior response exists, then we *can* start a new request.
842 # We are not allowed to begin fetching the response to this new
843 # request, however, until that prior response is complete.
844 #
845 if self.__state == _CS_IDLE:
846 self.__state = _CS_REQ_STARTED
847 else:
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +0000848 raise CannotSendRequest(self.__state)
Greg Stein5e0fa402000-06-26 08:28:01 +0000849
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +0000850 # Save the method we use, we need it later in the response phase
851 self._method = method
Greg Steindd6eefb2000-07-18 09:09:48 +0000852 if not url:
853 url = '/'
Martin v. Löwisdd5a8602007-06-30 09:22:09 +0000854 request = '%s %s %s' % (method, url, self._http_vsn_str)
Greg Stein5e0fa402000-06-26 08:28:01 +0000855
Martin v. Löwisdd5a8602007-06-30 09:22:09 +0000856 # Non-ASCII characters should have been eliminated earlier
857 self._output(request.encode('ascii'))
Greg Stein5e0fa402000-06-26 08:28:01 +0000858
Greg Steindd6eefb2000-07-18 09:09:48 +0000859 if self._http_vsn == 11:
860 # Issue some standard headers for better HTTP/1.1 compliance
Greg Stein5e0fa402000-06-26 08:28:01 +0000861
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000862 if not skip_host:
863 # this header is issued *only* for HTTP/1.1
864 # connections. more specifically, this means it is
865 # only issued when the client uses the new
866 # HTTPConnection() class. backwards-compat clients
867 # will be using HTTP/1.0 and those clients may be
868 # issuing this header themselves. we should NOT issue
869 # it twice; some web servers (such as Apache) barf
870 # when they see two Host: headers
Guido van Rossumf6922aa2001-01-14 21:03:01 +0000871
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000872 # If we need a non-standard port,include it in the
873 # header. If the request is going through a proxy,
874 # but the host of the actual URL, not the host of the
875 # proxy.
Jeremy Hylton8acf1e02002-03-08 19:35:51 +0000876
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000877 netloc = ''
878 if url.startswith('http'):
879 nil, netloc, nil, nil, nil = urlsplit(url)
880
881 if netloc:
Thomas Wouters477c8d52006-05-27 19:21:47 +0000882 try:
883 netloc_enc = netloc.encode("ascii")
884 except UnicodeEncodeError:
885 netloc_enc = netloc.encode("idna")
886 self.putheader('Host', netloc_enc)
Jeremy Hylton3921ff62002-03-09 06:07:23 +0000887 else:
Thomas Wouters477c8d52006-05-27 19:21:47 +0000888 try:
889 host_enc = self.host.encode("ascii")
890 except UnicodeEncodeError:
891 host_enc = self.host.encode("idna")
Senthil Kumaran2e89cf32010-11-14 03:14:52 +0000892
893 # As per RFC 273, IPv6 address should be wrapped with []
894 # when used as Host header
895
896 if self.host.find(':') >= 0:
897 host_enc = b'[' + host_enc + b']'
898
Georg Brandl86b2fb92008-07-16 03:43:04 +0000899 if self.port == self.default_port:
Thomas Wouters477c8d52006-05-27 19:21:47 +0000900 self.putheader('Host', host_enc)
901 else:
Guido van Rossum98297ee2007-11-06 21:34:58 +0000902 host_enc = host_enc.decode("ascii")
Thomas Wouters477c8d52006-05-27 19:21:47 +0000903 self.putheader('Host', "%s:%s" % (host_enc, self.port))
Greg Stein5e0fa402000-06-26 08:28:01 +0000904
Greg Steindd6eefb2000-07-18 09:09:48 +0000905 # note: we are assuming that clients will not attempt to set these
906 # headers since *this* library must deal with the
907 # consequences. this also means that when the supporting
908 # libraries are updated to recognize other forms, then this
909 # code should be changed (removed or updated).
Greg Stein5e0fa402000-06-26 08:28:01 +0000910
Greg Steindd6eefb2000-07-18 09:09:48 +0000911 # we only want a Content-Encoding of "identity" since we don't
912 # support encodings such as x-gzip or x-deflate.
