Lib/urllib2.py

   1 """An extensible library for opening URLs using a variety of protocols
   2
   3 The simplest way to use this module is to call the urlopen function,
   4 which accepts a string containing a URL or a Request object (described
   5 below).  It opens the URL and returns the results as file-like
   6 object; the returned object has some extra methods described below.
   7
   8 The OpenerDirector manages a collection of Handler objects that do
   9 all the actual work.  Each Handler implements a particular protocol or
  10 option.  The OpenerDirector is a composite object that invokes the
  11 Handlers needed to open the requested URL.  For example, the
  12 HTTPHandler performs HTTP GET and POST requests and deals with
  13 non-error returns.  The HTTPRedirectHandler automatically deals with
  14 HTTP 301, 302, 303 and 307 redirect errors, and the HTTPDigestAuthHandler
  15 deals with digest authentication.
  16
  17 urlopen(url, data=None) -- basic usage is the same as original
  18 urllib.  pass the url and optionally data to post to an HTTP URL, and
  19 get a file-like object back.  One difference is that you can also pass
  20 a Request instance instead of URL.  Raises a URLError (subclass of
  21 IOError); for HTTP errors, raises an HTTPError, which can also be
  22 treated as a valid response.
  23
  24 build_opener -- function that creates a new OpenerDirector instance.
  25 will install the default handlers.  accepts one or more Handlers as
  26 arguments, either instances or Handler classes that it will
  27 instantiate.  if one of the argument is a subclass of the default
  28 handler, the argument will be installed instead of the default.
  29
  30 install_opener -- installs a new opener as the default opener.
  31
  32 objects of interest:
  33 OpenerDirector --
  34
  35 Request -- an object that encapsulates the state of a request.  the
  36 state can be a simple as the URL.  it can also include extra HTTP
  37 headers, e.g. a User-Agent.
  38
  39 BaseHandler --
  40
  41 exceptions:
  42 URLError-- a subclass of IOError, individual protocols have their own
  43 specific subclass
  44
  45 HTTPError-- also a valid HTTP response, so you can treat an HTTP error
  46 as an exceptional event or valid response
  47
  48 internals:
  49 BaseHandler and parent
  50 _call_chain conventions
  51
  52 Example usage:
  53
  54 import urllib2
  55
  56 # set up authentication info
  57 authinfo = urllib2.HTTPBasicAuthHandler()
  58 authinfo.add_password('realm', 'host', 'username', 'password')
  59
  60 proxy_support = urllib2.ProxyHandler({"http" : "http://ahad-haam:3128"})
  61
  62 # build a new opener that adds authentication and caching FTP handlers
  63 opener = urllib2.build_opener(proxy_support, authinfo, urllib2.CacheFTPHandler)
  64
  65 # install it
  66 urllib2.install_opener(opener)
  67
  68 f = urllib2.urlopen('http://www.python.org/')
  69
  70
  71 """
  72
  73 # XXX issues:
  74 # If an authentication error handler that tries to perform
  75 # authentication for some reason but fails, how should the error be
  76 # signalled?  The client needs to know the HTTP error code.  But if
  77 # the handler knows that the problem was, e.g., that it didn't know
  78 # that hash algo that requested in the challenge, it would be good to
  79 # pass that information along to the client, too.
  80 # ftp errors aren't handled cleanly
  81 # check digest against correct (i.e. non-apache) implementation
  82
  83 # Possible extensions:
  84 # complex proxies  XXX not sure what exactly was meant by this
  85 # abstract factory for opener
  86
  87 import base64
  88 import hashlib
  89 import httplib
  90 import mimetools
  91 import os
  92 import posixpath
  93 import random
  94 import re
  95 import socket
  96 import sys
  97 import time
  98 import urlparse
  99 import bisect
 100
 101 try:
 102     from cStringIO import StringIO
 103 except ImportError:
 104     from StringIO import StringIO
 105
 106 from urllib import (unwrap, unquote, splittype, splithost, quote,
 107      addinfourl, splitport, splitgophertype, splitquery,
 108      splitattr, ftpwrapper, noheaders, splituser, splitpasswd, splitvalue)
 109
 110 # support for FileHandler, proxies via environment variables
 111 from urllib import localhost, url2pathname, getproxies
 112
 113 # used in User-Agent header sent
 114 __version__ = sys.version[:3]
 115
 116 _opener = None
 117 def urlopen(url, data=None):
 118     global _opener
 119     if _opener is None:
 120         _opener = build_opener()
 121     return _opener.open(url, data)
 122
 123 def install_opener(opener):
 124     global _opener
 125     _opener = opener
 126
 127 # do these error classes make sense?
 128 # make sure all of the IOError stuff is overridden.  we just want to be
 129 # subtypes.
 130
 131 class URLError(IOError):
 132     # URLError is a sub-type of IOError, but it doesn't share any of
 133     # the implementation.  need to override __init__ and __str__.
 134     # It sets self.args for compatibility with other EnvironmentError
 135     # subclasses, but args doesn't have the typical format with errno in
 136     # slot 0 and strerror in slot 1.  This may be better than nothing.
 137     def __init__(self, reason):
 138         self.args = reason,
 139         self.reason = reason
 140
 141     def __str__(self):
 142         return '<urlopen error %s>' % self.reason
 143
 144 class HTTPError(URLError, addinfourl):
 145     """Raised when HTTP error occurs, but also acts like non-error return"""
 146     __super_init = addinfourl.__init__
 147
 148     def __init__(self, url, code, msg, hdrs, fp):
 149         self.code = code
 150         self.msg = msg
 151         self.hdrs = hdrs
 152         self.fp = fp
 153         self.filename = url
 154         # The addinfourl classes depend on fp being a valid file
 155         # object.  In some cases, the HTTPError may not have a valid
 156         # file object.  If this happens, the simplest workaround is to
 157         # not initialize the base classes.
 158         if fp is not None:
 159             self.__super_init(fp, hdrs, url)
 160
 161     def __str__(self):
 162         return 'HTTP Error %s: %s' % (self.code, self.msg)
 163
 164 class GopherError(URLError):
 165     pass
 166
 167 # copied from cookielib.py
 168 _cut_port_re = re.compile(r":\d+$")
 169 def request_host(request):
 170     """Return request-host, as defined by RFC 2965.
