This commit was manufactured by cvs2svn to create tag 'r222'.
[python/dscho.git] / Lib / httplib.py
blob077fb255dc3b6f0c3d2acd521eef154300d102b5
1 """HTTP/1.1 client library
3 <intro stuff goes here>
4 <other stuff, too>
6 HTTPConnection go through a number of "states", which defines when a client
7 may legally make another request or fetch the response for a particular
8 request. This diagram details these state transitions:
10 (null)
12 | HTTPConnection()
14 Idle
16 | putrequest()
18 Request-started
20 | ( putheader() )* endheaders()
22 Request-sent
24 | response = getresponse()
26 Unread-response [Response-headers-read]
27 |\____________________
28 | |
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
38 | response.read()
40 Request-sent
42 This 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
48 Note: 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.
59 Logical State __state __response
60 ------------- ------- ----------
61 Idle _CS_IDLE None
62 Request-started _CS_REQ_STARTED None
63 Request-sent _CS_REQ_SENT None
64 Unread-response _CS_IDLE <response_class>
65 Req-started-unread-response _CS_REQ_STARTED <response_class>
66 Req-sent-unread-response _CS_REQ_SENT <response_class>
67 """
69 import errno
70 import mimetools
71 import socket
72 from urlparse import urlsplit
74 try:
75 from cStringIO import StringIO
76 except ImportError:
77 from StringIO import StringIO
79 __all__ = ["HTTP", "HTTPResponse", "HTTPConnection", "HTTPSConnection",
80 "HTTPException", "NotConnected", "UnknownProtocol",
81 "UnknownTransferEncoding", "UnimplementedFileMode",
82 "IncompleteRead", "InvalidURL", "ImproperConnectionState",
83 "CannotSendRequest", "CannotSendHeader", "ResponseNotReady",
84 "BadStatusLine", "error"]
86 HTTP_PORT = 80
87 HTTPS_PORT = 443
89 _UNKNOWN = 'UNKNOWN'
91 # connection states
92 _CS_IDLE = 'Idle'
93 _CS_REQ_STARTED = 'Request-started'
94 _CS_REQ_SENT = 'Request-sent'
96 class HTTPMessage(mimetools.Message):
98 def addheader(self, key, value):
99 """Add header for field key handling repeats."""
100 prev = self.dict.get(key)
101 if prev is None:
102 self.dict[key] = value
103 else:
104 combined = ", ".join((prev, value))
105 self.dict[key] = combined
107 def addcontinue(self, key, more):
108 """Add more field data from a continuation line."""
109 prev = self.dict[key]
110 self.dict[key] = prev + "\n " + more
112 def readheaders(self):
113 """Read header lines.
115 Read header lines up to the entirely blank line that terminates them.
116 The (normally blank) line that ends the headers is skipped, but not
117 included in the returned list. If a non-header line ends the headers,
118 (which is an error), an attempt is made to backspace over it; it is
119 never included in the returned list.
121 The variable self.status is set to the empty string if all went well,
122 otherwise it is an error message. The variable self.headers is a
123 completely uninterpreted list of lines contained in the header (so
124 printing them will reproduce the header exactly as it appears in the
125 file).
127 If multiple header fields with the same name occur, they are combined
128 according to the rules in RFC 2616 sec 4.2:
130 Appending each subsequent field-value to the first, each separated
131 by a comma. The order in which header fields with the same field-name
132 are received is significant to the interpretation of the combined
133 field value.
135 # XXX The implementation overrides the readheaders() method of
136 # rfc822.Message. The base class design isn't amenable to
137 # customized behavior here so the method here is a copy of the
138 # base class code with a few small changes.
140 self.dict = {}
141 self.unixfrom = ''
142 self.headers = list = []
143 self.status = ''
144 headerseen = ""
145 firstline = 1
146 startofline = unread = tell = None
147 if hasattr(self.fp, 'unread'):
148 unread = self.fp.unread
149 elif self.seekable:
150 tell = self.fp.tell
151 while 1:
152 if tell:
153 try:
154 startofline = tell()
155 except IOError:
156 startofline = tell = None
157 self.seekable = 0
158 line = self.fp.readline()
159 if not line:
160 self.status = 'EOF in headers'
161 break
162 # Skip unix From name time lines
163 if firstline and line.startswith('From '):
164 self.unixfrom = self.unixfrom + line
165 continue
166 firstline = 0
167 if headerseen and line[0] in ' \t':
168 # XXX Not sure if continuation lines are handled properly
169 # for http and/or for repeating headers
170 # It's a continuation line.
171 list.append(line)
172 x = self.dict[headerseen] + "\n " + line.strip()
173 self.addcontinue(headerseen, line.strip())
174 continue
175 elif self.iscomment(line):
176 # It's a comment. Ignore it.