Martin v. Löwisaf7dc8d2003-11-19 19:51:55 +0000913 if not skip_accept_encoding:
914 self.putheader('Accept-Encoding', 'identity')
Greg Stein5e0fa402000-06-26 08:28:01 +0000915
Greg Steindd6eefb2000-07-18 09:09:48 +0000916 # we can accept "chunked" Transfer-Encodings, but no others
917 # NOTE: no TE header implies *only* "chunked"
918 #self.putheader('TE', 'chunked')
Greg Stein5e0fa402000-06-26 08:28:01 +0000919
Greg Steindd6eefb2000-07-18 09:09:48 +0000920 # if TE is supplied in the header, then it must appear in a
921 # Connection header.
922 #self.putheader('Connection', 'TE')
Greg Stein5e0fa402000-06-26 08:28:01 +0000923
Greg Steindd6eefb2000-07-18 09:09:48 +0000924 else:
925 # For HTTP/1.0, the server will assume "not chunked"
926 pass
Greg Stein5e0fa402000-06-26 08:28:01 +0000927
Benjamin Petersonf608c612008-11-16 18:33:53 +0000928 def putheader(self, header, *values):
Greg Steindd6eefb2000-07-18 09:09:48 +0000929 """Send a request header line to the server.
Greg Stein5e0fa402000-06-26 08:28:01 +0000930
Greg Steindd6eefb2000-07-18 09:09:48 +0000931 For example: h.putheader('Accept', 'text/html')
932 """
933 if self.__state != _CS_REQ_STARTED:
934 raise CannotSendHeader()
Greg Stein5e0fa402000-06-26 08:28:01 +0000935
Guido van Rossum98297ee2007-11-06 21:34:58 +0000936 if hasattr(header, 'encode'):
937 header = header.encode('ascii')
Benjamin Petersonf608c612008-11-16 18:33:53 +0000938 values = list(values)
939 for i, one_value in enumerate(values):
940 if hasattr(one_value, 'encode'):
941 values[i] = one_value.encode('ascii')
Senthil Kumaran97304562010-10-03 18:25:01 +0000942 elif isinstance(one_value, int):
943 values[i] = str(one_value).encode('ascii')
Benjamin Petersonf608c612008-11-16 18:33:53 +0000944 value = b'\r\n\t'.join(values)
Guido van Rossum98297ee2007-11-06 21:34:58 +0000945 header = header + b': ' + value
946 self._output(header)
Greg Stein5e0fa402000-06-26 08:28:01 +0000947
Benjamin Peterson1742e402008-11-30 22:15:29 +0000948 def endheaders(self, message_body=None):
949 """Indicate that the last header line has been sent to the server.
Greg Stein5e0fa402000-06-26 08:28:01 +0000950
Benjamin Peterson1742e402008-11-30 22:15:29 +0000951 This method sends the request to the server. The optional
952 message_body argument can be used to pass message body
953 associated with the request. The message body will be sent in
954 the same packet as the message headers if possible. The
955 message_body should be a string.
956 """
Greg Steindd6eefb2000-07-18 09:09:48 +0000957 if self.__state == _CS_REQ_STARTED:
958 self.__state = _CS_REQ_SENT
959 else:
960 raise CannotSendHeader()
Benjamin Peterson1742e402008-11-30 22:15:29 +0000961 self._send_output(message_body)
Greg Stein5e0fa402000-06-26 08:28:01 +0000962
Greg Steindd6eefb2000-07-18 09:09:48 +0000963 def request(self, method, url, body=None, headers={}):
964 """Send a complete request to the server."""
Jeremy Hylton636950f2009-03-28 04:34:21 +0000965 self._send_request(method, url, body, headers)
Greg Stein5e0fa402000-06-26 08:28:01 +0000966
Benjamin Peterson1742e402008-11-30 22:15:29 +0000967 def _set_content_length(self, body):
968 # Set the content-length based on the body.
969 thelen = None
970 try:
971 thelen = str(len(body))
972 except TypeError as te:
973 # If this is a file-like object, try to
974 # fstat its file descriptor
Benjamin Peterson1742e402008-11-30 22:15:29 +0000975 try:
976 thelen = str(os.fstat(body.fileno()).st_size)
977 except (AttributeError, OSError):
978 # Don't send a length if this failed
979 if self.debuglevel > 0: print("Cannot stat!!")