 171
 172     Variation from RFC: returned value is lowercased, for convenient
 173     comparison.
 174
 175     """
 176     url = request.get_full_url()
 177     host = urlparse.urlparse(url)[1]
 178     if host == "":
 179         host = request.get_header("Host", "")
 180
 181     # remove port, if present
 182     host = _cut_port_re.sub("", host, 1)
 183     return host.lower()
 184
 185 class Request:
 186
 187     def __init__(self, url, data=None, headers={},
 188                  origin_req_host=None, unverifiable=False):
 189         # unwrap('<URL:type://host/path>') --> 'type://host/path'
 190         self.__original = unwrap(url)
 191         self.type = None
 192         # self.__r_type is what's left after doing the splittype
 193         self.host = None
 194         self.port = None
 195         self.data = data
 196         self.headers = {}
 197         for key, value in headers.items():
 198             self.add_header(key, value)
 199         self.unredirected_hdrs = {}
 200         if origin_req_host is None:
 201             origin_req_host = request_host(self)
 202         self.origin_req_host = origin_req_host
 203         self.unverifiable = unverifiable
 204
 205     def __getattr__(self, attr):
 206         # XXX this is a fallback mechanism to guard against these
 207         # methods getting called in a non-standard order.  this may be
 208         # too complicated and/or unnecessary.
 209         # XXX should the __r_XXX attributes be public?
 210         if attr[:12] == '_Request__r_':
 211             name = attr[12:]
 212             if hasattr(Request, 'get_' + name):
 213                 getattr(self, 'get_' + name)()
 214                 return getattr(self, attr)
 215         raise AttributeError, attr
 216
 217     def get_method(self):
 218         if self.has_data():
 219             return "POST"
 220         else:
 221             return "GET"
 222
 223     # XXX these helper methods are lame
 224
 225     def add_data(self, data):
 226         self.data = data
 227
 228     def has_data(self):
 229         return self.data is not None
 230
 231     def get_data(self):
 232         return self.data
 233
 234     def get_full_url(self):
 235         return self.__original
 236
 237     def get_type(self):
 238         if self.type is None:
 239             self.type, self.__r_type = splittype(self.__original)
 240             if self.type is None:
 241                 raise ValueError, "unknown url type: %s" % self.__original
 242         return self.type
 243
 244     def get_host(self):
 245         if self.host is None:
 246             self.host, self.__r_host = splithost(self.__r_type)
 247             if self.host:
 248                 self.host = unquote(self.host)
 249         return self.host
 250
 251     def get_selector(self):
 252         return self.__r_host
 253
 254     def set_proxy(self, host, type):
 255         self.host, self.type = host, type
 256         self.__r_host = self.__original
 257
 258     def get_origin_req_host(self):
 259         return self.origin_req_host
 260
 261     def is_unverifiable(self):
 262         return self.unverifiable
 263
 264     def add_header(self, key, val):
 265         # useful for something like authentication
 266         self.headers[key.capitalize()] = val
 267
 268     def add_unredirected_header(self, key, val):
 269         # will not be added to a redirected request
 270         self.unredirected_hdrs[key.capitalize()] = val
 271
 272     def has_header(self, header_name):
 273         return (header_name in self.headers or
 274                 header_name in self.unredirected_hdrs)
 275
 276     def get_header(self, header_name, default=None):
 277         return self.headers.get(
 278             header_name,
 279             self.unredirected_hdrs.get(header_name, default))
 280
 281     def header_items(self):
 282         hdrs = self.unredirected_hdrs.copy()
 283         hdrs.update(self.headers)
 284         return hdrs.items()
 285
 286 class OpenerDirector:
 287     def __init__(self):
 288         client_version = "Python-urllib/%s" % __version__
 289         self.addheaders = [('User-agent', client_version)]
 290         # manage the individual handlers
 291         self.handlers = []
 292         self.handle_open = {}
 293         self.handle_error = {}
 294         self.process_response = {}
 295         self.process_request = {}
 296
 297     def add_handler(self, handler):
 298         added = False
 299         for meth in dir(handler):
 300             if meth in ["redirect_request", "do_open", "proxy_open"]:
 301                 # oops, coincidental match
 302                 continue
 303
 304             i = meth.find("_")
 305             protocol = meth[:i]
 306             condition = meth[i+1:]
 307
 308             if condition.startswith("error"):
 309                 j = condition.find("_") + i + 1
 310                 kind = meth[j+1:]
 311                 try:
 312                     kind = int(kind)
 313                 except ValueError:
 314                     pass
 315                 lookup = self.handle_error.get(protocol, {})
 316                 self.handle_error[protocol] = lookup
 317             elif condition == "open":
 318                 kind = protocol
 319                 lookup = self.handle_open
 320             elif condition == "response":
 321                 kind = protocol
 322                 lookup = self.process_response
 323             elif condition == "request":
 324                 kind = protocol
 325                 lookup = self.process_request
 326             else:
 327                 continue
 328
 329             handlers = lookup.setdefault(kind, [])
 330             if handlers:
 331                 bisect.insort(handlers, handler)
 332             else:
 333                 handlers.append(handler)
 334             added = True
 335
 336         if added:
 337             # XXX why does self.handlers need to be sorted?
 338             bisect.insort(self.handlers, handler)
 339             handler.add_parent(self)
 340
 341     def close(self):
 342         # Only exists for backwards compatibility.
 343         pass
 344
 345     def _call_chain(self, chain, kind, meth_name, *args):
 346         # Handlers raise an exception if no one else should try to handle
 347         # the request, or return None if they can't but another handler
 348         # could.  Otherwise, they return the response.