177 continue
178 elif self.islast(line):
179 # Note! No pushback here! The delimiter line gets eaten.
180 break
181 headerseen = self.isheader(line)
182 if headerseen:
183 # It's a legal header line, save it.
184 list.append(line)
185 self.addheader(headerseen, line[len(headerseen)+1:].strip())
186 continue
187 else:
188 # It's not a header line; throw it back and stop here.
189 if not self.dict:
190 self.status = 'No headers'
191 else:
192 self.status = 'Non-header line where header expected'
193 # Try to undo the read.
194 if unread:
195 unread(line)
196 elif tell:
197 self.fp.seek(startofline)
198 else:
199 self.status = self.status + '; bad seek'
200 break
202 class HTTPResponse:
204 # strict: If true, raise BadStatusLine if the status line can't be
205 # parsed as a valid HTTP/1.0 or 1.1 status line. By default it is
206 # false because it prevents clients from talking to HTTP/0.9
207 # servers. Note that a response with a sufficiently corrupted
208 # status line will look like an HTTP/0.9 response.
210 # See RFC 2616 sec 19.6 and RFC 1945 sec 6 for details.
212 def __init__(self, sock, debuglevel=0, strict=0):
213 self.fp = sock.makefile('rb', 0)
214 self.debuglevel = debuglevel
215 self.strict = strict
217 self.msg = None
219 # from the Status-Line of the response
220 self.version = _UNKNOWN # HTTP-Version
221 self.status = _UNKNOWN # Status-Code
222 self.reason = _UNKNOWN # Reason-Phrase
224 self.chunked = _UNKNOWN # is "chunked" being used?
225 self.chunk_left = _UNKNOWN # bytes left to read in current chunk
226 self.length = _UNKNOWN # number of bytes left in response
227 self.will_close = _UNKNOWN # conn will close at end of response
229 def _read_status(self):
230 # Initialize with Simple-Response defaults
231 line = self.fp.readline()
232 if self.debuglevel > 0:
233 print "reply:", repr(line)
234 try:
235 [version, status, reason] = line.split(None, 2)
236 except ValueError:
237 try:
238 [version, status] = line.split(None, 1)
239 reason = ""
240 except ValueError:
241 # empty version will cause next test to fail and status
242 # will be treated as 0.9 response.
243 version = ""
244 if not version.startswith('HTTP/'):
245 if self.strict:
246 self.close()
247 raise BadStatusLine(line)
248 else:
249 # assume it's a Simple-Response from an 0.9 server
250 self.fp = LineAndFileWrapper(line, self.fp)
251 return "HTTP/0.9", 200, ""
253 # The status code is a three-digit number
254 try:
255 status = int(status)
256 if status < 100 or status > 999:
257 raise BadStatusLine(line)
258 except ValueError:
259 raise BadStatusLine(line)
260 return version, status, reason
262 def begin(self):
263 if self.msg is not None:
264 # we've already started reading the response
265 return
267 # read until we get a non-100 response
268 while 1:
269 version, status, reason = self._read_status()
270 if status != 100:
271 break
272 # skip the header from the 100 response
273 while 1:
274 skip = self.fp.readline().strip()
275 if not skip:
276 break
277 if self.debuglevel > 0:
278 print "header:", skip
280 self.status = status
281 self.reason = reason.strip()
282 if version == 'HTTP/1.0':
283 self.version = 10
284 elif version.startswith('HTTP/1.'):
285 self.version = 11 # use HTTP/1.1 code for HTTP/1.x where x>=1
286 elif version == 'HTTP/0.9':
287 self.version = 9
288 else:
289 raise UnknownProtocol(version)
291 if self.version == 9:
292 self.chunked = 0
293 self.will_close = 1
294 self.msg = HTTPMessage(StringIO())
295 return
297 self.msg = HTTPMessage(self.fp, 0)
298 if self.debuglevel > 0:
299 for hdr in self.msg.headers:
300 print "header:", hdr,
302 # don't let the msg keep an fp
303 self.msg.fp = None
305 # are we using the chunked-style of transfer encoding?
306 tr_enc = self.msg.getheader('transfer-encoding')
307 if tr_enc and tr_enc.lower() == "chunked":
308 self.chunked = 1
309 self.chunk_left = None
310 else:
311 self.chunked = 0
313 # will the connection close at the end of the response?
314 conn = self.msg.getheader('connection')
315 if conn:
316 conn = conn.lower()
317 # a "Connection: close" will always close the connection. if we
318 # don't see that and this is not HTTP/1.1, then the connection will
319 # close unless we see a Keep-Alive header.