980
981 if thelen is not None:
982 self.putheader('Content-Length', thelen)
983
Greg Steindd6eefb2000-07-18 09:09:48 +0000984 def _send_request(self, method, url, body, headers):
Jeremy Hylton636950f2009-03-28 04:34:21 +0000985 # Honor explicitly requested Host: and Accept-Encoding: headers.
Jeremy Hylton2c178252004-08-07 16:28:14 +0000986 header_names = dict.fromkeys([k.lower() for k in headers])
987 skips = {}
988 if 'host' in header_names:
989 skips['skip_host'] = 1
990 if 'accept-encoding' in header_names:
991 skips['skip_accept_encoding'] = 1
Greg Stein5e0fa402000-06-26 08:28:01 +0000992
Jeremy Hylton2c178252004-08-07 16:28:14 +0000993 self.putrequest(method, url, **skips)
994
995 if body and ('content-length' not in header_names):
Benjamin Peterson1742e402008-11-30 22:15:29 +0000996 self._set_content_length(body)
Guido van Rossumcc2b0162007-02-11 06:12:03 +0000997 for hdr, value in headers.items():
Greg Steindd6eefb2000-07-18 09:09:48 +0000998 self.putheader(hdr, value)
Jeremy Hyltonef9f48e2009-03-26 22:04:05 +0000999 if isinstance(body, str):
Jeremy Hylton236654b2009-03-27 20:24:34 +00001000 # RFC 2616 Section 3.7.1 says that text default has a
1001 # default charset of iso-8859-1.
1002 body = body.encode('iso-8859-1')
Jeremy Hyltonef9f48e2009-03-26 22:04:05 +00001003 self.endheaders(body)
Greg Stein5e0fa402000-06-26 08:28:01 +00001004
Greg Steindd6eefb2000-07-18 09:09:48 +00001005 def getresponse(self):
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +00001006 """Get the response from the server.
1007
1008 If the HTTPConnection is in the correct state, returns an
1009 instance of HTTPResponse or of whatever object is returned by
1010 class the response_class variable.
1011
1012 If a request has not been sent or if a previous response has
1013 not be handled, ResponseNotReady is raised. If the HTTP
1014 response indicates that the connection should be closed, then
1015 it will be closed before the response is returned. When the
1016 connection is closed, the underlying socket is closed.
1017 """
Greg Stein5e0fa402000-06-26 08:28:01 +00001018
Greg Stein616a58d2003-06-24 06:35:19 +00001019 # if a prior response has been completed, then forget about it.
Greg Steindd6eefb2000-07-18 09:09:48 +00001020 if self.__response and self.__response.isclosed():
1021 self.__response = None
Greg Stein5e0fa402000-06-26 08:28:01 +00001022
Greg Steindd6eefb2000-07-18 09:09:48 +00001023 # if a prior response exists, then it must be completed (otherwise, we
1024 # cannot read this response's header to determine the connection-close
1025 # behavior)
1026 #
1027 # note: if a prior response existed, but was connection-close, then the
1028 # socket and response were made independent of this HTTPConnection
1029 # object since a new request requires that we open a whole new
1030 # connection
1031 #
1032 # this means the prior response had one of two states:
1033 # 1) will_close: this connection was reset and the prior socket and
1034 # response operate independently
1035 # 2) persistent: the response was retained and we await its
1036 # isclosed() status to become true.
