Lib/cookielib.py

   1 """HTTP cookie handling for web clients.
   2
   3 This module has (now fairly distant) origins in Gisle Aas' Perl module
   4 HTTP::Cookies, from the libwww-perl library.
   5
   6 Docstrings, comments and debug strings in this code refer to the
   7 attributes of the HTTP cookie system as cookie-attributes, to distinguish
   8 them clearly from Python attributes.
   9
  10 Class diagram (note that BSDDBCookieJar and the MSIE* classes are not
  11 distributed with the Python standard library, but are available from
  12 http://wwwsearch.sf.net/):
  13
  14                         CookieJar____
  15                         /     \      \
  16             FileCookieJar      \      \
  17              /    |   \         \      \
  18  MozillaCookieJar | LWPCookieJar \      \
  19                   |               |      \
  20                   |   ---MSIEBase |       \
  21                   |  /      |     |        \
  22                   | /   MSIEDBCookieJar BSDDBCookieJar
  23                   |/
  24                MSIECookieJar
  25
  26 """
  27
  28 __all__ = ['Cookie', 'CookieJar', 'CookiePolicy', 'DefaultCookiePolicy',
  29            'FileCookieJar', 'LWPCookieJar', 'LoadError', 'MozillaCookieJar']
  30
  31 import re, urlparse, copy, time, urllib
  32 try:
  33     import threading as _threading
  34 except ImportError:
  35     import dummy_threading as _threading
  36 import httplib  # only for the default HTTP port
  37 from calendar import timegm
  38
  39 debug = False   # set to True to enable debugging via the logging module
  40 logger = None
  41
  42 def _debug(*args):
  43     if not debug:
  44         return
  45     global logger
  46     if not logger:
  47         import logging
  48         logger = logging.getLogger("cookielib")
  49     return logger.debug(*args)
  50
  51
  52 DEFAULT_HTTP_PORT = str(httplib.HTTP_PORT)
  53 MISSING_FILENAME_TEXT = ("a filename was not supplied (nor was the CookieJar "
  54                          "instance initialised with one)")
  55
  56 def _warn_unhandled_exception():
  57     # There are a few catch-all except: statements in this module, for
  58     # catching input that's bad in unexpected ways.  Warn if any
  59     # exceptions are caught there.
  60     import warnings, traceback, StringIO
  61     f = StringIO.StringIO()
  62     traceback.print_exc(None, f)
  63     msg = f.getvalue()
  64     warnings.warn("cookielib bug!\n%s" % msg, stacklevel=2)
  65
  66
  67 # Date/time conversion
  68 # -----------------------------------------------------------------------------
  69
  70 EPOCH_YEAR = 1970
  71 def _timegm(tt):
  72     year, month, mday, hour, min, sec = tt[:6]
  73     if ((year >= EPOCH_YEAR) and (1 <= month <= 12) and (1 <= mday <= 31) and
  74         (0 <= hour <= 24) and (0 <= min <= 59) and (0 <= sec <= 61)):
  75         return timegm(tt)
  76     else:
  77         return None
  78
  79 DAYS = ["Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"]
  80 MONTHS = ["Jan", "Feb", "Mar", "Apr", "May", "Jun",
  81           "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"]
  82 MONTHS_LOWER = []
  83 for month in MONTHS: MONTHS_LOWER.append(month.lower())
  84
  85 def time2isoz(t=None):
  86     """Return a string representing time in seconds since epoch, t.
  87
  88     If the function is called without an argument, it will use the current
  89     time.
  90
  91     The format of the returned string is like "YYYY-MM-DD hh:mm:ssZ",
  92     representing Universal Time (UTC, aka GMT).  An example of this format is:
  93
  94     1994-11-24 08:49:37Z
  95
  96     """
  97     if t is None: t = time.time()
  98     year, mon, mday, hour, min, sec = time.gmtime(t)[:6]
  99     return "%04d-%02d-%02d %02d:%02d:%02dZ" % (
 100         year, mon, mday, hour, min, sec)
 101
 102 def time2netscape(t=None):
 103     """Return a string representing time in seconds since epoch, t.
 104
 105     If the function is called without an argument, it will use the current
 106     time.
 107
 108     The format of the returned string is like this:
 109
 110     Wed, DD-Mon-YYYY HH:MM:SS GMT
 111
 112     """
 113     if t is None: t = time.time()
 114     year, mon, mday, hour, min, sec, wday = time.gmtime(t)[:7]
 115     return "%s %02d-%s-%04d %02d:%02d:%02d GMT" % (
 116         DAYS[wday], mday, MONTHS[mon-1], year, hour, min, sec)
 117
 118
 119 UTC_ZONES = {"GMT": None, "UTC": None, "UT": None, "Z": None}
 120
 121 TIMEZONE_RE = re.compile(r"^([-+])?(\d\d?):?(\d\d)?$")
 122 def offset_from_tz_string(tz):
 123     offset = None
 124     if tz in UTC_ZONES:
 125         offset = 0
 126     else:
 127         m = TIMEZONE_RE.search(tz)
 128         if m:
 129             offset = 3600 * int(m.group(2))
 130             if m.group(3):
 131                 offset = offset + 60 * int(m.group(3))
 132             if m.group(1) == '-':
 133                 offset = -offset
 134     return offset
 135
 136 def _str2time(day, mon, yr, hr, min, sec, tz):
 137     # translate month name to number
 138     # month numbers start with 1 (January)
 139     try:
 140         mon = MONTHS_LOWER.index(mon.lower())+1
 141     except ValueError:
 142         # maybe it's already a number
 143         try:
 144             imon = int(mon)
 145         except ValueError:
 146             return None
 147         if 1 <= imon <= 12:
 148             mon = imon
 149         else:
 150             return None
 151
 152     # make sure clock elements are defined
 153     if hr is None: hr = 0
 154     if min is None: min = 0
 155     if sec is None: sec = 0
 156
 157     yr = int(yr)
 158     day = int(day)
 159     hr = int(hr)
 160     min = int(min)
 161     sec = int(sec)
 162
 163     if yr < 1000:
 164         # find "obvious" year
 165         cur_yr = time.localtime(time.time())[0]
 166         m = cur_yr % 100
 167         tmp = yr
 168         yr = yr + cur_yr - m
 169         m = m - tmp
 170         if abs(m) > 50:
 171             if m > 0: yr = yr + 100
 172             else: yr = yr - 100
 173
 174     # convert UTC time tuple to seconds since epoch (not timezone-adjusted)
 175     t = _timegm((yr, mon, day, hr, min, sec, tz))
 176
 177     if t is not None:
 178         # adjust time using timezone string, to get absolute time since epoch
 179         if tz is None:
 180             tz = "UTC"
 181         tz = tz.upper()
 182         offset = offset_from_tz_string(tz)
 183         if offset is None:
 184             return None
 185         t = t - offset
 186
 187     return t
 188
 189 STRICT_DATE_RE = re.compile(
 190     r"^[SMTWF][a-z][a-z], (\d\d) ([JFMASOND][a-z][a-z]) "
 191     "(\d\d\d\d) (\d\d):(\d\d):(\d\d) GMT$")
 192 WEEKDAY_RE = re.compile(
 193     r"^(?:Sun|Mon|Tue|Wed|Thu|Fri|Sat)[a-z]*,?\s*", re.I)
 194 LOOSE_HTTP_DATE_RE = re.compile(
 195     r"""^
 196     (\d\d?)            # day
 197        (?:\s+|[-\/])
 198     (\w+)              # month
 199         (?:\s+|[-\/])
 200     (\d+)              # year
 201     (?:
 202           (?:\s+|:)    # separator before clock
 203        (\d\d?):(\d\d)  # hour:min
 204        (?::(\d\d))?    # optional seconds
 205     )?                 # optional clock
 206        \s*
 207     ([-+]?\d{2,4}|(?![APap][Mm]\b)[A-Za-z]+)? # timezone
 208        \s*
 209     (?:\(\w+\))?       # ASCII representation of timezone in parens.
 210        \s*$""", re.X)
 211 def http2time(text):
 212     """Returns time in seconds since epoch of time represented by a string.
 213
 214     Return value is an integer.
 215
 216     None is returned if the format of str is unrecognized, the time is outside
 217     the representable range, or the timezone string is not recognized.  If the
 218     string contains no timezone, UTC is assumed.