320 self.will_close = conn.find('close') != -1 or \
321 ( self.version != 11 and \
322 not self.msg.getheader('keep-alive') )
323 else:
324 # for HTTP/1.1, the connection will always remain open
325 # otherwise, it will remain open IFF we see a Keep-Alive header
326 self.will_close = self.version != 11 and \
327 not self.msg.getheader('keep-alive')
329 # do we have a Content-Length?
330 # NOTE: RFC 2616, S4.4, #3 says we ignore this if tr_enc is "chunked"
331 length = self.msg.getheader('content-length')
332 if length and not self.chunked:
333 try:
334 self.length = int(length)
335 except ValueError:
336 self.length = None
337 else:
338 self.length = None
340 # does the body have a fixed length? (of zero)
341 if (status == 204 or # No Content
342 status == 304 or # Not Modified
343 100 <= status < 200): # 1xx codes
344 self.length = 0
346 # if the connection remains open, and we aren't using chunked, and
347 # a content-length was not provided, then assume that the connection
348 # WILL close.
349 if not self.will_close and \
350 not self.chunked and \
351 self.length is None:
352 self.will_close = 1
354 def close(self):
355 if self.fp:
356 self.fp.close()
357 self.fp = None
359 def isclosed(self):
360 # NOTE: it is possible that we will not ever call self.close(). This
361 # case occurs when will_close is TRUE, length is None, and we
362 # read up to the last byte, but NOT past it.
364 # IMPLIES: if will_close is FALSE, then self.close() will ALWAYS be
365 # called, meaning self.isclosed() is meaningful.
366 return self.fp is None
368 def read(self, amt=None):
369 if self.fp is None:
370 return ''
372 if self.chunked:
373 return self._read_chunked(amt)
375 if amt is None:
376 # unbounded read
377 if self.will_close:
378 s = self.fp.read()
379 else:
380 s = self._safe_read(self.length)
381 self.close() # we read everything
382 return s
384 if self.length is not None:
385 if amt > self.length:
386 # clip the read to the "end of response"
387 amt = self.length
388 self.length -= amt
390 # we do not use _safe_read() here because this may be a .will_close
391 # connection, and the user is reading more bytes than will be provided
392 # (for example, reading in 1k chunks)
393 s = self.fp.read(amt)
395 return s
397 def _read_chunked(self, amt):
398 assert self.chunked != _UNKNOWN
399 chunk_left = self.chunk_left
400 value = ''
402 # XXX This accumulates chunks by repeated string concatenation,
403 # which is not efficient as the number or size of chunks gets big.
404 while 1:
405 if chunk_left is None:
406 line = self.fp.readline()
407 i = line.find(';')
408 if i >= 0:
409 line = line[:i] # strip chunk-extensions
410 chunk_left = int(line, 16)
411 if chunk_left == 0:
412 break
413 if amt is None:
414 value += self._safe_read(chunk_left)
415 elif amt < chunk_left:
416 value += self._safe_read(amt)
417 self.chunk_left = chunk_left - amt
418 return value
419 elif amt == chunk_left:
420 value += self._safe_read(amt)
421 self._safe_read(2) # toss the CRLF at the end of the chunk
422 self.chunk_left = None
423 return value
424 else:
425 value += self._safe_read(chunk_left)
426 amt -= chunk_left
428 # we read the whole chunk, get another
429 self._safe_read(2) # toss the CRLF at the end of the chunk
430 chunk_left = None
432 # read and discard trailer up to the CRLF terminator
433 ### note: we shouldn't have any trailers!
434 while 1:
435 line = self.fp.readline()
436 if line == '\r\n':
437 break
439 # we read everything; close the "file"
440 # XXX Shouldn't the client close the file?
441 self.close()
443 return value
445 def _safe_read(self, amt):
446 """Read the number of bytes requested, compensating for partial reads.
448 Normally, we have a blocking socket, but a read() can be interrupted
449 by a signal (resulting in a partial read).
451 Note that we cannot distinguish between EOF and an interrupt when zero
452 bytes have been read. IncompleteRead() will be raised in this
453 situation.