 349         handlers = chain.get(kind, ())
 350         for handler in handlers:
 351             func = getattr(handler, meth_name)
 352
 353             result = func(*args)
 354             if result is not None:
 355                 return result
 356
 357     def open(self, fullurl, data=None):
 358         # accept a URL or a Request object
 359         if isinstance(fullurl, basestring):
 360             req = Request(fullurl, data)
 361         else:
 362             req = fullurl
 363             if data is not None:
 364                 req.add_data(data)
 365
 366         protocol = req.get_type()
 367
 368         # pre-process request
 369         meth_name = protocol+"_request"
 370         for processor in self.process_request.get(protocol, []):
 371             meth = getattr(processor, meth_name)
 372             req = meth(req)
 373
 374         response = self._open(req, data)
 375
 376         # post-process response
 377         meth_name = protocol+"_response"
 378         for processor in self.process_response.get(protocol, []):
 379             meth = getattr(processor, meth_name)
 380             response = meth(req, response)
 381
 382         return response
 383
 384     def _open(self, req, data=None):
 385         result = self._call_chain(self.handle_open, 'default',
 386                                   'default_open', req)
 387         if result:
 388             return result
 389
 390         protocol = req.get_type()
 391         result = self._call_chain(self.handle_open, protocol, protocol +
 392                                   '_open', req)
 393         if result:
 394             return result
 395
 396         return self._call_chain(self.handle_open, 'unknown',
 397                                 'unknown_open', req)
 398
 399     def error(self, proto, *args):
 400         if proto in ('http', 'https'):
 401             # XXX http[s] protocols are special-cased
 402             dict = self.handle_error['http'] # https is not different than http
 403             proto = args[2]  # YUCK!
 404             meth_name = 'http_error_%s' % proto
 405             http_err = 1
 406             orig_args = args
 407         else:
 408             dict = self.handle_error
 409             meth_name = proto + '_error'
 410             http_err = 0
 411         args = (dict, proto, meth_name) + args
 412         result = self._call_chain(*args)
 413         if result:
 414             return result
 415
 416         if http_err:
 417             args = (dict, 'default', 'http_error_default') + orig_args
 418             return self._call_chain(*args)
 419
 420 # XXX probably also want an abstract factory that knows when it makes
 421 # sense to skip a superclass in favor of a subclass and when it might
 422 # make sense to include both
 423
 424 def build_opener(*handlers):
 425     """Create an opener object from a list of handlers.
 426
 427     The opener will use several default handlers, including support
 428     for HTTP and FTP.
 429
 430     If any of the handlers passed as arguments are subclasses of the
 431     default handlers, the default handlers will not be used.
 432     """
 433     import types
 434     def isclass(obj):
 435         return isinstance(obj, types.ClassType) or hasattr(obj, "__bases__")
 436
 437     opener = OpenerDirector()
 438     default_classes = [ProxyHandler, UnknownHandler, HTTPHandler,
 439                        HTTPDefaultErrorHandler, HTTPRedirectHandler,
 440                        FTPHandler, FileHandler, HTTPErrorProcessor]
 441     if hasattr(httplib, 'HTTPS'):
 442         default_classes.append(HTTPSHandler)
 443     skip = []
 444     for klass in default_classes:
 445         for check in handlers:
 446             if isclass(check):
 447                 if issubclass(check, klass):
 448                     skip.append(klass)
 449             elif isinstance(check, klass):
 450                 skip.append(klass)
 451     for klass in skip:
 452         default_classes.remove(klass)
 453
 454     for klass in default_classes:
 455         opener.add_handler(klass())
 456
 457     for h in handlers:
 458         if isclass(h):
 459             h = h()
 460         opener.add_handler(h)
 461     return opener
 462
 463 class BaseHandler:
 464     handler_order = 500
 465
 466     def add_parent(self, parent):
 467         self.parent = parent
 468
 469     def close(self):
 470         # Only exists for backwards compatibility
 471         pass
 472
 473     def __lt__(self, other):
 474         if not hasattr(other, "handler_order"):
 475             # Try to preserve the old behavior of having custom classes
 476             # inserted after default ones (works only for custom user
 477             # classes which are not aware of handler_order).
 478             return True
 479         return self.handler_order < other.handler_order
 480
 481
 482 class HTTPErrorProcessor(BaseHandler):
 483     """Process HTTP error responses."""
 484     handler_order = 1000  # after all other processing
 485
 486     def http_response(self, request, response):
 487         code, msg, hdrs = response.code, response.msg, response.info()
 488
 489         if code not in (200, 206):
 490             response = self.parent.error(
 491                 'http', request, response, code, msg, hdrs)
 492
 493         return response
 494
 495     https_response = http_response
 496
 497 class HTTPDefaultErrorHandler(BaseHandler):
 498     def http_error_default(self, req, fp, code, msg, hdrs):
 499         raise HTTPError(req.get_full_url(), code, msg, hdrs, fp)
 500
 501 class HTTPRedirectHandler(BaseHandler):
 502     # maximum number of redirections to any single URL
 503     # this is needed because of the state that cookies introduce
 504     max_repeats = 4
 505     # maximum total number of redirections (regardless of URL) before
 506     # assuming we're in a loop
 507     max_redirections = 10
 508
 509     def redirect_request(self, req, fp, code, msg, headers, newurl):
 510         """Return a Request or None in response to a redirect.
 511
 512         This is called by the http_error_30x methods when a
 513         redirection response is received.  If a redirection should
 514         take place, return a new Request to allow http_error_30x to
 515         perform the redirect.  Otherwise, raise HTTPError if no-one
 516         else should try to handle this url.  Return None if you can't
 517         but another Handler might.
 518         """
 519         m = req.get_method()
 520         if (code in (301, 302, 303, 307) and m in ("GET", "HEAD")
 521             or code in (301, 302, 303) and m == "POST"):
 522             # Strictly (according to RFC 2616), 301 or 302 in response
 523             # to a POST MUST NOT cause a redirection without confirmation
 524             # from the user (of urllib2, in this case).  In practice,
 525             # essentially all clients do redirect in this case, so we
 526             # do the same.
 527             # be conciliant with URIs containing a space
 528             newurl = newurl.replace(' ', '%20')
 529             return Request(newurl,
 530                            headers=req.headers,
 531                            origin_req_host=req.get_origin_req_host(),
 532                            unverifiable=True)
 533         else:
 534             raise HTTPError(req.get_full_url(), code, msg, headers, fp)
 535
 536     # Implementation note: To avoid the server sending us into an
 537     # infinite loop, the request object needs to track what URLs we
 538     # have already seen.  Do this by adding a handler-specific
 539     # attribute to the Request object.