 219
 220     The timezone in the string may be numerical (like "-0800" or "+0100") or a
 221     string timezone (like "UTC", "GMT", "BST" or "EST").  Currently, only the
 222     timezone strings equivalent to UTC (zero offset) are known to the function.
 223
 224     The function loosely parses the following formats:
 225
 226     Wed, 09 Feb 1994 22:23:32 GMT       -- HTTP format
 227     Tuesday, 08-Feb-94 14:15:29 GMT     -- old rfc850 HTTP format
 228     Tuesday, 08-Feb-1994 14:15:29 GMT   -- broken rfc850 HTTP format
 229     09 Feb 1994 22:23:32 GMT            -- HTTP format (no weekday)
 230     08-Feb-94 14:15:29 GMT              -- rfc850 format (no weekday)
 231     08-Feb-1994 14:15:29 GMT            -- broken rfc850 format (no weekday)
 232
 233     The parser ignores leading and trailing whitespace.  The time may be
 234     absent.
 235
 236     If the year is given with only 2 digits, the function will select the
 237     century that makes the year closest to the current date.
 238
 239     """
 240     # fast exit for strictly conforming string
 241     m = STRICT_DATE_RE.search(text)
 242     if m:
 243         g = m.groups()
 244         mon = MONTHS_LOWER.index(g[1].lower()) + 1
 245         tt = (int(g[2]), mon, int(g[0]),
 246               int(g[3]), int(g[4]), float(g[5]))
 247         return _timegm(tt)
 248
 249     # No, we need some messy parsing...
 250
 251     # clean up
 252     text = text.lstrip()
 253     text = WEEKDAY_RE.sub("", text, 1)  # Useless weekday
 254
 255     # tz is time zone specifier string
 256     day, mon, yr, hr, min, sec, tz = [None]*7
 257
 258     # loose regexp parse
 259     m = LOOSE_HTTP_DATE_RE.search(text)
 260     if m is not None:
 261         day, mon, yr, hr, min, sec, tz = m.groups()
 262     else:
 263         return None  # bad format
 264
 265     return _str2time(day, mon, yr, hr, min, sec, tz)
 266
 267 ISO_DATE_RE = re.compile(
 268     """^
 269     (\d{4})              # year
 270        [-\/]?
 271     (\d\d?)              # numerical month
 272        [-\/]?
 273     (\d\d?)              # day
 274    (?:
 275          (?:\s+|[-:Tt])  # separator before clock
 276       (\d\d?):?(\d\d)    # hour:min
 277       (?::?(\d\d(?:\.\d*)?))?  # optional seconds (and fractional)
 278    )?                    # optional clock
 279       \s*
 280    ([-+]?\d\d?:?(:?\d\d)?
 281     |Z|z)?               # timezone  (Z is "zero meridian", i.e. GMT)
 282       \s*$""", re.X)
 283 def iso2time(text):
 284     """
 285     As for http2time, but parses the ISO 8601 formats:
 286
 287     1994-02-03 14:15:29 -0100    -- ISO 8601 format
 288     1994-02-03 14:15:29          -- zone is optional
 289     1994-02-03                   -- only date
 290     1994-02-03T14:15:29          -- Use T as separator
 291     19940203T141529Z             -- ISO 8601 compact format
 292     19940203                     -- only date
 293
 294     """
 295     # clean up
 296     text = text.lstrip()
 297
 298     # tz is time zone specifier string
 299     day, mon, yr, hr, min, sec, tz = [None]*7
 300
 301     # loose regexp parse
 302     m = ISO_DATE_RE.search(text)
 303     if m is not None:
 304         # XXX there's an extra bit of the timezone I'm ignoring here: is
 305         #   this the right thing to do?
 306         yr, mon, day, hr, min, sec, tz, _ = m.groups()
 307     else:
 308         return None  # bad format
 309
 310     return _str2time(day, mon, yr, hr, min, sec, tz)
 311
 312
 313 # Header parsing
 314 # -----------------------------------------------------------------------------
 315
 316 def unmatched(match):
 317     """Return unmatched part of re.Match object."""
 318     start, end = match.span(0)
 319     return match.string[:start]+match.string[end:]
 320
 321 HEADER_TOKEN_RE =        re.compile(r"^\s*([^=\s;,]+)")
 322 HEADER_QUOTED_VALUE_RE = re.compile(r"^\s*=\s*\"([^\"\\]*(?:\\.[^\"\\]*)*)\"")
 323 HEADER_VALUE_RE =        re.compile(r"^\s*=\s*([^\s;,]*)")
 324 HEADER_ESCAPE_RE = re.compile(r"\\(.)")
 325 def split_header_words(header_values):
 326     r"""Parse header values into a list of lists containing key,value pairs.
 327
 328     The function knows how to deal with ",", ";" and "=" as well as quoted
 329     values after "=".  A list of space separated tokens are parsed as if they
 330     were separated by ";".
 331
 332     If the header_values passed as argument contains multiple values, then they
 333     are treated as if they were a single value separated by comma ",".
 334
 335     This means that this function is useful for parsing header fields that
 336     follow this syntax (BNF as from the HTTP/1.1 specification, but we relax
 337     the requirement for tokens).
 338
 339       headers           = #header
 340       header            = (token | parameter) *( [";"] (token | parameter))
 341
 342       token             = 1*<any CHAR except CTLs or separators>
 343       separators        = "(" | ")" | "<" | ">" | "@"
 344                         | "," | ";" | ":" | "\" | <">
 345                         | "/" | "[" | "]" | "?" | "="
 346                         | "{" | "}" | SP | HT
 347
 348       quoted-string     = ( <"> *(qdtext | quoted-pair ) <"> )
 349       qdtext            = <any TEXT except <">>
 350       quoted-pair       = "\" CHAR
 351
 352       parameter         = attribute "=" value
 353       attribute         = token
 354       value             = token | quoted-string
 355
 356     Each header is represented by a list of key/value pairs.  The value for a
 357     simple token (not part of a parameter) is None.  Syntactically incorrect
 358     headers will not necessarily be parsed as you would want.
 359
 360     This is easier to describe with some examples:
 361
 362     >>> split_header_words(['foo="bar"; port="80,81"; discard, bar=baz'])
 363     [[('foo', 'bar'), ('port', '80,81'), ('discard', None)], [('bar', 'baz')]]
 364     >>> split_header_words(['text/html; charset="iso-8859-1"'])
 365     [[('text/html', None), ('charset', 'iso-8859-1')]]
 366     >>> split_header_words([r'Basic realm="\"foo\bar\""'])
 367     [[('Basic', None), ('realm', '"foobar"')]]
 368
 369     """
 370     assert not isinstance(header_values, basestring)
 371     result = []
 372     for text in header_values:
 373         orig_text = text
 374         pairs = []
 375         while text:
 376             m = HEADER_TOKEN_RE.search(text)
 377             if m:
 378                 text = unmatched(m)
 379                 name = m.group(1)
 380                 m = HEADER_QUOTED_VALUE_RE.search(text)
 381                 if m:  # quoted value
 382                     text = unmatched(m)
 383                     value = m.group(1)
 384                     value = HEADER_ESCAPE_RE.sub(r"\1", value)
 385                 else:
 386                     m = HEADER_VALUE_RE.search(text)
 387                     if m:  # unquoted value
 388                         text = unmatched(m)
 389                         value = m.group(1)
 390                         value = value.rstrip()
 391                     else:
 392                         # no value, a lone token
 393                         value = None
 394                 pairs.append((name, value))
 395             elif text.lstrip().startswith(","):
 396                 # concatenated headers, as per RFC 2616 section 4.2
 397                 text = text.lstrip()[1:]
 398                 if pairs: result.append(pairs)
 399                 pairs = []
 400             else:
 401                 # skip junk
 402                 non_junk, nr_junk_chars = re.subn("^[=\s;]*", "", text)
 403                 assert nr_junk_chars > 0, (
 404                     "split_header_words bug: '%s', '%s', %s" %
 405                     (orig_text, text, pairs))
 406                 text = non_junk
 407         if pairs: result.append(pairs)
 408     return result
 409
 410 HEADER_JOIN_ESCAPE_RE = re.compile(r"([\"\\])")
 411 def join_header_words(lists):
 412     """Do the inverse (almost) of the conversion done by split_header_words.
 413
 414     Takes a list of lists of (key, value) pairs and produces a single header
 415     value.  Attribute values are quoted if needed.