455 This function should be used when <amt> bytes "should" be present for
456 reading. If the bytes are truly not available (due to EOF), then the
457 IncompleteRead exception can be used to detect the problem.
459 s = ''
460 while amt > 0:
461 chunk = self.fp.read(amt)
462 if not chunk:
463 raise IncompleteRead(s)
464 s = s + chunk
465 amt = amt - len(chunk)
466 return s
468 def getheader(self, name, default=None):
469 if self.msg is None:
470 raise ResponseNotReady()
471 return self.msg.getheader(name, default)
474 class HTTPConnection:
476 _http_vsn = 11
477 _http_vsn_str = 'HTTP/1.1'
479 response_class = HTTPResponse
480 default_port = HTTP_PORT
481 auto_open = 1
482 debuglevel = 0
483 strict = 0
485 def __init__(self, host, port=None, strict=None):
486 self.sock = None
487 self._buffer = []
488 self.__response = None
489 self.__state = _CS_IDLE
491 self._set_hostport(host, port)
492 if strict is not None:
493 self.strict = strict
495 def _set_hostport(self, host, port):
496 if port is None:
497 i = host.find(':')
498 if i >= 0:
499 try:
500 port = int(host[i+1:])
501 except ValueError:
502 raise InvalidURL("nonnumeric port: '%s'" % host[i+1:])
503 host = host[:i]
504 else:
505 port = self.default_port
506 self.host = host
507 self.port = port
509 def set_debuglevel(self, level):
510 self.debuglevel = level
512 def connect(self):
513 """Connect to the host and port specified in __init__."""
514 msg = "getaddrinfo returns an empty list"
515 for res in socket.getaddrinfo(self.host, self.port, 0,
516 socket.SOCK_STREAM):
517 af, socktype, proto, canonname, sa = res
518 try:
519 self.sock = socket.socket(af, socktype, proto)
520 if self.debuglevel > 0:
521 print "connect: (%s, %s)" % (self.host, self.port)
522 self.sock.connect(sa)
523 except socket.error, msg:
524 if self.debuglevel > 0:
525 print 'connect fail:', (self.host, self.port)
526 if self.sock:
527 self.sock.close()
528 self.sock = None
529 continue
530 break
531 if not self.sock:
532 raise socket.error, msg
534 def close(self):
535 """Close the connection to the HTTP server."""
536 if self.sock:
537 self.sock.close() # close it manually... there may be other refs
538 self.sock = None
539 if self.__response:
540 self.__response.close()
541 self.__response = None
542 self.__state = _CS_IDLE
544 def send(self, str):
545 """Send `str' to the server."""
546 if self.sock is None:
547 if self.auto_open:
548 self.connect()
549 else:
550 raise NotConnected()
552 # send the data to the server. if we get a broken pipe, then close
553 # the socket. we want to reconnect when somebody tries to send again.
555 # NOTE: we DO propagate the error, though, because we cannot simply
556 # ignore the error... the caller will know if they can retry.
557 if self.debuglevel > 0:
558 print "send:", repr(str)
559 try:
560 self.sock.sendall(str)
561 except socket.error, v:
562 if v[0] == 32: # Broken pipe
563 self.close()
564 raise
566 def _output(self, s):
567 """Add a line of output to the current request buffer.
569 Assumes that the line does *not* end with \\r\\n.
571 self._buffer.append(s)
573 def _send_output(self):
574 """Send the currently buffered request and clear the buffer.
576 Appends an extra \\r\\n to the buffer.
578 self._buffer.extend(("", ""))
579 msg = "\r\n".join(self._buffer)
580 del self._buffer[:]
581 self.send(msg)
583 def putrequest(self, method, url, skip_host=0):
584 """Send a request to the server.
586 `method' specifies an HTTP request method, e.g. 'GET'.
587 `url' specifies the object being requested, e.g. '/index.html'.
590 # check if a prior response has been completed
591 # XXX What if it hasn't?
592 if self.__response and self.__response.isclosed():
593 self.__response = None
596 # in certain cases, we cannot issue another request on this connection.
597 # this occurs when:
598 # 1) we are in the process of sending a request. (_CS_REQ_STARTED)
599 # 2) a response to a previous request has signalled that it is going
600 # to close the connection upon completion.
601 # 3) the headers for the previous response have not been read, thus
602 # we cannot determine whether point (2) is true. (_CS_REQ_SENT)
604 # if there is no prior response, then we can request at will.
606 # if point (2) is true, then we will have passed the socket to the
607 # response (effectively meaning, "there is no prior response"), and
608 # will open a new one when a new request is made.
610 # Note: if a prior response exists, then we *can* start a new request.
611 # We are not allowed to begin fetching the response to this new
612 # request, however, until that prior response is complete.
614 if self.__state == _CS_IDLE:
615 self.__state = _CS_REQ_STARTED
616 else:
617 raise CannotSendRequest()
619 if not url:
620 url = '/'
621 str = '%s %s %s' % (method, url, self._http_vsn_str)
623 self._output(str)
625 if self._http_vsn == 11:
626 # Issue some standard headers for better HTTP/1.1 compliance
628 if not skip_host:
629 # this header is issued *only* for HTTP/1.1
630 # connections. more specifically, this means it is
631 # only issued when the client uses the new
632 # HTTPConnection() class. backwards-compat clients
633 # will be using HTTP/1.0 and those clients may be
634 # issuing this header themselves. we should NOT issue
635 # it twice; some web servers (such as Apache) barf
636 # when they see two Host: headers
638 # If we need a non-standard port,include it in the
639 # header. If the request is going through a proxy,
640 # but the host of the actual URL, not the host of the
641 # proxy.