 540     def http_error_302(self, req, fp, code, msg, headers):
 541         # Some servers (incorrectly) return multiple Location headers
 542         # (so probably same goes for URI).  Use first header.
 543         if 'location' in headers:
 544             newurl = headers.getheaders('location')[0]
 545         elif 'uri' in headers:
 546             newurl = headers.getheaders('uri')[0]
 547         else:
 548             return
 549         newurl = urlparse.urljoin(req.get_full_url(), newurl)
 550
 551         # XXX Probably want to forget about the state of the current
 552         # request, although that might interact poorly with other
 553         # handlers that also use handler-specific request attributes
 554         new = self.redirect_request(req, fp, code, msg, headers, newurl)
 555         if new is None:
 556             return
 557
 558         # loop detection
 559         # .redirect_dict has a key url if url was previously visited.
 560         if hasattr(req, 'redirect_dict'):
 561             visited = new.redirect_dict = req.redirect_dict
 562             if (visited.get(newurl, 0) >= self.max_repeats or
 563                 len(visited) >= self.max_redirections):
 564                 raise HTTPError(req.get_full_url(), code,
 565                                 self.inf_msg + msg, headers, fp)
 566         else:
 567             visited = new.redirect_dict = req.redirect_dict = {}
 568         visited[newurl] = visited.get(newurl, 0) + 1
 569
 570         # Don't close the fp until we are sure that we won't use it
 571         # with HTTPError.
 572         fp.read()
 573         fp.close()
 574
 575         return self.parent.open(new)
 576
 577     http_error_301 = http_error_303 = http_error_307 = http_error_302
 578
 579     inf_msg = "The HTTP server returned a redirect error that would " \
 580               "lead to an infinite loop.\n" \
 581               "The last 30x error message was:\n"
 582
 583
 584 def _parse_proxy(proxy):
 585     """Return (scheme, user, password, host/port) given a URL or an authority.
 586
 587     If a URL is supplied, it must have an authority (host:port) component.
 588     According to RFC 3986, having an authority component means the URL must
 589     have two slashes after the scheme:
 590
 591     >>> _parse_proxy('file:/ftp.example.com/')
 592     Traceback (most recent call last):
 593     ValueError: proxy URL with no authority: 'file:/ftp.example.com/'
 594
 595     The first three items of the returned tuple may be None.
 596
 597     Examples of authority parsing:
 598
 599     >>> _parse_proxy('proxy.example.com')
 600     (None, None, None, 'proxy.example.com')
 601     >>> _parse_proxy('proxy.example.com:3128')
 602     (None, None, None, 'proxy.example.com:3128')
 603
 604     The authority component may optionally include userinfo (assumed to be
 605     username:password):
 606
 607     >>> _parse_proxy('joe:password@proxy.example.com')
 608     (None, 'joe', 'password', 'proxy.example.com')
 609     >>> _parse_proxy('joe:password@proxy.example.com:3128')
 610     (None, 'joe', 'password', 'proxy.example.com:3128')
 611
 612     Same examples, but with URLs instead:
 613
 614     >>> _parse_proxy('http://proxy.example.com/')
 615     ('http', None, None, 'proxy.example.com')
 616     >>> _parse_proxy('http://proxy.example.com:3128/')
 617     ('http', None, None, 'proxy.example.com:3128')
 618     >>> _parse_proxy('http://joe:password@proxy.example.com/')
 619     ('http', 'joe', 'password', 'proxy.example.com')
 620     >>> _parse_proxy('http://joe:password@proxy.example.com:3128')
 621     ('http', 'joe', 'password', 'proxy.example.com:3128')
 622
 623     Everything after the authority is ignored:
 624
 625     >>> _parse_proxy('ftp://joe:password@proxy.example.com/rubbish:3128')
 626     ('ftp', 'joe', 'password', 'proxy.example.com')
 627
 628     Test for no trailing '/' case:
 629
 630     >>> _parse_proxy('http://joe:password@proxy.example.com')
 631     ('http', 'joe', 'password', 'proxy.example.com')
 632
 633     """
 634     scheme, r_scheme = splittype(proxy)
 635     if not r_scheme.startswith("/"):
 636         # authority
 637         scheme = None
 638         authority = proxy
 639     else:
 640         # URL
 641         if not r_scheme.startswith("//"):
 642             raise ValueError("proxy URL with no authority: %r" % proxy)
 643         # We have an authority, so for RFC 3986-compliant URLs (by ss 3.
 644         # and 3.3.), path is empty or starts with '/'
 645         end = r_scheme.find("/", 2)
 646         if end == -1:
 647             end = None
 648         authority = r_scheme[2:end]
 649     userinfo, hostport = splituser(authority)
 650     if userinfo is not None:
 651         user, password = splitpasswd(userinfo)
 652     else:
 653         user = password = None
 654     return scheme, user, password, hostport
 655
 656 class ProxyHandler(BaseHandler):
 657     # Proxies must be in front
 658     handler_order = 100
 659
 660     def __init__(self, proxies=None):
 661         if proxies is None:
 662             proxies = getproxies()
 663         assert hasattr(proxies, 'has_key'), "proxies must be a mapping"
 664         self.proxies = proxies
 665         for type, url in proxies.items():
 666             setattr(self, '%s_open' % type,
 667                     lambda r, proxy=url, type=type, meth=self.proxy_open: \
 668                     meth(r, proxy, type))
 669
 670     def proxy_open(self, req, proxy, type):
 671         orig_type = req.get_type()
 672         proxy_type, user, password, hostport = _parse_proxy(proxy)
 673         if proxy_type is None:
 674             proxy_type = orig_type
 675         if user and password:
 676             user_pass = '%s:%s' % (unquote(user), unquote(password))
 677             creds = base64.encodestring(user_pass).strip()
 678             req.add_header('Proxy-authorization', 'Basic ' + creds)
 679         hostport = unquote(hostport)
 680         req.set_proxy(hostport, proxy_type)
 681         if orig_type == proxy_type:
 682             # let other handlers take care of it
 683             return None
 684         else:
 685             # need to start over, because the other handlers don't
 686             # grok the proxy's URL type
 687             # e.g. if we have a constructor arg proxies like so:
 688             # {'http': 'ftp://proxy.example.com'}, we may end up turning
 689             # a request for http://acme.example.com/a into one for
 690             # ftp://proxy.example.com/a
 691             return self.parent.open(req)
 692
 693 class HTTPPasswordMgr:
 694
 695     def __init__(self):
 696         self.passwd = {}
 697
 698     def add_password(self, realm, uri, user, passwd):
 699         # uri could be a single URI or a sequence
 700         if isinstance(uri, basestring):
 701             uri = [uri]
 702         if not realm in self.passwd:
 703             self.passwd[realm] = {}
 704         for default_port in True, False:
 705             reduced_uri = tuple(
 706                 [self.reduce_uri(u, default_port) for u in uri])
 707             self.passwd[realm][reduced_uri] = (user, passwd)
 708
 709     def find_user_password(self, realm, authuri):
 710         domains = self.passwd.get(realm, {})
 711         for default_port in True, False:
 712             reduced_authuri = self.reduce_uri(authuri, default_port)
 713             for uris, authinfo in domains.iteritems():
 714                 for uri in uris:
 715                     if self.is_suburi(uri, reduced_authuri):
 716                         return authinfo
 717         return None, None
 718
 719     def reduce_uri(self, uri, default_port=True):
 720         """Accept authority or URI and extract only the authority and path."""
