def getURLString(self, surt=False, public_suffix=False):
if None != self.opaque:
return self.opaque
if 'dns' == self.scheme:
s = self.scheme + ':' ###java version adds :// regardless of scheme
else: ###java version uses opaque type for dns urls, but this version supports dns urls
s = self.scheme + '://'
if surt:
s += "("
if self.authUser:
s += self.authUser
if self.authPass:
s += self.authPass
s += '@'
hostSrc = self.host
if public_suffix:
hostSrc = self.getPublicSuffix()
if surt:
hostSrc = hostToSURT(hostSrc)
s += hostSrc
if self.port != self.DEFAULT_PORT:
s += ":%d" % self.port
if surt:
s += ')'
hasPath = (None != self.path) and (len(self.path) > 0)
if hasPath:
s += self.path
else:
if (None != self.query) or (None != self.hash):
#must have '/' with query or hash:
s += '/'
if None != self.query:
s += '?' + self.query
if None != self.hash:
s += '#' + self.hash
if None != self.last_delimiter:
s += self.last_delimiter
return s
def getURLString(self, surt=False, public_suffix=False):
if None != self.opaque:
return self.opaque
if 'dns' == self.scheme:
s = self.scheme + ':' ###java version adds :// regardless of scheme
else: ###java version uses opaque type for dns urls, but this version supports dns urls
s = self.scheme + '://'
if surt:
s += "("
if self.authUser:
s += self.authUser
if self.authPass:
s += self.authPass
s += '@'
hostSrc = self.host
if public_suffix:
hostSrc = self.getPublicSuffix()
if surt:
hostSrc = hostToSURT(hostSrc)
s += hostSrc
if self.port != self.DEFAULT_PORT:
s += ":%d" % self.port
if surt:
s += ')'
hasPath = (None != self.path) and (len(self.path) > 0)
if hasPath:
s += self.path
else:
if (None != self.query) or (None != self.hash):
#must have '/' with query or hash:
s += '/'
if None != self.query:
s += '?' + self.query
if None != self.hash:
s += '#' + self.hash
if None != self.last_delimiter:
s += self.last_delimiter
return s
def surt(url):
"""
These doctests are from WaybackURLKeyMakerTest.java
>>> surt(None)
'-'
>>> surt('')
'-'
>>> surt("filedesc:foo.arc.gz")
'filedesc:foo.arc.gz'
>>> surt("filedesc:/foo.arc.gz")
'filedesc:/foo.arc.gz'
>>> surt("filedesc://foo.arc.gz")
'filedesc://foo.arc.gz'
>>> surt("warcinfo:foo.warc.gz")
'warcinfo:foo.warc.gz'
>>> surt("dns:alexa.com")
'com,alexa)'
>>> surt("dns:archive.org")
'org,archive)'
>>> surt("http://www.archive.org/")
'org,archive)/'
>>> surt("http://archive.org/")
'org,archive)/'
>>> surt("http://archive.org/goo/")
'org,archive)/goo'
>>> surt("http://archive.org/goo/?")
'org,archive)/goo'
>>> surt("http://archive.org/goo/?b&a")
'org,archive)/goo?a&b'
>>> surt("http://archive.org/goo/?a=2&b&a=1")
'org,archive)/goo?a=1&a=2&b'
PHP session id:
>>> surt("http://archive.org/index.php?PHPSESSID=0123456789abcdefghijklemopqrstuv&action=profile;u=4221")
'org,archive)/index.php?action=profile;u=4221'
WHOIS url:
>>> surt("whois://whois.isoc.org.il/shaveh.co.il")
'whois://whois.isoc.org.il/shaveh.co.il'
Yahoo web bug. See https://github.com/internetarchive/surt/issues/1
>>> surt('http://visit.webhosting.yahoo.com/visit.gif?&r=http%3A//web.archive.org/web/20090517140029/http%3A//anthonystewarthead.electric-chi.com/&b=Netscape%205.0%20%28Windows%3B%20en-US%29&s=1366x768&o=Win32&c=24&j=true&v=1.2')
'com,yahoo,webhosting,visit)/visit.gif?&b=netscape%205.0%20(windows;%20en-us)&c=24&j=true&o=win32&r=http://web.archive.org/web/20090517140029/http://anthonystewarthead.electric-chi.com/&s=1366x768&v=1.2'
"""
if not url:
return "-"
if url.startswith("filedesc"):
return url
if url.startswith("warcinfo"):
return url
if url.startswith("dns:"):
return hostToSURT(url[4:]) + ')'
if url.startswith("whois://"):
return url
hurl = canonicalize(handyurl.parse(url))
key = hurl.getURLString(surt=True)
parenIdx = key.find('(')
if -1 == parenIdx:
return url #something very wrong
return key[parenIdx + 1:]
def surt(url):
"""
These doctests are from WaybackURLKeyMakerTest.java
>>> surt(None)
'-'
>>> surt('')
'-'
>>> surt("filedesc:foo.arc.gz")
'filedesc:foo.arc.gz'
>>> surt("filedesc:/foo.arc.gz")
'filedesc:/foo.arc.gz'
>>> surt("filedesc://foo.arc.gz")
'filedesc://foo.arc.gz'
>>> surt("warcinfo:foo.warc.gz")
'warcinfo:foo.warc.gz'
>>> surt("dns:alexa.com")
'com,alexa)'
>>> surt("dns:archive.org")
'org,archive)'
>>> surt("http://www.archive.org/")
'org,archive)/'
>>> surt("http://archive.org/")
'org,archive)/'
>>> surt("http://archive.org/goo/")
'org,archive)/goo'
>>> surt("http://archive.org/goo/?")
'org,archive)/goo'
>>> surt("http://archive.org/goo/?b&a")
'org,archive)/goo?a&b'
>>> surt("http://archive.org/goo/?a=2&b&a=1")
'org,archive)/goo?a=1&a=2&b'
PHP session id:
>>> surt("http://archive.org/index.php?PHPSESSID=0123456789abcdefghijklemopqrstuv&action=profile;u=4221")
'org,archive)/index.php?action=profile;u=4221'
WHOIS url:
>>> surt("whois://whois.isoc.org.il/shaveh.co.il")
'whois://whois.isoc.org.il/shaveh.co.il'
Yahoo web bug. See https://github.com/internetarchive/surt/issues/1
>>> surt('http://visit.webhosting.yahoo.com/visit.gif?&r=http%3A//web.archive.org/web/20090517140029/http%3A//anthonystewarthead.electric-chi.com/&b=Netscape%205.0%20%28Windows%3B%20en-US%29&s=1366x768&o=Win32&c=24&j=true&v=1.2')
'com,yahoo,webhosting,visit)/visit.gif?&b=netscape%205.0%20(windows;%20en-us)&c=24&j=true&o=win32&r=http://web.archive.org/web/20090517140029/http://anthonystewarthead.electric-chi.com/&s=1366x768&v=1.2'
"""
if not url:
return "-"
if url.startswith("filedesc"):
return url
if url.startswith("warcinfo"):
return url
if url.startswith("dns:"):
return hostToSURT(url[4:]) + ')'
if url.startswith("whois://"):
return url
hurl = canonicalize(handyurl.parse(url))
key = hurl.getURLString(surt=True)
parenIdx = key.find('(')
if -1 == parenIdx:
return url #something very wrong
return key[parenIdx+1:]