643 netloc = ''
644 if url.startswith('http'):
645 nil, netloc, nil, nil, nil = urlsplit(url)
647 if netloc:
648 self.putheader('Host', netloc)
649 elif self.port == HTTP_PORT:
650 self.putheader('Host', self.host)
651 else:
652 self.putheader('Host', "%s:%s" % (self.host, self.port))
654 # note: we are assuming that clients will not attempt to set these
655 # headers since *this* library must deal with the
656 # consequences. this also means that when the supporting
657 # libraries are updated to recognize other forms, then this
658 # code should be changed (removed or updated).
660 # we only want a Content-Encoding of "identity" since we don't
661 # support encodings such as x-gzip or x-deflate.
662 self.putheader('Accept-Encoding', 'identity')
664 # we can accept "chunked" Transfer-Encodings, but no others
665 # NOTE: no TE header implies *only* "chunked"
666 #self.putheader('TE', 'chunked')
668 # if TE is supplied in the header, then it must appear in a
669 # Connection header.
670 #self.putheader('Connection', 'TE')
672 else:
673 # For HTTP/1.0, the server will assume "not chunked"
674 pass
676 def putheader(self, header, value):
677 """Send a request header line to the server.
679 For example: h.putheader('Accept', 'text/html')
681 if self.__state != _CS_REQ_STARTED:
682 raise CannotSendHeader()
684 str = '%s: %s' % (header, value)
685 self._output(str)
687 def endheaders(self):
688 """Indicate that the last header line has been sent to the server."""
690 if self.__state == _CS_REQ_STARTED:
691 self.__state = _CS_REQ_SENT
692 else:
693 raise CannotSendHeader()
695 self._send_output()
697 def request(self, method, url, body=None, headers={}):
698 """Send a complete request to the server."""
700 try:
701 self._send_request(method, url, body, headers)
702 except socket.error, v:
703 # trap 'Broken pipe' if we're allowed to automatically reconnect
704 if v[0] != 32 or not self.auto_open:
705 raise
706 # try one more time
707 self._send_request(method, url, body, headers)
709 def _send_request(self, method, url, body, headers):
710 # If headers already contains a host header, then define the
711 # optional skip_host argument to putrequest(). The check is
712 # harder because field names are case insensitive.
713 if 'Host' in (headers
714 or [k for k in headers.iterkeys() if k.lower() == "host"]):
715 self.putrequest(method, url, skip_host=1)
716 else:
717 self.putrequest(method, url)
719 if body:
720 self.putheader('Content-Length', str(len(body)))
721 for hdr, value in headers.items():
722 self.putheader(hdr, value)
723 self.endheaders()
725 if body:
726 self.send(body)
728 def getresponse(self):
729 "Get the response from the server."
731 # check if a prior response has been completed
732 if self.__response and self.__response.isclosed():
733 self.__response = None
736 # if a prior response exists, then it must be completed (otherwise, we
737 # cannot read this response's header to determine the connection-close
738 # behavior)
740 # note: if a prior response existed, but was connection-close, then the
741 # socket and response were made independent of this HTTPConnection
742 # object since a new request requires that we open a whole new
743 # connection
745 # this means the prior response had one of two states:
746 # 1) will_close: this connection was reset and the prior socket and
747 # response operate independently
748 # 2) persistent: the response was retained and we await its
749 # isclosed() status to become true.
751 if self.__state != _CS_REQ_SENT or self.__response:
752 raise ResponseNotReady()
754 if self.debuglevel > 0:
755 response = self.response_class(self.sock, self.debuglevel,
756 strict=self.strict)
757 else:
758 response = self.response_class(self.sock, strict=self.strict)
760 response.begin()
761 assert response.will_close != _UNKNOWN
762 self.__state = _CS_IDLE
764 if response.will_close:
765 # this effectively passes the connection to the response
766 self.close()
767 else:
768 # remember this, so we can tell when it is complete
769 self.__response = response
771 return response
773 # The next several classes are used to define FakeSocket,a socket-like
774 # interface to an SSL connection.
776 # The primary complexity comes from faking a makefile() method. The
777 # standard socket makefile() implementation calls dup() on the socket
778 # file descriptor. As a consequence, clients can call close() on the
779 # parent socket and its makefile children in any order. The underlying
780 # socket isn't closed until they are all closed.
782 # The implementation uses reference counting to keep the socket open
783 # until the last client calls close(). SharedSocket keeps track of
784 # the reference counting and SharedSocketClient provides an constructor
785 # and close() method that call incref() and decref() correctly.