1037 #
1038 if self.__state != _CS_REQ_SENT or self.__response:
Jeremy Hylton6c5e28c2009-03-31 14:35:53 +00001039 raise ResponseNotReady(self.__state)
Greg Stein5e0fa402000-06-26 08:28:01 +00001040
Jeremy Hylton30f86742000-09-18 22:50:38 +00001041 if self.debuglevel > 0:
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001042 response = self.response_class(self.sock, self.debuglevel,
Tim Petersc2659cf2003-05-12 20:19:37 +00001043 strict=self.strict,
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +00001044 method=self._method)
Jeremy Hylton30f86742000-09-18 22:50:38 +00001045 else:
Jeremy Hyltonc1b2cb92003-05-05 16:13:58 +00001046 response = self.response_class(self.sock, strict=self.strict,
1047 method=self._method)
Greg Stein5e0fa402000-06-26 08:28:01 +00001048
Jeremy Hylton39c03802002-07-12 14:04:09 +00001049 response.begin()
Jeremy Hyltonbe4fcf12002-06-28 22:38:01 +00001050 assert response.will_close != _UNKNOWN
Greg Steindd6eefb2000-07-18 09:09:48 +00001051 self.__state = _CS_IDLE
Greg Stein5e0fa402000-06-26 08:28:01 +00001052
Greg Steindd6eefb2000-07-18 09:09:48 +00001053 if response.will_close:
Guido van Rossumd8faa362007-04-27 19:54:29 +00001054 # this effectively passes the connection to the response
1055 self.close()
Greg Steindd6eefb2000-07-18 09:09:48 +00001056 else:
1057 # remember this, so we can tell when it is complete
1058 self.__response = response
Greg Stein5e0fa402000-06-26 08:28:01 +00001059
Greg Steindd6eefb2000-07-18 09:09:48 +00001060 return response
Greg Stein5e0fa402000-06-26 08:28:01 +00001061
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001062try:
1063 import ssl
1064except ImportError:
1065 pass
1066else:
1067 class HTTPSConnection(HTTPConnection):
1068 "This class allows communication via SSL."
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001069
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001070 default_port = HTTPS_PORT
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001071
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001072 def __init__(self, host, port=None, key_file=None, cert_file=None,
Georg Brandlf78e02b2008-06-10 17:40:04 +00001073 strict=None, timeout=socket._GLOBAL_DEFAULT_TIMEOUT):
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001074 HTTPConnection.__init__(self, host, port, strict, timeout)
1075 self.key_file = key_file
1076 self.cert_file = cert_file
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001077
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001078 def connect(self):
1079 "Connect to a host on a given (SSL) port."
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001080
Jeremy Hylton636950f2009-03-28 04:34:21 +00001081 sock = socket.create_connection((self.host, self.port),
1082 self.timeout)
Senthil Kumaran0ac1f832009-07-26 12:39:47 +00001083
1084 if self._tunnel_host:
1085 self.sock = sock
1086 self._tunnel()
1087
Thomas Wouters1b7f8912007-09-19 03:06:30 +00001088 self.sock = ssl.wrap_socket(sock, self.key_file, self.cert_file)
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001089
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001090
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001091 def FakeSocket (sock, sslobj):
Thomas Wouters89d996e2007-09-08 17:39:28 +00001092 warnings.warn("FakeSocket is deprecated, and won't be in 3.x. " +
Thomas Wouters1b7f8912007-09-19 03:06:30 +00001093 "Use the result of ssl.wrap_socket() directly instead.",
Thomas Wouters89d996e2007-09-08 17:39:28 +00001094 DeprecationWarning, stacklevel=2)
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001095 return sslobj
Jeremy Hylton29d27ac2002-07-09 21:22:36 +00001096
Thomas Wouters47b49bf2007-08-30 22:15:33 +00001097 __all__.append("HTTPSConnection")
Greg Stein5e0fa402000-06-26 08:28:01 +00001098
Greg Stein5e0fa402000-06-26 08:28:01 +00001099class HTTPException(Exception):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001100 # Subclasses that define an __init__ must call Exception.__init__
1101 # or define self.args. Otherwise, str() will fail.