 416
 417     >>> join_header_words([[("text/plain", None), ("charset", "iso-8859/1")]])
 418     'text/plain; charset="iso-8859/1"'
 419     >>> join_header_words([[("text/plain", None)], [("charset", "iso-8859/1")]])
 420     'text/plain, charset="iso-8859/1"'
 421
 422     """
 423     headers = []
 424     for pairs in lists:
 425         attr = []
 426         for k, v in pairs:
 427             if v is not None:
 428                 if not re.search(r"^\w+$", v):
 429                     v = HEADER_JOIN_ESCAPE_RE.sub(r"\\\1", v)  # escape " and \
 430                     v = '"%s"' % v
 431                 k = "%s=%s" % (k, v)
 432             attr.append(k)
 433         if attr: headers.append("; ".join(attr))
 434     return ", ".join(headers)
 435
 436 def parse_ns_headers(ns_headers):
 437     """Ad-hoc parser for Netscape protocol cookie-attributes.
 438
 439     The old Netscape cookie format for Set-Cookie can for instance contain
 440     an unquoted "," in the expires field, so we have to use this ad-hoc
 441     parser instead of split_header_words.
 442
 443     XXX This may not make the best possible effort to parse all the crap
 444     that Netscape Cookie headers contain.  Ronald Tschalar's HTTPClient
 445     parser is probably better, so could do worse than following that if
 446     this ever gives any trouble.
 447
 448     Currently, this is also used for parsing RFC 2109 cookies.
 449
 450     """
 451     known_attrs = ("expires", "domain", "path", "secure",
 452                    # RFC 2109 attrs (may turn up in Netscape cookies, too)
 453                    "port", "max-age")
 454
 455     result = []
 456     for ns_header in ns_headers:
 457         pairs = []
 458         version_set = False
 459         for ii, param in enumerate(re.split(r";\s*", ns_header)):
 460             param = param.rstrip()
 461             if param == "": continue
 462             if "=" not in param:
 463                 k, v = param, None
 464             else:
 465                 k, v = re.split(r"\s*=\s*", param, 1)
 466                 k = k.lstrip()
 467             if ii != 0:
 468                 lc = k.lower()
 469                 if lc in known_attrs:
 470                     k = lc
 471                 if k == "version":
 472                     # This is an RFC 2109 cookie.
 473                     version_set = True
 474                 if k == "expires":
 475                     # convert expires date to seconds since epoch
 476                     if v.startswith('"'): v = v[1:]
 477                     if v.endswith('"'): v = v[:-1]
 478                     v = http2time(v)  # None if invalid
 479             pairs.append((k, v))
 480
 481         if pairs:
 482             if not version_set:
 483                 pairs.append(("version", "0"))
 484             result.append(pairs)
 485
 486     return result
 487
 488
 489 IPV4_RE = re.compile(r"\.\d+$")
 490 def is_HDN(text):
 491     """Return True if text is a host domain name."""
 492     # XXX
 493     # This may well be wrong.  Which RFC is HDN defined in, if any (for
 494     #  the purposes of RFC 2965)?
 495     # For the current implementation, what about IPv6?  Remember to look
 496     #  at other uses of IPV4_RE also, if change this.
 497     if IPV4_RE.search(text):
 498         return False
 499     if text == "":
 500         return False
 501     if text[0] == "." or text[-1] == ".":
 502         return False
 503     return True
 504
 505 def domain_match(A, B):
 506     """Return True if domain A domain-matches domain B, according to RFC 2965.
 507
 508     A and B may be host domain names or IP addresses.
 509
 510     RFC 2965, section 1:
 511
 512     Host names can be specified either as an IP address or a HDN string.
 513     Sometimes we compare one host name with another.  (Such comparisons SHALL
 514     be case-insensitive.)  Host A's name domain-matches host B's if
 515
 516          *  their host name strings string-compare equal; or
 517
 518          * A is a HDN string and has the form NB, where N is a non-empty
 519             name string, B has the form .B', and B' is a HDN string.  (So,
 520             x.y.com domain-matches .Y.com but not Y.com.)
 521
 522     Note that domain-match is not a commutative operation: a.b.c.com
 523     domain-matches .c.com, but not the reverse.
 524
 525     """
 526     # Note that, if A or B are IP addresses, the only relevant part of the
 527     # definition of the domain-match algorithm is the direct string-compare.
 528     A = A.lower()
 529     B = B.lower()
 530     if A == B:
 531         return True
 532     if not is_HDN(A):
 533         return False
 534     i = A.rfind(B)
 535     if i == -1 or i == 0:
 536         # A does not have form NB, or N is the empty string
 537         return False
 538     if not B.startswith("."):
 539         return False
 540     if not is_HDN(B[1:]):
 541         return False
 542     return True
 543
 544 def liberal_is_HDN(text):
 545     """Return True if text is a sort-of-like a host domain name.
 546
 547     For accepting/blocking domains.
 548
 549     """
 550     if IPV4_RE.search(text):
 551         return False
 552     return True
 553
 554 def user_domain_match(A, B):
 555     """For blocking/accepting domains.
 556
 557     A and B may be host domain names or IP addresses.
 558
 559     """
 560     A = A.lower()
 561     B = B.lower()
 562     if not (liberal_is_HDN(A) and liberal_is_HDN(B)):
 563         if A == B:
 564             # equal IP addresses
 565             return True
 566         return False
 567     initial_dot = B.startswith(".")
 568     if initial_dot and A.endswith(B):
 569         return True
 570     if not initial_dot and A == B:
 571         return True
 572     return False
 573
 574 cut_port_re = re.compile(r":\d+$")
 575 def request_host(request):
 576     """Return request-host, as defined by RFC 2965.
 577
 578     Variation from RFC: returned value is lowercased, for convenient
 579     comparison.
 580
 581     """
 582     url = request.get_full_url()
 583     host = urlparse.urlparse(url)[1]
 584     if host == "":
 585         host = request.get_header("Host", "")
 586
 587     # remove port, if present
 588     host = cut_port_re.sub("", host, 1)
 589     return host.lower()
 590
 591 def eff_request_host(request):
 592     """Return a tuple (request-host, effective request-host name).
 593
 594     As defined by RFC 2965, except both are lowercased.
 595
 596     """
 597     erhn = req_host = request_host(request)
 598     if req_host.find(".") == -1 and not IPV4_RE.search(req_host):
 599         erhn = req_host + ".local"
 600     return req_host, erhn
 601
 602 def request_path(request):
 603     """request-URI, as defined by RFC 2965."""
 604     url = request.get_full_url()
 605     #scheme, netloc, path, parameters, query, frag = urlparse.urlparse(url)
 606     #req_path = escape_path("".join(urlparse.urlparse(url)[2:]))
 607     path, parameters, query, frag = urlparse.urlparse(url)[2:]
 608     if parameters:
 609         path = "%s;%s" % (path, parameters)
 610     path = escape_path(path)
 611     req_path = urlparse.urlunparse(("", "", path, "", query, frag))
 612     if not req_path.startswith("/"):
 613         # fix bad RFC 2396 absoluteURI
 614         req_path = "/"+req_path
 615     return req_path
 616
 617 def request_port(request):
 618     host = request.get_host()
 619     i = host.find(':')
 620     if i >= 0:
 621         port = host[i+1:]
 622         try:
 623             int(port)
 624         except ValueError:
 625             _debug("nonnumeric port: '%s'", port)
 626             return None
 627     else:
 628         port = DEFAULT_HTTP_PORT
 629     return port
 630
 631 # Characters in addition to A-Z, a-z, 0-9, '_', '.', and '-' that don't
 632 # need to be escaped to form a valid HTTP URL (RFCs 2396 and 1738).
 633 HTTP_PATH_SAFE = "%/;:@&=+$,!~*'()"
 634 ESCAPED_CHAR_RE = re.compile(r"%([0-9a-fA-F][0-9a-fA-F])")
 635 def uppercase_escaped_char(match):
 636     return "%%%s" % match.group(1).upper()
 637 def escape_path(path):
 638     """Escape any invalid characters in HTTP URL, and uppercase all escapes."""