 721         # note HTTP URLs do not have a userinfo component
 722         parts = urlparse.urlsplit(uri)
 723         if parts[1]:
 724             # URI
 725             scheme = parts[0]
 726             authority = parts[1]
 727             path = parts[2] or '/'
 728         else:
 729             # host or host:port
 730             scheme = None
 731             authority = uri
 732             path = '/'
 733         host, port = splitport(authority)
 734         if default_port and port is None and scheme is not None:
 735             dport = {"http": 80,
 736                      "https": 443,
 737                      }.get(scheme)
 738             if dport is not None:
 739                 authority = "%s:%d" % (host, dport)
 740         return authority, path
 741
 742     def is_suburi(self, base, test):
 743         """Check if test is below base in a URI tree
 744
 745         Both args must be URIs in reduced form.
 746         """
 747         if base == test:
 748             return True
 749         if base[0] != test[0]:
 750             return False
 751         common = posixpath.commonprefix((base[1], test[1]))
 752         if len(common) == len(base[1]):
 753             return True
 754         return False
 755
 756
 757 class HTTPPasswordMgrWithDefaultRealm(HTTPPasswordMgr):
 758
 759     def find_user_password(self, realm, authuri):
 760         user, password = HTTPPasswordMgr.find_user_password(self, realm,
 761                                                             authuri)
 762         if user is not None:
 763             return user, password
 764         return HTTPPasswordMgr.find_user_password(self, None, authuri)
 765
 766
 767 class AbstractBasicAuthHandler:
 768
 769     rx = re.compile('[ \t]*([^ \t]+)[ \t]+realm="([^"]*)"', re.I)
 770
 771     # XXX there can actually be multiple auth-schemes in a
 772     # www-authenticate header.  should probably be a lot more careful
 773     # in parsing them to extract multiple alternatives
 774
 775     # XXX could pre-emptively send auth info already accepted (RFC 2617,
 776     # end of section 2, and section 1.2 immediately after "credentials"
 777     # production).
 778
 779     def __init__(self, password_mgr=None):
 780         if password_mgr is None:
 781             password_mgr = HTTPPasswordMgr()
 782         self.passwd = password_mgr
 783         self.add_password = self.passwd.add_password
 784
 785     def http_error_auth_reqed(self, authreq, host, req, headers):
 786         # host may be an authority (without userinfo) or a URL with an
 787         # authority
 788         # XXX could be multiple headers
 789         authreq = headers.get(authreq, None)
 790         if authreq:
 791             mo = AbstractBasicAuthHandler.rx.search(authreq)
 792             if mo:
 793                 scheme, realm = mo.groups()
 794                 if scheme.lower() == 'basic':
 795                     return self.retry_http_basic_auth(host, req, realm)
 796
 797     def retry_http_basic_auth(self, host, req, realm):
 798         user, pw = self.passwd.find_user_password(realm, host)
 799         if pw is not None:
 800             raw = "%s:%s" % (user, pw)
 801             auth = 'Basic %s' % base64.encodestring(raw).strip()
 802             if req.headers.get(self.auth_header, None) == auth:
 803                 return None
 804             req.add_header(self.auth_header, auth)
 805             return self.parent.open(req)
 806         else:
 807             return None
 808
 809
 810 class HTTPBasicAuthHandler(AbstractBasicAuthHandler, BaseHandler):
 811
 812     auth_header = 'Authorization'
 813
 814     def http_error_401(self, req, fp, code, msg, headers):
 815         url = req.get_full_url()
 816         return self.http_error_auth_reqed('www-authenticate',
 817                                           url, req, headers)
 818
 819
 820 class ProxyBasicAuthHandler(AbstractBasicAuthHandler, BaseHandler):
 821
 822     auth_header = 'Proxy-authorization'
 823
 824     def http_error_407(self, req, fp, code, msg, headers):
 825         # http_error_auth_reqed requires that there is no userinfo component in
 826         # authority.  Assume there isn't one, since urllib2 does not (and
 827         # should not, RFC 3986 s. 3.2.1) support requests for URLs containing
 828         # userinfo.
 829         authority = req.get_host()
 830         return self.http_error_auth_reqed('proxy-authenticate',
 831                                           authority, req, headers)
 832
 833
 834 def randombytes(n):
 835     """Return n random bytes."""
 836     # Use /dev/urandom if it is available.  Fall back to random module
 837     # if not.  It might be worthwhile to extend this function to use
 838     # other platform-specific mechanisms for getting random bytes.
 839     if os.path.exists("/dev/urandom"):
 840         f = open("/dev/urandom")
 841         s = f.read(n)
 842         f.close()
 843         return s
 844     else:
 845         L = [chr(random.randrange(0, 256)) for i in range(n)]
 846         return "".join(L)
 847
 848 class AbstractDigestAuthHandler:
 849     # Digest authentication is specified in RFC 2617.