787 class SharedSocket:
789 def __init__(self, sock):
790 self.sock = sock
791 self._refcnt = 0
793 def incref(self):
794 self._refcnt += 1
796 def decref(self):
797 self._refcnt -= 1
798 assert self._refcnt >= 0
799 if self._refcnt == 0:
800 self.sock.close()
802 def __del__(self):
803 self.sock.close()
805 class SharedSocketClient:
807 def __init__(self, shared):
808 self._closed = 0
809 self._shared = shared
810 self._shared.incref()
811 self._sock = shared.sock
813 def close(self):
814 if not self._closed:
815 self._shared.decref()
816 self._closed = 1
817 self._shared = None
819 class SSLFile(SharedSocketClient):
820 """File-like object wrapping an SSL socket."""
822 BUFSIZE = 8192
824 def __init__(self, sock, ssl, bufsize=None):
825 SharedSocketClient.__init__(self, sock)
826 self._ssl = ssl
827 self._buf = ''
828 self._bufsize = bufsize or self.__class__.BUFSIZE
830 def _read(self):
831 buf = ''
832 # put in a loop so that we retry on transient errors
833 while 1:
834 try:
835 buf = self._ssl.read(self._bufsize)
836 except socket.sslerror, err:
837 if (err[0] == socket.SSL_ERROR_WANT_READ
838 or err[0] == socket.SSL_ERROR_WANT_WRITE):
839 continue
840 if (err[0] == socket.SSL_ERROR_ZERO_RETURN
841 or err[0] == socket.SSL_ERROR_EOF):
842 break
843 raise
844 except socket.error, err:
845 if err[0] == errno.EINTR:
846 continue
847 if err[0] == errno.EBADF:
848 # XXX socket was closed?
849 break
850 raise
851 else:
852 break
853 return buf
855 def read(self, size=None):
856 L = [self._buf]
857 avail = len(self._buf)
858 while size is None or avail < size:
859 s = self._read()
860 if s == '':
861 break
862 L.append(s)
863 avail += len(s)
864 all = "".join(L)
865 if size is None:
866 self._buf = ''
867 return all
868 else:
869 self._buf = all[size:]
870 return all[:size]
872 def readline(self):
873 L = [self._buf]
874 self._buf = ''
875 while 1:
876 i = L[-1].find("\n")
877 if i >= 0:
878 break
879 s = self._read()
880 if s == '':
881 break
882 L.append(s)
883 if i == -1:
884 # loop exited because there is no more data
885 return "".join(L)
886 else:
887 all = "".join(L)
888 # XXX could do enough bookkeeping not to do a 2nd search
889 i = all.find("\n") + 1
890 line = all[:i]
891 self._buf = all[i:]
892 return line
894 class FakeSocket(SharedSocketClient):
896 class _closedsocket:
897 def __getattr__(self, name):
898 raise error(9, 'Bad file descriptor')
900 def __init__(self, sock, ssl):
901 sock = SharedSocket(sock)
902 SharedSocketClient.__init__(self, sock)
903 self._ssl = ssl
905 def close(self):
906 SharedSocketClient.close(self)
907 self._sock = self.__class__._closedsocket()
909 def makefile(self, mode, bufsize=None):
910 if mode != 'r' and mode != 'rb':
911 raise UnimplementedFileMode()
912 return SSLFile(self._shared, self._ssl, bufsize)
914 def send(self, stuff, flags = 0):
915 return self._ssl.write(stuff)
917 sendall = send
919 def recv(self, len = 1024, flags = 0):
920 return self._ssl.read(len)
922 def __getattr__(self, attr):
923 return getattr(self._sock, attr)
926 class HTTPSConnection(HTTPConnection):
927 "This class allows communication via SSL."
929 default_port = HTTPS_PORT
931 def __init__(self, host, port=None, key_file=None, cert_file=None,
932 strict=None):
933 HTTPConnection.__init__(self, host, port, strict)
934 self.key_file = key_file
935 self.cert_file = cert_file
937 def connect(self):
938 "Connect to a host on a given (SSL) port."
940 sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
941 sock.connect((self.host, self.port))
942 realsock = sock
943 if hasattr(sock, "_sock"):
944 realsock = sock._sock
945 ssl = socket.ssl(realsock, self.key_file, self.cert_file)
946 self.sock = FakeSocket(sock, ssl)
949 class HTTP:
950 "Compatibility class with httplib.py from 1.5."
952 _http_vsn = 10
953 _http_vsn_str = 'HTTP/1.0'
955 debuglevel = 0
957 _connection_class = HTTPConnection
959 def __init__(self, host='', port=None, strict=None):
960 "Provide a default host, since the superclass requires one."