 639     # There's no knowing what character encoding was used to create URLs
 640     # containing %-escapes, but since we have to pick one to escape invalid
 641     # path characters, we pick UTF-8, as recommended in the HTML 4.0
 642     # specification:
 643     # http://www.w3.org/TR/REC-html40/appendix/notes.html#h-B.2.1
 644     # And here, kind of: draft-fielding-uri-rfc2396bis-03
 645     # (And in draft IRI specification: draft-duerst-iri-05)
 646     # (And here, for new URI schemes: RFC 2718)
 647     if isinstance(path, unicode):
 648         path = path.encode("utf-8")
 649     path = urllib.quote(path, HTTP_PATH_SAFE)
 650     path = ESCAPED_CHAR_RE.sub(uppercase_escaped_char, path)
 651     return path
 652
 653 def reach(h):
 654     """Return reach of host h, as defined by RFC 2965, section 1.
 655
 656     The reach R of a host name H is defined as follows:
 657
 658        *  If
 659
 660           -  H is the host domain name of a host; and,
 661
 662           -  H has the form A.B; and
 663
 664           -  A has no embedded (that is, interior) dots; and
 665
 666           -  B has at least one embedded dot, or B is the string "local".
 667              then the reach of H is .B.
 668
 669        *  Otherwise, the reach of H is H.
 670
 671     >>> reach("www.acme.com")
 672     '.acme.com'
 673     >>> reach("acme.com")
 674     'acme.com'
 675     >>> reach("acme.local")
 676     '.local'
 677
 678     """
 679     i = h.find(".")
 680     if i >= 0:
 681         #a = h[:i]  # this line is only here to show what a is
 682         b = h[i+1:]
 683         i = b.find(".")
 684         if is_HDN(h) and (i >= 0 or b == "local"):
 685             return "."+b
 686     return h
 687
 688 def is_third_party(request):
 689     """
 690
 691     RFC 2965, section 3.3.6:
 692
 693         An unverifiable transaction is to a third-party host if its request-
 694         host U does not domain-match the reach R of the request-host O in the
 695         origin transaction.
 696
 697     """
 698     req_host = request_host(request)
 699     if not domain_match(req_host, reach(request.get_origin_req_host())):
 700         return True
 701     else:
 702         return False
 703
 704
 705 class Cookie:
 706     """HTTP Cookie.
 707
 708     This class represents both Netscape and RFC 2965 cookies.
 709
 710     This is deliberately a very simple class.  It just holds attributes.  It's
 711     possible to construct Cookie instances that don't comply with the cookie
 712     standards.  CookieJar.make_cookies is the factory function for Cookie
 713     objects -- it deals with cookie parsing, supplying defaults, and
 714     normalising to the representation used in this class.  CookiePolicy is
 715     responsible for checking them to see whether they should be accepted from
 716     and returned to the server.
 717
 718     Note that the port may be present in the headers, but unspecified ("Port"
 719     rather than"Port=80", for example); if this is the case, port is None.
 720
 721     """
 722
 723     def __init__(self, version, name, value,
 724                  port, port_specified,
 725                  domain, domain_specified, domain_initial_dot,
 726                  path, path_specified,
 727                  secure,
 728                  expires,
 729                  discard,
 730                  comment,
 731                  comment_url,
 732                  rest,
 733                  rfc2109=False,
 734                  ):
 735
 736         if version is not None: version = int(version)
 737         if expires is not None: expires = int(expires)
 738         if port is None and port_specified is True:
 739             raise ValueError("if port is None, port_specified must be false")
 740
 741         self.version = version
 742         self.name = name
 743         self.value = value
 744         self.port = port
 745         self.port_specified = port_specified
 746         # normalise case, as per RFC 2965 section 3.3.3
 747         self.domain = domain.lower()
 748         self.domain_specified = domain_specified
 749         # Sigh.  We need to know whether the domain given in the
 750         # cookie-attribute had an initial dot, in order to follow RFC 2965
 751         # (as clarified in draft errata).  Needed for the returned $Domain
 752         # value.
 753         self.domain_initial_dot = domain_initial_dot
 754         self.path = path
 755         self.path_specified = path_specified
 756         self.secure = secure
 757         self.expires = expires
 758         self.discard = discard
 759         self.comment = comment
 760         self.comment_url = comment_url
 761         self.rfc2109 = rfc2109
 762
 763         self._rest = copy.copy(rest)
 764
 765     def has_nonstandard_attr(self, name):
 766         return name in self._rest
 767     def get_nonstandard_attr(self, name, default=None):
 768         return self._rest.get(name, default)
 769     def set_nonstandard_attr(self, name, value):
 770         self._rest[name] = value
 771
 772     def is_expired(self, now=None):
 773         if now is None: now = time.time()
 774         if (self.expires is not None) and (self.expires <= now):
 775             return True
 776         return False
 777
 778     def __str__(self):
 779         if self.port is None: p = ""
 780         else: p = ":"+self.port
 781         limit = self.domain + p + self.path
 782         if self.value is not None:
 783             namevalue = "%s=%s" % (self.name, self.value)
 784         else:
 785             namevalue = self.name
 786         return "<Cookie %s for %s>" % (namevalue, limit)
 787
 788     def __repr__(self):
 789         args = []
 790         for name in ("version", "name", "value",
 791                      "port", "port_specified",
 792                      "domain", "domain_specified", "domain_initial_dot",
 793                      "path", "path_specified",
 794                      "secure", "expires", "discard", "comment", "comment_url",
 795                      ):
 796             attr = getattr(self, name)
 797             args.append("%s=%s" % (name, repr(attr)))
 798         args.append("rest=%s" % repr(self._rest))
 799         args.append("rfc2109=%s" % repr(self.rfc2109))
 800         return "Cookie(%s)" % ", ".join(args)
 801
 802
 803 class CookiePolicy:
 804     """Defines which cookies get accepted from and returned to server.
 805
 806     May also modify cookies, though this is probably a bad idea.
 807
 808     The subclass DefaultCookiePolicy defines the standard rules for Netscape
 809     and RFC 2965 cookies -- override that if you want a customised policy.
 810
 811     """
 812     def set_ok(self, cookie, request):
 813         """Return true if (and only if) cookie should be accepted from server.
 814
 815         Currently, pre-expired cookies never get this far -- the CookieJar
 816         class deletes such cookies itself.
 817
 818         """
 819         raise NotImplementedError()
 820
 821     def return_ok(self, cookie, request):
 822         """Return true if (and only if) cookie should be returned to server."""
 823         raise NotImplementedError()
 824
 825     def domain_return_ok(self, domain, request):
 826         """Return false if cookies should not be returned, given cookie domain.
 827         """
 828         return True
 829
 830     def path_return_ok(self, path, request):
 831         """Return false if cookies should not be returned, given cookie path.
 832         """
 833         return True
 834
 835
 836 class DefaultCookiePolicy(CookiePolicy):
 837     """Implements the standard rules for accepting and returning cookies."""
 838
 839     DomainStrictNoDots = 1
 840     DomainStrictNonDomain = 2
 841     DomainRFC2965Match = 4
 842
 843     DomainLiberal = 0
 844     DomainStrict = DomainStrictNoDots|DomainStrictNonDomain
 845
 846     def __init__(self,
 847                  blocked_domains=None, allowed_domains=None,
 848                  netscape=True, rfc2965=False,
 849                  rfc2109_as_netscape=None,
 850                  hide_cookie2=False,
 851                  strict_domain=False,
 852                  strict_rfc2965_unverifiable=True,
 853                  strict_ns_unverifiable=False,
 854                  strict_ns_domain=DomainLiberal,
 855                  strict_ns_set_initial_dollar=False,
 856                  strict_ns_set_path=False,
 857                  ):
 858         """Constructor arguments should be passed as keyword arguments only."""
 859         self.netscape = netscape
 860         self.rfc2965 = rfc2965
 861         self.rfc2109_as_netscape = rfc2109_as_netscape
 862         self.hide_cookie2 = hide_cookie2
 863         self.strict_domain = strict_domain
 864         self.strict_rfc2965_unverifiable = strict_rfc2965_unverifiable
 865         self.strict_ns_unverifiable = strict_ns_unverifiable
 866         self.strict_ns_domain = strict_ns_domain
 867         self.strict_ns_set_initial_dollar = strict_ns_set_initial_dollar
 868         self.strict_ns_set_path = strict_ns_set_path
 869
 870         if blocked_domains is not None:
 871             self._blocked_domains = tuple(blocked_domains)
 872         else:
 873             self._blocked_domains = ()
 874
 875         if allowed_domains is not None:
 876             allowed_domains = tuple(allowed_domains)
 877         self._allowed_domains = allowed_domains
 878
 879     def blocked_domains(self):
 880         """Return the sequence of blocked domains (as a tuple)."""