 850
 851     # XXX The client does not inspect the Authentication-Info header
 852     # in a successful response.
 853
 854     # XXX It should be possible to test this implementation against
 855     # a mock server that just generates a static set of challenges.
 856
 857     # XXX qop="auth-int" supports is shaky
 858
 859     def __init__(self, passwd=None):
 860         if passwd is None:
 861             passwd = HTTPPasswordMgr()
 862         self.passwd = passwd
 863         self.add_password = self.passwd.add_password
 864         self.retried = 0
 865         self.nonce_count = 0
 866
 867     def reset_retry_count(self):
 868         self.retried = 0
 869
 870     def http_error_auth_reqed(self, auth_header, host, req, headers):
 871         authreq = headers.get(auth_header, None)
 872         if self.retried > 5:
 873             # Don't fail endlessly - if we failed once, we'll probably
 874             # fail a second time. Hm. Unless the Password Manager is
 875             # prompting for the information. Crap. This isn't great
 876             # but it's better than the current 'repeat until recursion
 877             # depth exceeded' approach <wink>
 878             raise HTTPError(req.get_full_url(), 401, "digest auth failed",
 879                             headers, None)
 880         else:
 881             self.retried += 1
 882         if authreq:
 883             scheme = authreq.split()[0]
 884             if scheme.lower() == 'digest':
 885                 return self.retry_http_digest_auth(req, authreq)
 886
 887     def retry_http_digest_auth(self, req, auth):
 888         token, challenge = auth.split(' ', 1)
 889         chal = parse_keqv_list(parse_http_list(challenge))
 890         auth = self.get_authorization(req, chal)
 891         if auth:
 892             auth_val = 'Digest %s' % auth
 893             if req.headers.get(self.auth_header, None) == auth_val:
 894                 return None
 895             req.add_unredirected_header(self.auth_header, auth_val)
 896             resp = self.parent.open(req)
 897             return resp
 898
 899     def get_cnonce(self, nonce):
 900         # The cnonce-value is an opaque
 901         # quoted string value provided by the client and used by both client
 902         # and server to avoid chosen plaintext attacks, to provide mutual
 903         # authentication, and to provide some message integrity protection.
 904         # This isn't a fabulous effort, but it's probably Good Enough.
 905         dig = hashlib.sha1("%s:%s:%s:%s" % (self.nonce_count, nonce, time.ctime(),
 906                                             randombytes(8))).hexdigest()
 907         return dig[:16]
 908
 909     def get_authorization(self, req, chal):
 910         try:
 911             realm = chal['realm']
 912             nonce = chal['nonce']
 913             qop = chal.get('qop')
 914             algorithm = chal.get('algorithm', 'MD5')
 915             # mod_digest doesn't send an opaque, even though it isn't
 916             # supposed to be optional
 917             opaque = chal.get('opaque', None)
 918         except KeyError:
 919             return None
 920
 921         H, KD = self.get_algorithm_impls(algorithm)
 922         if H is None:
 923             return None
 924
 925         user, pw = self.passwd.find_user_password(realm, req.get_full_url())
 926         if user is None:
 927             return None
 928
 929         # XXX not implemented yet
 930         if req.has_data():
 931             entdig = self.get_entity_digest(req.get_data(), chal)
 932         else:
 933             entdig = None
 934
 935         A1 = "%s:%s:%s" % (user, realm, pw)
 936         A2 = "%s:%s" % (req.get_method(),
 937                         # XXX selector: what about proxies and full urls
 938                         req.get_selector())
 939         if qop == 'auth':
 940             self.nonce_count += 1
 941             ncvalue = '%08x' % self.nonce_count
 942             cnonce = self.get_cnonce(nonce)
 943             noncebit = "%s:%s:%s:%s:%s" % (nonce, ncvalue, cnonce, qop, H(A2))
 944             respdig = KD(H(A1), noncebit)
 945         elif qop is None:
 946             respdig = KD(H(A1), "%s:%s" % (nonce, H(A2)))
 947         else:
 948             # XXX handle auth-int.
 949             pass
 950
 951         # XXX should the partial digests be encoded too?
 952
 953         base = 'username="%s", realm="%s", nonce="%s", uri="%s", ' \
 954                'response="%s"' % (user, realm, nonce, req.get_selector(),
 955                                   respdig)
 956         if opaque:
 957             base += ', opaque="%s"' % opaque
 958         if entdig:
 959             base += ', digest="%s"' % entdig
 960         base += ', algorithm="%s"' % algorithm
 961         if qop:
 962             base += ', qop=auth, nc=%s, cnonce="%s"' % (ncvalue, cnonce)
 963         return base
 964
 965     def get_algorithm_impls(self, algorithm):
 966         # lambdas assume digest modules are imported at the top level
 967         if algorithm == 'MD5':
 968             H = lambda x: hashlib.md5(x).hexdigest()
 969         elif algorithm == 'SHA':
 970             H = lambda x: hashlib.sha1(x).hexdigest()
 971         # XXX MD5-sess
 972         KD = lambda s, d: H("%s:%s" % (s, d))
 973         return H, KD
 974
 975     def get_entity_digest(self, data, chal):
 976         # XXX not implemented yet
 977         return None
 978
 979
 980 class HTTPDigestAuthHandler(BaseHandler, AbstractDigestAuthHandler):
 981     """An authentication protocol defined by RFC 2069
 982
 983     Digest authentication improves on basic authentication because it
 984     does not transmit passwords in the clear.