962 # some joker passed 0 explicitly, meaning default port
963 if port == 0:
964 port = None
966 # Note that we may pass an empty string as the host; this will throw
967 # an error when we attempt to connect. Presumably, the client code
968 # will call connect before then, with a proper host.
969 self._setup(self._connection_class(host, port, strict))
971 def _setup(self, conn):
972 self._conn = conn
974 # set up delegation to flesh out interface
975 self.send = conn.send
976 self.putrequest = conn.putrequest
977 self.endheaders = conn.endheaders
978 self.set_debuglevel = conn.set_debuglevel
980 conn._http_vsn = self._http_vsn
981 conn._http_vsn_str = self._http_vsn_str
983 self.file = None
985 def connect(self, host=None, port=None):
986 "Accept arguments to set the host/port, since the superclass doesn't."
988 if host is not None:
989 self._conn._set_hostport(host, port)
990 self._conn.connect()
992 def getfile(self):
993 "Provide a getfile, since the superclass' does not use this concept."
994 return self.file
996 def putheader(self, header, *values):
997 "The superclass allows only one value argument."
998 self._conn.putheader(header, '\r\n\t'.join(values))
1000 def getreply(self):
1001 """Compat definition since superclass does not define it.
1003 Returns a tuple consisting of:
1004 - server status code (e.g. '200' if all goes well)
1005 - server "reason" corresponding to status code
1006 - any RFC822 headers in the response from the server
1008 try:
1009 response = self._conn.getresponse()
1010 except BadStatusLine, e:
1011 ### hmm. if getresponse() ever closes the socket on a bad request,
1012 ### then we are going to have problems with self.sock
1014 ### should we keep this behavior? do people use it?
1015 # keep the socket open (as a file), and return it
1016 self.file = self._conn.sock.makefile('rb', 0)
1018 # close our socket -- we want to restart after any protocol error
1019 self.close()
1021 self.headers = None
1022 return -1, e.line, None
1024 self.headers = response.msg
1025 self.file = response.fp
1026 return response.status, response.reason, response.msg
1028 def close(self):
1029 self._conn.close()
1031 # note that self.file == response.fp, which gets closed by the
1032 # superclass. just clear the object ref here.
1033 ### hmm. messy. if status==-1, then self.file is owned by us.
1034 ### well... we aren't explicitly closing, but losing this ref will
1035 ### do it
1036 self.file = None
1038 if hasattr(socket, 'ssl'):
1039 class HTTPS(HTTP):
1040 """Compatibility with 1.5 httplib interface
1042 Python 1.5.2 did not have an HTTPS class, but it defined an
1043 interface for sending http requests that is also useful for
1044 https.
1047 _connection_class = HTTPSConnection
1049 def __init__(self, host='', port=None, key_file=None, cert_file=None,
1050 strict=None):
1051 # provide a default host, pass the X509 cert info
1053 # urf. compensate for bad input.
1054 if port == 0:
1055 port = None
1056 self._setup(self._connection_class(host, port, key_file,
1057 cert_file, strict))
1059 # we never actually use these for anything, but we keep them
1060 # here for compatibility with post-1.5.2 CVS.
1061 self.key_file = key_file
1062 self.cert_file = cert_file
1065 class HTTPException(Exception):
1066 # Subclasses that define an __init__ must call Exception.__init__
1067 # or define self.args. Otherwise, str() will fail.
1068 pass
1070 class NotConnected(HTTPException):
1071 pass
1073 class InvalidURL(HTTPException):
1074 pass
1076 class UnknownProtocol(HTTPException):
1077 def __init__(self, version):
1078 self.args = version,
1079 self.version = version
1081 class UnknownTransferEncoding(HTTPException):
1082 pass
1084 class UnimplementedFileMode(HTTPException):
1085 pass
1087 class IncompleteRead(HTTPException):
1088 def __init__(self, partial):
1089 self.args = partial,
1090 self.partial = partial
1092 class ImproperConnectionState(HTTPException):
1093 pass
1095 class CannotSendRequest(ImproperConnectionState):
1096 pass
1098 class CannotSendHeader(ImproperConnectionState):
1099 pass
1101 class ResponseNotReady(ImproperConnectionState):
1102 pass
1104 class BadStatusLine(HTTPException):
1105 def __init__(self, line):
1106 self.args = line,
1107 self.line = line
1109 # for backwards compatibility
1110 error = HTTPException
1112 class LineAndFileWrapper:
1113 """A limited file-like object for HTTP/0.9 responses."""
1115 # The status-line parsing code calls readline(), which normally
1116 # get the HTTP status line. For a 0.9 response, however, this is
1117 # actually the first line of the body! Clients need to get a
1118 # readable file object that contains that line.