Greg Steindd6eefb2000-07-18 09:09:48 +00001102 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001103
1104class NotConnected(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001105 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001106
Skip Montanaro9d389972002-03-24 16:53:50 +00001107class InvalidURL(HTTPException):
1108 pass
1109
Greg Stein5e0fa402000-06-26 08:28:01 +00001110class UnknownProtocol(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001111 def __init__(self, version):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001112 self.args = version,
Greg Steindd6eefb2000-07-18 09:09:48 +00001113 self.version = version
Greg Stein5e0fa402000-06-26 08:28:01 +00001114
1115class UnknownTransferEncoding(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001116 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001117
Greg Stein5e0fa402000-06-26 08:28:01 +00001118class UnimplementedFileMode(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001119 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001120
1121class IncompleteRead(HTTPException):
Benjamin Peterson6accb982009-03-02 22:50:25 +00001122 def __init__(self, partial, expected=None):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001123 self.args = partial,
Greg Steindd6eefb2000-07-18 09:09:48 +00001124 self.partial = partial
Benjamin Peterson6accb982009-03-02 22:50:25 +00001125 self.expected = expected
1126 def __repr__(self):
1127 if self.expected is not None:
1128 e = ', %i more expected' % self.expected
1129 else:
1130 e = ''
1131 return 'IncompleteRead(%i bytes read%s)' % (len(self.partial), e)
1132 def __str__(self):
1133 return repr(self)
Greg Stein5e0fa402000-06-26 08:28:01 +00001134
1135class ImproperConnectionState(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001136 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001137
1138class CannotSendRequest(ImproperConnectionState):
Greg Steindd6eefb2000-07-18 09:09:48 +00001139 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001140
1141class CannotSendHeader(ImproperConnectionState):
Greg Steindd6eefb2000-07-18 09:09:48 +00001142 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001143
1144class ResponseNotReady(ImproperConnectionState):
Greg Steindd6eefb2000-07-18 09:09:48 +00001145 pass
Greg Stein5e0fa402000-06-26 08:28:01 +00001146
1147class BadStatusLine(HTTPException):
Greg Steindd6eefb2000-07-18 09:09:48 +00001148 def __init__(self, line):
Jeremy Hylton12f4f352002-07-06 18:55:01 +00001149 self.args = line,
Greg Steindd6eefb2000-07-18 09:09:48 +00001150 self.line = line
Greg Stein5e0fa402000-06-26 08:28:01 +00001151
Antoine Pitrouff1bbba2010-12-18 18:04:38 +00001152class LineTooLong(HTTPException):
1153 def __init__(self, line_type):
1154 HTTPException.__init__(self, "got more than %d bytes when reading %s"
1155 % (_MAXLINE, line_type))
1156
Greg Stein5e0fa402000-06-26 08:28:01 +00001157# for backwards compatibility
1158error = HTTPException
1159
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001160class LineAndFileWrapper:
1161 """A limited file-like object for HTTP/0.9 responses."""
1162
1163 # The status-line parsing code calls readline(), which normally
1164 # get the HTTP status line. For a 0.9 response, however, this is
1165 # actually the first line of the body! Clients need to get a
1166 # readable file object that contains that line.
1167
1168 def __init__(self, line, file):
1169 self._line = line
1170 self._file = file
1171 self._line_consumed = 0
1172 self._line_offset = 0
1173 self._line_left = len(line)
1174
1175 def __getattr__(self, attr):
1176 return getattr(self._file, attr)
1177
1178 def _done(self):
1179 # called when the last byte is read from the line. After the
1180 # call, all read methods are delegated to the underlying file
Skip Montanaro74b9a7a2003-02-25 17:48:15 +00001181 # object.
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001182 self._line_consumed = 1
1183 self.read = self._file.read
1184 self.readline = self._file.readline
1185 self.readlines = self._file.readlines
1186
1187 def read(self, amt=None):
Hye-Shik Chang39aef792004-06-05 13:30:56 +00001188 if self._line_consumed:
1189 return self._file.read(amt)
1190 assert self._line_left
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001191 if amt is None or amt > self._line_left:
1192 s = self._line[self._line_offset:]
1193 self._done()
1194 if amt is None:
1195 return s + self._file.read()
1196 else:
Tim Petersc411dba2002-07-16 21:35:23 +00001197 return s + self._file.read(amt - len(s))
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001198 else:
1199 assert amt <= self._line_left
1200 i = self._line_offset
1201 j = i + amt
1202 s = self._line[i:j]
1203 self._line_offset = j
1204 self._line_left -= amt
1205 if self._line_left == 0:
1206 self._done()
1207 return s
Tim Petersc411dba2002-07-16 21:35:23 +00001208
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001209 def readline(self):
Hye-Shik Chang39aef792004-06-05 13:30:56 +00001210 if self._line_consumed:
1211 return self._file.readline()
1212 assert self._line_left
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001213 s = self._line[self._line_offset:]
1214 self._done()
1215 return s
1216
1217 def readlines(self, size=None):
Hye-Shik Chang39aef792004-06-05 13:30:56 +00001218 if self._line_consumed:
1219 return self._file.readlines(size)
1220 assert self._line_left
Jeremy Hyltond46aa372002-07-06 18:48:07 +00001221 L = [self._line[self._line_offset:]]
1222 self._done()
1223 if size is None:
1224 return L + self._file.readlines()
1225 else:
1226 return L + self._file.readlines(size)