 881         return self._blocked_domains
 882     def set_blocked_domains(self, blocked_domains):
 883         """Set the sequence of blocked domains."""
 884         self._blocked_domains = tuple(blocked_domains)
 885
 886     def is_blocked(self, domain):
 887         for blocked_domain in self._blocked_domains:
 888             if user_domain_match(domain, blocked_domain):
 889                 return True
 890         return False
 891
 892     def allowed_domains(self):
 893         """Return None, or the sequence of allowed domains (as a tuple)."""
 894         return self._allowed_domains
 895     def set_allowed_domains(self, allowed_domains):
 896         """Set the sequence of allowed domains, or None."""
 897         if allowed_domains is not None:
 898             allowed_domains = tuple(allowed_domains)
 899         self._allowed_domains = allowed_domains
 900
 901     def is_not_allowed(self, domain):
 902         if self._allowed_domains is None:
 903             return False
 904         for allowed_domain in self._allowed_domains:
 905             if user_domain_match(domain, allowed_domain):
 906                 return False
 907         return True
 908
 909     def set_ok(self, cookie, request):
 910         """
 911         If you override .set_ok(), be sure to call this method.  If it returns
 912         false, so should your subclass (assuming your subclass wants to be more
 913         strict about which cookies to accept).
 914
 915         """
 916         _debug(" - checking cookie %s=%s", cookie.name, cookie.value)
 917
 918         assert cookie.name is not None
 919
 920         for n in "version", "verifiability", "name", "path", "domain", "port":
 921             fn_name = "set_ok_"+n
 922             fn = getattr(self, fn_name)
 923             if not fn(cookie, request):
 924                 return False
 925
 926         return True
 927
 928     def set_ok_version(self, cookie, request):
 929         if cookie.version is None:
 930             # Version is always set to 0 by parse_ns_headers if it's a Netscape
 931             # cookie, so this must be an invalid RFC 2965 cookie.
 932             _debug("   Set-Cookie2 without version attribute (%s=%s)",
 933                    cookie.name, cookie.value)
 934             return False
 935         if cookie.version > 0 and not self.rfc2965:
 936             _debug("   RFC 2965 cookies are switched off")
 937             return False
 938         elif cookie.version == 0 and not self.netscape:
 939             _debug("   Netscape cookies are switched off")
 940             return False
 941         return True
 942
 943     def set_ok_verifiability(self, cookie, request):
 944         if request.is_unverifiable() and is_third_party(request):
 945             if cookie.version > 0 and self.strict_rfc2965_unverifiable:
 946                 _debug("   third-party RFC 2965 cookie during "
 947                              "unverifiable transaction")
 948                 return False
 949             elif cookie.version == 0 and self.strict_ns_unverifiable:
 950                 _debug("   third-party Netscape cookie during "
 951                              "unverifiable transaction")
 952                 return False
 953         return True
 954
 955     def set_ok_name(self, cookie, request):
 956         # Try and stop servers setting V0 cookies designed to hack other
 957         # servers that know both V0 and V1 protocols.
 958         if (cookie.version == 0 and self.strict_ns_set_initial_dollar and
 959             cookie.name.startswith("$")):
 960             _debug("   illegal name (starts with '$'): '%s'", cookie.name)
 961             return False
 962         return True
 963
 964     def set_ok_path(self, cookie, request):
 965         if cookie.path_specified:
 966             req_path = request_path(request)
 967             if ((cookie.version > 0 or
 968                  (cookie.version == 0 and self.strict_ns_set_path)) and
 969                 not req_path.startswith(cookie.path)):
 970                 _debug("   path attribute %s is not a prefix of request "
 971                        "path %s", cookie.path, req_path)
 972                 return False
 973         return True
 974
 975     def set_ok_domain(self, cookie, request):
 976         if self.is_blocked(cookie.domain):
 977             _debug("   domain %s is in user block-list", cookie.domain)
 978             return False
 979         if self.is_not_allowed(cookie.domain):
 980             _debug("   domain %s is not in user allow-list", cookie.domain)
 981             return False
 982         if cookie.domain_specified:
 983             req_host, erhn = eff_request_host(request)
 984             domain = cookie.domain
 985             if self.strict_domain and (domain.count(".") >= 2):
 986                 # XXX This should probably be compared with the Konqueror
 987                 # (kcookiejar.cpp) and Mozilla implementations, but it's a
 988                 # losing battle.
 989                 i = domain.rfind(".")
 990                 j = domain.rfind(".", 0, i)
 991                 if j == 0:  # domain like .foo.bar
 992                     tld = domain[i+1:]
 993                     sld = domain[j+1:i]
 994                     if sld.lower() in ("co", "ac", "com", "edu", "org", "net",
 995                        "gov", "mil", "int", "aero", "biz", "cat", "coop",
 996                        "info", "jobs", "mobi", "museum", "name", "pro",
 997                        "travel", "eu") and len(tld) == 2:
 998                         # domain like .co.uk
 999                         _debug("   country-code second level domain %s", domain)
1000                         return False
1001             if domain.startswith("."):
1002                 undotted_domain = domain[1:]
1003             else:
1004                 undotted_domain = domain
1005             embedded_dots = (undotted_domain.find(".") >= 0)
1006             if not embedded_dots and domain != ".local":
1007                 _debug("   non-local domain %s contains no embedded dot",
1008                        domain)
1009                 return False
1010             if cookie.version == 0:
1011                 if (not erhn.endswith(domain) and
1012                     (not erhn.startswith(".") and
1013                      not ("."+erhn).endswith(domain))):
1014                     _debug("   effective request-host %s (even with added "
1015                            "initial dot) does not end end with %s",
1016                            erhn, domain)
1017                     return False
1018             if (cookie.version > 0 or
1019                 (self.strict_ns_domain & self.DomainRFC2965Match)):
1020                 if not domain_match(erhn, domain):
1021                     _debug("   effective request-host %s does not domain-match "
1022                            "%s", erhn, domain)
1023                     return False
1024             if (cookie.version > 0 or
1025                 (self.strict_ns_domain & self.DomainStrictNoDots)):
1026                 host_prefix = req_host[:-len(domain)]
1027                 if (host_prefix.find(".") >= 0 and
1028                     not IPV4_RE.search(req_host)):
1029                     _debug("   host prefix %s for domain %s contains a dot",
1030                            host_prefix, domain)
1031                     return False
1032         return True
1033
1034     def set_ok_port(self, cookie, request):
1035         if cookie.port_specified:
1036             req_port = request_port(request)
1037             if req_port is None:
1038                 req_port = "80"
1039             else:
1040                 req_port = str(req_port)
1041             for p in cookie.port.split(","):
1042                 try:
1043                     int(p)
1044                 except ValueError:
1045                     _debug("   bad port %s (not numeric)", p)
1046                     return False
1047                 if p == req_port:
1048                     break
1049             else:
1050                 _debug("   request port (%s) not found in %s",
1051                        req_port, cookie.port)
1052                 return False
1053         return True
1054
1055     def return_ok(self, cookie, request):
1056         """
1057         If you override .return_ok(), be sure to call this method.  If it
1058         returns false, so should your subclass (assuming your subclass wants to
1059         be more strict about which cookies to return).
1060
1061         """
1062         # Path has already been checked by .path_return_ok(), and domain
1063         # blocking done by .domain_return_ok().