1120 def __init__(self, line, file):
1121 self._line = line
1122 self._file = file
1123 self._line_consumed = 0
1124 self._line_offset = 0
1125 self._line_left = len(line)
1127 def __getattr__(self, attr):
1128 return getattr(self._file, attr)
1130 def _done(self):
1131 # called when the last byte is read from the line. After the
1132 # call, all read methods are delegated to the underlying file
1133 # obhect.
1134 self._line_consumed = 1
1135 self.read = self._file.read
1136 self.readline = self._file.readline
1137 self.readlines = self._file.readlines
1139 def read(self, amt=None):
1140 assert not self._line_consumed and self._line_left
1141 if amt is None or amt > self._line_left:
1142 s = self._line[self._line_offset:]
1143 self._done()
1144 if amt is None:
1145 return s + self._file.read()
1146 else:
1147 return s + self._file.read(amt - len(s))
1148 else:
1149 assert amt <= self._line_left
1150 i = self._line_offset
1151 j = i + amt
1152 s = self._line[i:j]
1153 self._line_offset = j
1154 self._line_left -= amt
1155 if self._line_left == 0:
1156 self._done()
1157 return s
1159 def readline(self):
1160 s = self._line[self._line_offset:]
1161 self._done()
1162 return s
1164 def readlines(self, size=None):
1165 L = [self._line[self._line_offset:]]
1166 self._done()
1167 if size is None:
1168 return L + self._file.readlines()
1169 else:
1170 return L + self._file.readlines(size)
1172 def test():
1173 """Test this module.
1175 A hodge podge of tests collected here, because they have too many
1176 external dependencies for the regular test suite.
1179 import sys
1180 import getopt
1181 opts, args = getopt.getopt(sys.argv[1:], 'd')
1182 dl = 0
1183 for o, a in opts:
1184 if o == '-d': dl = dl + 1
1185 host = 'www.python.org'
1186 selector = '/'
1187 if args[0:]: host = args[0]
1188 if args[1:]: selector = args[1]
1189 h = HTTP()
1190 h.set_debuglevel(dl)
1191 h.connect(host)
1192 h.putrequest('GET', selector)
1193 h.endheaders()
1194 status, reason, headers = h.getreply()
1195 print 'status =', status
1196 print 'reason =', reason
1197 print "read", len(h.getfile().read())
1198 print
1199 if headers:
1200 for header in headers.headers: print header.strip()
1201 print
1203 # minimal test that code to extract host from url works
1204 class HTTP11(HTTP):
1205 _http_vsn = 11
1206 _http_vsn_str = 'HTTP/1.1'
1208 h = HTTP11('www.python.org')
1209 h.putrequest('GET', 'http://www.python.org/~jeremy/')
1210 h.endheaders()
1211 h.getreply()
1212 h.close()
1214 if hasattr(socket, 'ssl'):
1216 for host, selector in (('sourceforge.net', '/projects/python'),
1217 ('dbserv2.theopalgroup.com', '/mediumfile'),
1218 ('dbserv2.theopalgroup.com', '/smallfile'),
1220 print "https://%s%s" % (host, selector)
1221 hs = HTTPS()
1222 hs.set_debuglevel(dl)
1223 hs.connect(host)
1224 hs.putrequest('GET', selector)
1225 hs.endheaders()
1226 status, reason, headers = hs.getreply()
1227 print 'status =', status
1228 print 'reason =', reason
1229 print "read", len(hs.getfile().read())
1230 print
1231 if headers:
1232 for header in headers.headers: print header.strip()
1233 print
1235 # Test a buggy server -- returns garbled status line.
1236 # http://www.yahoo.com/promotions/mom_com97/supermom.html
1237 c = HTTPConnection("promotions.yahoo.com")
1238 c.set_debuglevel(1)
1239 c.connect()
1240 c.request("GET", "/promotions/mom_com97/supermom.html")
1241 r = c.getresponse()
1242 print r.status, r.version
1243 lines = r.read().split("\n")
1244 print "\n".join(lines[:5])
1246 c = HTTPConnection("promotions.yahoo.com", strict=1)
1247 c.set_debuglevel(1)
1248 c.connect()
1249 c.request("GET", "/promotions/mom_com97/supermom.html")
1250 try:
1251 r = c.getresponse()
1252 except BadStatusLine, err:
1253 print "strict mode failed as expected"
1254 print err
1255 else:
1256 print "XXX strict mode should have failed"
1258 for strict in 0, 1:
1259 h = HTTP(strict=strict)
1260 h.connect("promotions.yahoo.com")
1261 h.putrequest('GET', "/promotions/mom_com97/supermom.html")
1262 h.endheaders()
1263 status, reason, headers = h.getreply()
1264 assert (strict and status == -1) or status == 200, (strict, status)
1266 if __name__ == '__main__':
1267 test()