 985     """
 986
 987     auth_header = 'Authorization'
 988     handler_order = 490  # before Basic auth
 989
 990     def http_error_401(self, req, fp, code, msg, headers):
 991         host = urlparse.urlparse(req.get_full_url())[1]
 992         retry = self.http_error_auth_reqed('www-authenticate',
 993                                            host, req, headers)
 994         self.reset_retry_count()
 995         return retry
 996
 997
 998 class ProxyDigestAuthHandler(BaseHandler, AbstractDigestAuthHandler):
 999
1000     auth_header = 'Proxy-Authorization'
1001     handler_order = 490  # before Basic auth
1002
1003     def http_error_407(self, req, fp, code, msg, headers):
1004         host = req.get_host()
1005         retry = self.http_error_auth_reqed('proxy-authenticate',
1006                                            host, req, headers)
1007         self.reset_retry_count()
1008         return retry
1009
1010 class AbstractHTTPHandler(BaseHandler):
1011
1012     def __init__(self, debuglevel=0):
1013         self._debuglevel = debuglevel
1014
1015     def set_http_debuglevel(self, level):
1016         self._debuglevel = level
1017
1018     def do_request_(self, request):
1019         host = request.get_host()
1020         if not host:
1021             raise URLError('no host given')
1022
1023         if request.has_data():  # POST
1024             data = request.get_data()
1025             if not request.has_header('Content-type'):
1026                 request.add_unredirected_header(
1027                     'Content-type',
1028                     'application/x-www-form-urlencoded')
1029             if not request.has_header('Content-length'):
1030                 request.add_unredirected_header(
1031                     'Content-length', '%d' % len(data))
1032
1033         scheme, sel = splittype(request.get_selector())
1034         sel_host, sel_path = splithost(sel)
1035         if not request.has_header('Host'):
1036             request.add_unredirected_header('Host', sel_host or host)
1037         for name, value in self.parent.addheaders:
1038             name = name.capitalize()
1039             if not request.has_header(name):
1040                 request.add_unredirected_header(name, value)
1041
1042         return request
1043
1044     def do_open(self, http_class, req):
1045         """Return an addinfourl object for the request, using http_class.
1046
1047         http_class must implement the HTTPConnection API from httplib.
1048         The addinfourl return value is a file-like object.  It also
1049         has methods and attributes including:
1050             - info(): return a mimetools.Message object for the headers
1051             - geturl(): return the original request URL
1052             - code: HTTP status code
1053         """
1054         host = req.get_host()
1055         if not host:
1056             raise URLError('no host given')
1057
1058         h = http_class(host) # will parse host:port
1059         h.set_debuglevel(self._debuglevel)
1060
1061         headers = dict(req.headers)
1062         headers.update(req.unredirected_hdrs)
1063         # We want to make an HTTP/1.1 request, but the addinfourl
1064         # class isn't prepared to deal with a persistent connection.
1065         # It will try to read all remaining data from the socket,
1066         # which will block while the server waits for the next request.
1067         # So make sure the connection gets closed after the (only)
1068         # request.
1069         headers["Connection"] = "close"
1070         try:
1071             h.request(req.get_method(), req.get_selector(), req.data, headers)
1072             r = h.getresponse()
1073         except socket.error, err: # XXX what error?
1074             raise URLError(err)
1075
1076         # Pick apart the HTTPResponse object to get the addinfourl
1077         # object initialized properly.
1078
1079         # Wrap the HTTPResponse object in socket's file object adapter
1080         # for Windows.  That adapter calls recv(), so delegate recv()
1081         # to read().  This weird wrapping allows the returned object to
1082         # have readline() and readlines() methods.
1083
1084         # XXX It might be better to extract the read buffering code
1085         # out of socket._fileobject() and into a base class.
1086
1087         r.recv = r.read
1088         fp = socket._fileobject(r)
1089
1090         resp = addinfourl(fp, r.msg, req.get_full_url())
1091         resp.code = r.status
1092         resp.msg = r.reason
1093         return resp
1094
1095
1096 class HTTPHandler(AbstractHTTPHandler):
1097
1098     def http_open(self, req):
1099         return self.do_open(httplib.HTTPConnection, req)
1100
1101     http_request = AbstractHTTPHandler.do_request_
1102
1103 if hasattr(httplib, 'HTTPS'):
1104     class HTTPSHandler(AbstractHTTPHandler):
1105
1106         def https_open(self, req):
1107             return self.do_open(httplib.HTTPSConnection, req)
1108
1109         https_request = AbstractHTTPHandler.do_request_
1110
1111 class HTTPCookieProcessor(BaseHandler):
1112     def __init__(self, cookiejar=None):
1113         import cookielib
1114         if cookiejar is None:
1115             cookiejar = cookielib.CookieJar()
1116         self.cookiejar = cookiejar
1117
1118     def http_request(self, request):
1119         self.cookiejar.add_cookie_header(request)
1120         return request
1121
1122     def http_response(self, request, response):
1123         self.cookiejar.extract_cookies(response, request)
1124         return response
1125
1126     https_request = http_request
1127     https_response = http_response
1128
1129 class UnknownHandler(BaseHandler):
1130     def unknown_open(self, req):
1131         type = req.get_type()
1132         raise URLError('unknown url type: %s' % type)
1133
1134 def parse_keqv_list(l):
1135     """Parse list of key=value strings where keys are not duplicated."""
1136     parsed = {}
1137     for elt in l:
1138         k, v = elt.split('=', 1)
1139         if v[0] == '"' and v[-1] == '"':
1140             v = v[1:-1]
1141         parsed[k] = v
1142     return parsed
1143
1144 def parse_http_list(s):
1145     """Parse lists as described by RFC 2068 Section 2.
1146
1147     In particular, parse comma-separated lists where the elements of
1148     the list may include quoted-strings.  A quoted-string could
1149     contain a comma.  A non-quoted string could have quotes in the
1150     middle.  Neither commas nor quotes count if they are escaped.