1064         _debug(" - checking cookie %s=%s", cookie.name, cookie.value)
1065
1066         for n in "version", "verifiability", "secure", "expires", "port", "domain":
1067             fn_name = "return_ok_"+n
1068             fn = getattr(self, fn_name)
1069             if not fn(cookie, request):
1070                 return False
1071         return True
1072
1073     def return_ok_version(self, cookie, request):
1074         if cookie.version > 0 and not self.rfc2965:
1075             _debug("   RFC 2965 cookies are switched off")
1076             return False
1077         elif cookie.version == 0 and not self.netscape:
1078             _debug("   Netscape cookies are switched off")
1079             return False
1080         return True
1081
1082     def return_ok_verifiability(self, cookie, request):
1083         if request.is_unverifiable() and is_third_party(request):
1084             if cookie.version > 0 and self.strict_rfc2965_unverifiable:
1085                 _debug("   third-party RFC 2965 cookie during unverifiable "
1086                        "transaction")
1087                 return False
1088             elif cookie.version == 0 and self.strict_ns_unverifiable:
1089                 _debug("   third-party Netscape cookie during unverifiable "
1090                        "transaction")
1091                 return False
1092         return True
1093
1094     def return_ok_secure(self, cookie, request):
1095         if cookie.secure and request.get_type() != "https":
1096             _debug("   secure cookie with non-secure request")
1097             return False
1098         return True
1099
1100     def return_ok_expires(self, cookie, request):
1101         if cookie.is_expired(self._now):
1102             _debug("   cookie expired")
1103             return False
1104         return True
1105
1106     def return_ok_port(self, cookie, request):
1107         if cookie.port:
1108             req_port = request_port(request)
1109             if req_port is None:
1110                 req_port = "80"
1111             for p in cookie.port.split(","):
1112                 if p == req_port:
1113                     break
1114             else:
1115                 _debug("   request port %s does not match cookie port %s",
1116                        req_port, cookie.port)
1117                 return False
1118         return True
1119
1120     def return_ok_domain(self, cookie, request):
1121         req_host, erhn = eff_request_host(request)
1122         domain = cookie.domain
1123
1124         # strict check of non-domain cookies: Mozilla does this, MSIE5 doesn't
1125         if (cookie.version == 0 and
1126             (self.strict_ns_domain & self.DomainStrictNonDomain) and
1127             not cookie.domain_specified and domain != erhn):
1128             _debug("   cookie with unspecified domain does not string-compare "
1129                    "equal to request domain")
1130             return False
1131
1132         if cookie.version > 0 and not domain_match(erhn, domain):
1133             _debug("   effective request-host name %s does not domain-match "
1134                    "RFC 2965 cookie domain %s", erhn, domain)
1135             return False
1136         if cookie.version == 0 and not ("."+erhn).endswith(domain):
1137             _debug("   request-host %s does not match Netscape cookie domain "
1138                    "%s", req_host, domain)
1139             return False
1140         return True
1141
1142     def domain_return_ok(self, domain, request):
1143         # Liberal check of.  This is here as an optimization to avoid
1144         # having to load lots of MSIE cookie files unless necessary.
1145         req_host, erhn = eff_request_host(request)
1146         if not req_host.startswith("."):
1147             req_host = "."+req_host
1148         if not erhn.startswith("."):
1149             erhn = "."+erhn
1150         if not (req_host.endswith(domain) or erhn.endswith(domain)):
1151             #_debug("   request domain %s does not match cookie domain %s",
1152             #       req_host, domain)
1153             return False
1154
1155         if self.is_blocked(domain):
1156             _debug("   domain %s is in user block-list", domain)
1157             return False
1158         if self.is_not_allowed(domain):
1159             _debug("   domain %s is not in user allow-list", domain)
1160             return False
1161
1162         return True
1163
1164     def path_return_ok(self, path, request):
1165         _debug("- checking cookie path=%s", path)
1166         req_path = request_path(request)
1167         if not req_path.startswith(path):
1168             _debug("  %s does not path-match %s", req_path, path)
1169             return False
1170         return True
1171
1172
1173 def vals_sorted_by_key(adict):
1174     keys = adict.keys()
1175     keys.sort()
1176     return map(adict.get, keys)
1177
1178 def deepvalues(mapping):
1179     """Iterates over nested mapping, depth-first, in sorted order by key."""
1180     values = vals_sorted_by_key(mapping)
1181     for obj in values:
1182         mapping = False
1183         try:
1184             obj.items
1185         except AttributeError:
1186             pass
1187         else:
1188             mapping = True
1189             for subobj in deepvalues(obj):
1190                 yield subobj
1191         if not mapping:
1192             yield obj
1193
1194
1195 # Used as second parameter to dict.get() method, to distinguish absent
1196 # dict key from one with a None value.
1197 class Absent: pass
1198
1199 class CookieJar:
1200     """Collection of HTTP cookies.
1201
1202     You may not need to know about this class: try
1203     urllib2.build_opener(HTTPCookieProcessor).open(url).
1204
1205     """
1206
1207     non_word_re = re.compile(r"\W")
1208     quote_re = re.compile(r"([\"\\])")
1209     strict_domain_re = re.compile(r"\.?[^.]*")
1210     domain_re = re.compile(r"[^.]*")
1211     dots_re = re.compile(r"^\.+")
1212
1213     magic_re = r"^\#LWP-Cookies-(\d+\.\d+)"
1214
1215     def __init__(self, policy=None):
1216         if policy is None:
1217             policy = DefaultCookiePolicy()
1218         self._policy = policy
1219
1220         self._cookies_lock = _threading.RLock()
1221         self._cookies = {}
1222
1223     def set_policy(self, policy):
1224         self._policy = policy
1225
1226     def _cookies_for_domain(self, domain, request):
1227         cookies = []
1228         if not self._policy.domain_return_ok(domain, request):
1229             return []
1230         _debug("Checking %s for cookies to return", domain)
1231         cookies_by_path = self._cookies[domain]
1232         for path in cookies_by_path.keys():
1233             if not self._policy.path_return_ok(path, request):
1234                 continue
1235             cookies_by_name = cookies_by_path[path]
1236             for cookie in cookies_by_name.values():
1237                 if not self._policy.return_ok(cookie, request):
1238                     _debug("   not returning cookie")
1239                     continue
1240                 _debug("   it's a match")
1241                 cookies.append(cookie)
1242         return cookies
1243
1244     def _cookies_for_request(self, request):
1245         """Return a list of cookies to be returned to server."""
1246         cookies = []
1247         for domain in self._cookies.keys():
1248             cookies.extend(self._cookies_for_domain(domain, request))
1249         return cookies
1250
1251     def _cookie_attrs(self, cookies):
1252         """Return a list of cookie-attributes to be returned to server.
1253
1254         like ['foo="bar"; $Path="/"', ...]
1255
1256         The $Version attribute is also added when appropriate (currently only
1257         once per request).
1258
1259         """
1260         # add cookies in order of most specific (ie. longest) path first
1261         def decreasing_size(a, b): return cmp(len(b.path), len(a.path))
1262         cookies.sort(decreasing_size)
1263
1264         version_set = False
1265
1266         attrs = []
1267         for cookie in cookies:
1268             # set version of Cookie header
1269             # XXX
1270             # What should it be if multiple matching Set-Cookie headers have
1271             #  different versions themselves?
1272             # Answer: there is no answer; was supposed to be settled by
1273             #  RFC 2965 errata, but that may never appear...
1274             version = cookie.version
1275             if not version_set:
1276                 version_set = True
1277                 if version > 0:
1278                     attrs.append("$Version=%s" % version)
1279
1280             # quote cookie value if necessary
1281             # (not for Netscape protocol, which already has any quotes
1282             #  intact, due to the poorly-specified Netscape Cookie: syntax)
1283             if ((cookie.value is not None) and
1284                 self.non_word_re.search(cookie.value) and version > 0):
1285                 value = self.quote_re.sub(r"\\\1", cookie.value)
1286             else:
1287                 value = cookie.value
1288
1289             # add cookie-attributes to be returned in Cookie header
1290             if cookie.value is None:
1291                 attrs.append(cookie.name)
1292             else:
1293                 attrs.append("%s=%s" % (cookie.name, value))
1294             if version > 0:
1295                 if cookie.path_specified:
1296                     attrs.append('$Path="%s"' % cookie.path)
1297                 if cookie.domain.startswith("."):
1298                     domain = cookie.domain
1299                     if (not cookie.domain_initial_dot and
1300                         domain.startswith(".")):
1301                         domain = domain[1:]
1302                     attrs.append('$Domain="%s"' % domain)
1303                 if cookie.port is not None:
1304                     p = "$Port"
1305                     if cookie.port_specified:
1306                         p = p + ('="%s"' % cookie.port)
1307                     attrs.append(p)
1308
1309         return attrs
1310
1311     def add_cookie_header(self, request):
1312         """Add correct Cookie: header to request (urllib2.Request object).
1313
1314         The Cookie2 header is also added unless policy.hide_cookie2 is true.
1315
1316         """
1317         _debug("add_cookie_header")
1318         self._cookies_lock.acquire()
1319         try:
1320
1321             self._policy._now = self._now = int(time.time())
1322
1323             cookies = self._cookies_for_request(request)
1324
1325             attrs = self._cookie_attrs(cookies)
1326             if attrs:
1327                 if not request.has_header("Cookie"):
1328                     request.add_unredirected_header(
1329                         "Cookie", "; ".join(attrs))
1330
1331             # if necessary, advertise that we know RFC 2965
1332             if (self._policy.rfc2965 and not self._policy.hide_cookie2 and
1333                 not request.has_header("Cookie2")):
1334                 for cookie in cookies:
1335                     if cookie.version != 1:
1336                         request.add_unredirected_header("Cookie2", '$Version="1"')
1337                         break
1338
1339         finally:
1340             self._cookies_lock.release()
1341
1342         self.clear_expired_cookies()
1343
1344     def _normalized_cookie_tuples(self, attrs_set):
1345         """Return list of tuples containing normalised cookie information.
1346
1347         attrs_set is the list of lists of key,value pairs extracted from
1348         the Set-Cookie or Set-Cookie2 headers.
1349
1350         Tuples are name, value, standard, rest, where name and value are the
1351         cookie name and value, standard is a dictionary containing the standard
1352         cookie-attributes (discard, secure, version, expires or max-age,
1353         domain, path and port) and rest is a dictionary containing the rest of
1354         the cookie-attributes.
1355
1356         """
1357         cookie_tuples = []
1358
1359         boolean_attrs = "discard", "secure"
1360         value_attrs = ("version",
1361                        "expires", "max-age",
1362                        "domain", "path", "port",
1363                        "comment", "commenturl")
1364
1365         for cookie_attrs in attrs_set:
1366             name, value = cookie_attrs[0]
1367
1368             # Build dictionary of standard cookie-attributes (standard) and
1369             # dictionary of other cookie-attributes (rest).
1370
1371             # Note: expiry time is normalised to seconds since epoch.  V0
1372             # cookies should have the Expires cookie-attribute, and V1 cookies
1373             # should have Max-Age, but since V1 includes RFC 2109 cookies (and
1374             # since V0 cookies may be a mish-mash of Netscape and RFC 2109), we
1375             # accept either (but prefer Max-Age).
1376             max_age_set = False
1377
1378             bad_cookie = False
1379
1380             standard = {}
1381             rest = {}
1382             for k, v in cookie_attrs[1:]:
1383                 lc = k.lower()
1384                 # don't lose case distinction for unknown fields
1385                 if lc in value_attrs or lc in boolean_attrs:
1386                     k = lc
1387                 if k in boolean_attrs and v is None:
1388                     # boolean cookie-attribute is present, but has no value
1389                     # (like "discard", rather than "port=80")
1390                     v = True
1391                 if k in standard:
1392                     # only first value is significant
1393                     continue
1394                 if k == "domain":
1395                     if v is None:
1396                         _debug("   missing value for domain attribute")
1397                         bad_cookie = True
1398                         break
1399                     # RFC 2965 section 3.3.3
1400                     v = v.lower()
1401                 if k == "expires":
1402                     if max_age_set:
1403                         # Prefer max-age to expires (like Mozilla)
1404                         continue
1405                     if v is None:
1406                         _debug("   missing or invalid value for expires "
1407                               "attribute: treating as session cookie")
1408                         continue
1409                 if k == "max-age":
1410                     max_age_set = True
1411                     try:
1412                         v = int(v)
1413                     except ValueError:
1414                         _debug("   missing or invalid (non-numeric) value for "
1415                               "max-age attribute")
1416                         bad_cookie = True
1417                         break
1418                     # convert RFC 2965 Max-Age to seconds since epoch
1419                     # XXX Strictly you're supposed to follow RFC 2616
1420                     #   age-calculation rules.  Remember that zero Max-Age is a
1421                     #   is a request to discard (old and new) cookie, though.
1422                     k = "expires"
1423                     v = self._now + v
1424                 if (k in value_attrs) or (k in boolean_attrs):
1425                     if (v is None and
1426                         k not in ("port", "comment", "commenturl")):
1427                         _debug("   missing value for %s attribute" % k)
1428                         bad_cookie = True
1429                         break
1430                     standard[k] = v
1431                 else:
1432                     rest[k] = v
1433
1434             if bad_cookie:
1435                 continue
1436
1437             cookie_tuples.append((name, value, standard, rest))
1438
1439         return cookie_tuples
1440
1441     def _cookie_from_cookie_tuple(self, tup, request):
1442         # standard is dict of standard cookie-attributes, rest is dict of the
1443         # rest of them
1444         name, value, standard, rest = tup
1445
1446         domain = standard.get("domain", Absent)
1447         path = standard.get("path", Absent)
1448         port = standard.get("port", Absent)
1449         expires = standard.get("expires", Absent)
1450
1451         # set the easy defaults
1452         version = standard.get("version", None)
1453         if version is not None: version = int(version)
1454         secure = standard.get("secure", False)
1455         # (discard is also set if expires is Absent)
1456         discard = standard.get("discard", False)
1457         comment = standard.get("comment", None)
1458         comment_url = standard.get("commenturl", None)
1459
1460         # set default path
1461         if path is not Absent and path != "":
1462             path_specified = True
1463             path = escape_path(path)
1464         else:
1465             path_specified = False
1466             path = request_path(request)
1467             i = path.rfind("/")
1468             if i != -1:
1469                 if version == 0:
1470                     # Netscape spec parts company from reality here
1471                     path = path[:i]
1472                 else:
1473                     path = path[:i+1]
1474             if len(path) == 0: path = "/"
1475
1476         # set default domain
1477         domain_specified = domain is not Absent
1478         # but first we have to remember whether it starts with a dot
1479         domain_initial_dot = False
1480         if domain_specified:
1481             domain_initial_dot = bool(domain.startswith("."))
1482         if domain is Absent:
1483             req_host, erhn = eff_request_host(request)
1484             domain = erhn
1485         elif not domain.startswith("."):
1486             domain = "."+domain
1487
1488         # set default port
1489         port_specified = False
1490         if port is not Absent:
1491             if port is None:
1492                 # Port attr present, but has no value: default to request port.
1493                 # Cookie should then only be sent back on that port.
1494                 port = request_port(request)
1495             else:
1496                 port_specified = True
1497                 port = re.sub(r"\s+", "", port)
1498         else:
1499             # No port attr present.  Cookie can be sent back on any port.
1500             port = None
1501
1502         # set default expires and discard
1503         if expires is Absent:
1504             expires = None
1505             discard = True
1506         elif expires <= self._now:
1507             # Expiry date in past is request to delete cookie.  This can't be
1508             # in DefaultCookiePolicy, because can't delete cookies there.
1509             try:
1510                 self.clear(domain, path, name)
1511             except KeyError:
1512                 pass
1513             _debug("Expiring cookie, domain='%s', path='%s', name='%s'",
1514                    domain, path, name)
1515             return None
1516
1517         return Cookie(version,
1518                       name, value,
1519                       port, port_specified,
1520                       domain, domain_specified, domain_initial_dot,
1521                       path, path_specified,
1522                       secure,
1523                       expires,
1524                       discard,
1525                       comment,
1526                       comment_url,
1527                       rest)
1528
1529     def _cookies_from_attrs_set(self, attrs_set, request):
1530         cookie_tuples = self._normalized_cookie_tuples(attrs_set)
1531
1532         cookies = []
1533         for tup in cookie_tuples:
1534             cookie = self._cookie_from_cookie_tuple(tup, request)
1535             if cookie: cookies.append(cookie)
1536         return cookies
1537
1538     def _process_rfc2109_cookies(self, cookies):
1539         rfc2109_as_ns = getattr(self._policy, 'rfc2109_as_netscape', None)
1540         if rfc2109_as_ns is None:
1541             rfc2109_as_ns = not self._policy.rfc2965
1542         for cookie in cookies:
1543             if cookie.version == 1:
1544                 cookie.rfc2109 = True
1545                 if rfc2109_as_ns:
1546                     # treat 2109 cookies as Netscape cookies rather than
1547                     # as RFC2965 cookies
1548                     cookie.version = 0
1549
1550     def make_cookies(self, response, request):
1551         """Return sequence of Cookie objects extracted from response object."""