1151     Only double-quotes count, not single-quotes.
1152     """
1153     res = []
1154     part = ''
1155
1156     escape = quote = False
1157     for cur in s:
1158         if escape:
1159             part += cur
1160             escape = False
1161             continue
1162         if quote:
1163             if cur == '\\':
1164                 escape = True
1165                 continue
1166             elif cur == '"':
1167                 quote = False
1168             part += cur
1169             continue
1170
1171         if cur == ',':
1172             res.append(part)
1173             part = ''
1174             continue
1175
1176         if cur == '"':
1177             quote = True
1178
1179         part += cur
1180
1181     # append last part
1182     if part:
1183         res.append(part)
1184
1185     return [part.strip() for part in res]
1186
1187 class FileHandler(BaseHandler):
1188     # Use local file or FTP depending on form of URL
1189     def file_open(self, req):
1190         url = req.get_selector()
1191         if url[:2] == '//' and url[2:3] != '/':
1192             req.type = 'ftp'
1193             return self.parent.open(req)
1194         else:
1195             return self.open_local_file(req)
1196
1197     # names for the localhost
1198     names = None
1199     def get_names(self):
1200         if FileHandler.names is None:
1201             try:
1202                 FileHandler.names = (socket.gethostbyname('localhost'),
1203                                     socket.gethostbyname(socket.gethostname()))
1204             except socket.gaierror:
1205                 FileHandler.names = (socket.gethostbyname('localhost'),)
1206         return FileHandler.names
1207
1208     # not entirely sure what the rules are here
1209     def open_local_file(self, req):
1210         import email.Utils
1211         import mimetypes
1212         host = req.get_host()
1213         file = req.get_selector()
1214         localfile = url2pathname(file)
1215         stats = os.stat(localfile)
1216         size = stats.st_size
1217         modified = email.Utils.formatdate(stats.st_mtime, usegmt=True)
1218         mtype = mimetypes.guess_type(file)[0]
1219         headers = mimetools.Message(StringIO(
1220             'Content-type: %s\nContent-length: %d\nLast-modified: %s\n' %
1221             (mtype or 'text/plain', size, modified)))
1222         if host:
1223             host, port = splitport(host)
1224         if not host or \
1225            (not port and socket.gethostbyname(host) in self.get_names()):
1226             return addinfourl(open(localfile, 'rb'),
1227                               headers, 'file:'+file)
1228         raise URLError('file not on local host')
1229
1230 class FTPHandler(BaseHandler):
1231     def ftp_open(self, req):
1232         import ftplib
1233         import mimetypes
1234         host = req.get_host()
1235         if not host:
1236             raise IOError, ('ftp error', 'no host given')
1237         host, port = splitport(host)
1238         if port is None:
1239             port = ftplib.FTP_PORT
1240         else:
1241             port = int(port)
1242
1243         # username/password handling
1244         user, host = splituser(host)
1245         if user:
1246             user, passwd = splitpasswd(user)
1247         else:
1248             passwd = None
1249         host = unquote(host)
1250         user = unquote(user or '')
1251         passwd = unquote(passwd or '')
1252
1253         try:
1254             host = socket.gethostbyname(host)
1255         except socket.error, msg:
1256             raise URLError(msg)
1257         path, attrs = splitattr(req.get_selector())
1258         dirs = path.split('/')
1259         dirs = map(unquote, dirs)
1260         dirs, file = dirs[:-1], dirs[-1]
1261         if dirs and not dirs[0]:
1262             dirs = dirs[1:]
1263         try:
1264             fw = self.connect_ftp(user, passwd, host, port, dirs)
1265             type = file and 'I' or 'D'
1266             for attr in attrs:
1267                 attr, value = splitvalue(attr)
1268                 if attr.lower() == 'type' and \
1269                    value in ('a', 'A', 'i', 'I', 'd', 'D'):
1270                     type = value.upper()
1271             fp, retrlen = fw.retrfile(file, type)
1272             headers = ""
1273             mtype = mimetypes.guess_type(req.get_full_url())[0]
1274             if mtype:
1275                 headers += "Content-type: %s\n" % mtype
1276             if retrlen is not None and retrlen >= 0:
1277                 headers += "Content-length: %d\n" % retrlen
1278             sf = StringIO(headers)
1279             headers = mimetools.Message(sf)
1280             return addinfourl(fp, headers, req.get_full_url())
1281         except ftplib.all_errors, msg:
1282             raise IOError, ('ftp error', msg), sys.exc_info()[2]
1283
1284     def connect_ftp(self, user, passwd, host, port, dirs):
1285         fw = ftpwrapper(user, passwd, host, port, dirs)
1286 ##        fw.ftp.set_debuglevel(1)
1287         return fw
1288
1289 class CacheFTPHandler(FTPHandler):
1290     # XXX would be nice to have pluggable cache strategies
1291     # XXX this stuff is definitely not thread safe
1292     def __init__(self):
1293         self.cache = {}
1294         self.timeout = {}
1295         self.soonest = 0
1296         self.delay = 60
1297         self.max_conns = 16
1298
1299     def setTimeout(self, t):
1300         self.delay = t
1301
1302     def setMaxConns(self, m):
1303         self.max_conns = m
1304
1305     def connect_ftp(self, user, passwd, host, port, dirs):
1306         key = user, host, port, '/'.join(dirs)
1307         if key in self.cache:
1308             self.timeout[key] = time.time() + self.delay
1309         else:
1310             self.cache[key] = ftpwrapper(user, passwd, host, port, dirs)
1311             self.timeout[key] = time.time() + self.delay
1312         self.check_cache()
1313         return self.cache[key]
1314
1315     def check_cache(self):
1316         # first check for old ones
1317         t = time.time()
1318         if self.soonest <= t:
1319             for k, v in self.timeout.items():
1320                 if v < t:
1321                     self.cache[k].close()
1322                     del self.cache[k]
1323                     del self.timeout[k]
1324         self.soonest = min(self.timeout.values())
1325
1326         # then check the size
1327         if len(self.cache) == self.max_conns:
1328             for k, v in self.timeout.items():
1329                 if v == self.soonest:
1330                     del self.cache[k]
1331                     del self.timeout[k]
1332                     break
1333             self.soonest = min(self.timeout.values())
1334
1335 class GopherHandler(BaseHandler):
1336     def gopher_open(self, req):
1337         # XXX can raise socket.error
1338         import gopherlib  # this raises DeprecationWarning in 2.5
1339         host = req.get_host()
1340         if not host:
1341             raise GopherError('no host given')
1342         host = unquote(host)
1343         selector = req.get_selector()
1344         type, selector = splitgophertype(selector)
1345         selector, query = splitquery(selector)
1346         selector = unquote(selector)
1347         if query:
1348             query = unquote(query)
1349             fp = gopherlib.send_query(selector, query, host)
1350         else:
1351             fp = gopherlib.send_selector(selector, host)
1352         return addinfourl(fp, noheaders(), req.get_full_url())