1552         # get cookie-attributes for RFC 2965 and Netscape protocols
1553         headers = response.info()
1554         rfc2965_hdrs = headers.getheaders("Set-Cookie2")
1555         ns_hdrs = headers.getheaders("Set-Cookie")
1556
1557         rfc2965 = self._policy.rfc2965
1558         netscape = self._policy.netscape
1559
1560         if ((not rfc2965_hdrs and not ns_hdrs) or
1561             (not ns_hdrs and not rfc2965) or
1562             (not rfc2965_hdrs and not netscape) or
1563             (not netscape and not rfc2965)):
1564             return []  # no relevant cookie headers: quick exit
1565
1566         try:
1567             cookies = self._cookies_from_attrs_set(
1568                 split_header_words(rfc2965_hdrs), request)
1569         except Exception:
1570             _warn_unhandled_exception()
1571             cookies = []
1572
1573         if ns_hdrs and netscape:
1574             try:
1575                 # RFC 2109 and Netscape cookies
1576                 ns_cookies = self._cookies_from_attrs_set(
1577                     parse_ns_headers(ns_hdrs), request)
1578             except Exception:
1579                 _warn_unhandled_exception()
1580                 ns_cookies = []
1581             self._process_rfc2109_cookies(ns_cookies)
1582
1583             # Look for Netscape cookies (from Set-Cookie headers) that match
1584             # corresponding RFC 2965 cookies (from Set-Cookie2 headers).
1585             # For each match, keep the RFC 2965 cookie and ignore the Netscape
1586             # cookie (RFC 2965 section 9.1).  Actually, RFC 2109 cookies are
1587             # bundled in with the Netscape cookies for this purpose, which is
1588             # reasonable behaviour.
1589             if rfc2965:
1590                 lookup = {}
1591                 for cookie in cookies:
1592                     lookup[(cookie.domain, cookie.path, cookie.name)] = None
1593
1594                 def no_matching_rfc2965(ns_cookie, lookup=lookup):
1595                     key = ns_cookie.domain, ns_cookie.path, ns_cookie.name
1596                     return key not in lookup
1597                 ns_cookies = filter(no_matching_rfc2965, ns_cookies)
1598
1599             if ns_cookies:
1600                 cookies.extend(ns_cookies)
1601
1602         return cookies
1603
1604     def set_cookie_if_ok(self, cookie, request):
1605         """Set a cookie if policy says it's OK to do so."""
1606         self._cookies_lock.acquire()
1607         try:
1608             self._policy._now = self._now = int(time.time())
1609
1610             if self._policy.set_ok(cookie, request):
1611                 self.set_cookie(cookie)
1612
1613
1614         finally:
1615             self._cookies_lock.release()
1616
1617     def set_cookie(self, cookie):
1618         """Set a cookie, without checking whether or not it should be set."""
1619         c = self._cookies
1620         self._cookies_lock.acquire()
1621         try:
1622             if cookie.domain not in c: c[cookie.domain] = {}
1623             c2 = c[cookie.domain]
1624             if cookie.path not in c2: c2[cookie.path] = {}
1625             c3 = c2[cookie.path]
1626             c3[cookie.name] = cookie
1627         finally:
1628             self._cookies_lock.release()
1629
1630     def extract_cookies(self, response, request):
1631         """Extract cookies from response, where allowable given the request."""
1632         _debug("extract_cookies: %s", response.info())
1633         self._cookies_lock.acquire()
1634         try:
1635             self._policy._now = self._now = int(time.time())
1636
1637             for cookie in self.make_cookies(response, request):
1638                 if self._policy.set_ok(cookie, request):
1639                     _debug(" setting cookie: %s", cookie)
1640                     self.set_cookie(cookie)
1641         finally:
1642             self._cookies_lock.release()
1643
1644     def clear(self, domain=None, path=None, name=None):
1645         """Clear some cookies.
1646
1647         Invoking this method without arguments will clear all cookies.  If
1648         given a single argument, only cookies belonging to that domain will be
1649         removed.  If given two arguments, cookies belonging to the specified
1650         path within that domain are removed.  If given three arguments, then
1651         the cookie with the specified name, path and domain is removed.
1652
1653         Raises KeyError if no matching cookie exists.
1654
1655         """
1656         if name is not None:
1657             if (domain is None) or (path is None):
1658                 raise ValueError(
1659                     "domain and path must be given to remove a cookie by name")
1660             del self._cookies[domain][path][name]
1661         elif path is not None:
1662             if domain is None:
1663                 raise ValueError(
1664                     "domain must be given to remove cookies by path")
1665             del self._cookies[domain][path]
1666         elif domain is not None:
1667             del self._cookies[domain]
1668         else:
1669             self._cookies = {}
1670
1671     def clear_session_cookies(self):
1672         """Discard all session cookies.
1673
1674         Note that the .save() method won't save session cookies anyway, unless
1675         you ask otherwise by passing a true ignore_discard argument.
1676
1677         """
1678         self._cookies_lock.acquire()
1679         try:
1680             for cookie in self:
1681                 if cookie.discard:
1682                     self.clear(cookie.domain, cookie.path, cookie.name)
1683         finally:
1684             self._cookies_lock.release()
1685
1686     def clear_expired_cookies(self):
1687         """Discard all expired cookies.
1688
1689         You probably don't need to call this method: expired cookies are never
1690         sent back to the server (provided you're using DefaultCookiePolicy),
1691         this method is called by CookieJar itself every so often, and the
1692         .save() method won't save expired cookies anyway (unless you ask
1693         otherwise by passing a true ignore_expires argument).
1694
1695         """
1696         self._cookies_lock.acquire()
1697         try:
1698             now = time.time()
1699             for cookie in self:
1700                 if cookie.is_expired(now):
1701                     self.clear(cookie.domain, cookie.path, cookie.name)
1702         finally:
1703             self._cookies_lock.release()
1704
1705     def __iter__(self):
1706         return deepvalues(self._cookies)
1707
1708     def __len__(self):
1709         """Return number of contained cookies."""
1710         i = 0
1711         for cookie in self: i = i + 1
1712         return i
1713
1714     def __repr__(self):
1715         r = []
1716         for cookie in self: r.append(repr(cookie))
1717         return "<%s[%s]>" % (self.__class__, ", ".join(r))
1718
1719     def __str__(self):
1720         r = []
1721         for cookie in self: r.append(str(cookie))
1722         return "<%s[%s]>" % (self.__class__, ", ".join(r))
1723
1724
1725 # derives from IOError for backwards-compatibility with Python 2.4.0
1726 class LoadError(IOError): pass
1727
1728 class FileCookieJar(CookieJar):
1729     """CookieJar that can be loaded from and saved to a file."""
1730
1731     def __init__(self, filename=None, delayload=False, policy=None):
1732         """
1733         Cookies are NOT loaded from the named file until either the .load() or
1734         .revert() method is called.
1735
1736         """
1737         CookieJar.__init__(self, policy)
1738         if filename is not None:
1739             try:
1740                 filename+""
1741             except:
1742                 raise ValueError("filename must be string-like")
1743         self.filename = filename
1744         self.delayload = bool(delayload)
1745
1746     def save(self, filename=None, ignore_discard=False, ignore_expires=False):
1747         """Save cookies to a file."""
1748         raise NotImplementedError()
1749
1750     def load(self, filename=None, ignore_discard=False, ignore_expires=False):
1751         """Load cookies from a file."""
1752         if filename is None:
1753             if self.filename is not None: filename = self.filename
1754             else: raise ValueError(MISSING_FILENAME_TEXT)
1755
1756         f = open(filename)
1757         try:
1758             self._really_load(f, filename, ignore_discard, ignore_expires)
1759         finally:
1760             f.close()
1761
1762     def revert(self, filename=None,
1763                ignore_discard=False, ignore_expires=False):
1764         """Clear all cookies and reload cookies from a saved file.
1765
1766         Raises LoadError (or IOError) if reversion is not successful; the
1767         object's state will not be altered if this happens.
1768
1769         """
1770         if filename is None:
1771             if self.filename is not None: filename = self.filename
1772             else: raise ValueError(MISSING_FILENAME_TEXT)
1773
1774         self._cookies_lock.acquire()
1775         try:
1776
1777             old_state = copy.deepcopy(self._cookies)
1778             self._cookies = {}
1779             try:
1780                 self.load(filename, ignore_discard, ignore_expires)
1781             except (LoadError, IOError):
1782                 self._cookies = old_state
1783                 raise
1784
1785         finally:
1786             self._cookies_lock.release()
1787
1788 from _LWPCookieJar import LWPCookieJar, lwp_cookie_str
1789 from _MozillaCookieJar import MozillaCookieJar