mirror of
https://github.com/SickGear/SickGear.git
synced 2024-12-01 00:43:37 +00:00
111 lines
3.8 KiB
Python
111 lines
3.8 KiB
Python
|
"""
|
||
|
iri2uri
|
||
|
|
||
|
Converts an IRI to a URI.
|
||
|
|
||
|
"""
|
||
|
__author__ = "Joe Gregorio (joe@bitworking.org)"
|
||
|
__copyright__ = "Copyright 2006, Joe Gregorio"
|
||
|
__contributors__ = []
|
||
|
__version__ = "1.0.0"
|
||
|
__license__ = "MIT"
|
||
|
__history__ = """
|
||
|
"""
|
||
|
|
||
|
import urlparse
|
||
|
|
||
|
|
||
|
# Convert an IRI to a URI following the rules in RFC 3987
|
||
|
#
|
||
|
# The characters we need to enocde and escape are defined in the spec:
|
||
|
#
|
||
|
# iprivate = %xE000-F8FF / %xF0000-FFFFD / %x100000-10FFFD
|
||
|
# ucschar = %xA0-D7FF / %xF900-FDCF / %xFDF0-FFEF
|
||
|
# / %x10000-1FFFD / %x20000-2FFFD / %x30000-3FFFD
|
||
|
# / %x40000-4FFFD / %x50000-5FFFD / %x60000-6FFFD
|
||
|
# / %x70000-7FFFD / %x80000-8FFFD / %x90000-9FFFD
|
||
|
# / %xA0000-AFFFD / %xB0000-BFFFD / %xC0000-CFFFD
|
||
|
# / %xD0000-DFFFD / %xE1000-EFFFD
|
||
|
|
||
|
escape_range = [
|
||
|
(0xA0, 0xD7FF ),
|
||
|
(0xE000, 0xF8FF ),
|
||
|
(0xF900, 0xFDCF ),
|
||
|
(0xFDF0, 0xFFEF),
|
||
|
(0x10000, 0x1FFFD ),
|
||
|
(0x20000, 0x2FFFD ),
|
||
|
(0x30000, 0x3FFFD),
|
||
|
(0x40000, 0x4FFFD ),
|
||
|
(0x50000, 0x5FFFD ),
|
||
|
(0x60000, 0x6FFFD),
|
||
|
(0x70000, 0x7FFFD ),
|
||
|
(0x80000, 0x8FFFD ),
|
||
|
(0x90000, 0x9FFFD),
|
||
|
(0xA0000, 0xAFFFD ),
|
||
|
(0xB0000, 0xBFFFD ),
|
||
|
(0xC0000, 0xCFFFD),
|
||
|
(0xD0000, 0xDFFFD ),
|
||
|
(0xE1000, 0xEFFFD),
|
||
|
(0xF0000, 0xFFFFD ),
|
||
|
(0x100000, 0x10FFFD)
|
||
|
]
|
||
|
|
||
|
def encode(c):
|
||
|
retval = c
|
||
|
i = ord(c)
|
||
|
for low, high in escape_range:
|
||
|
if i < low:
|
||
|
break
|
||
|
if i >= low and i <= high:
|
||
|
retval = "".join(["%%%2X" % ord(o) for o in c.encode('utf-8')])
|
||
|
break
|
||
|
return retval
|
||
|
|
||
|
|
||
|
def iri2uri(uri):
|
||
|
"""Convert an IRI to a URI. Note that IRIs must be
|
||
|
passed in a unicode strings. That is, do not utf-8 encode
|
||
|
the IRI before passing it into the function."""
|
||
|
if isinstance(uri ,unicode):
|
||
|
(scheme, authority, path, query, fragment) = urlparse.urlsplit(uri)
|
||
|
authority = authority.encode('idna')
|
||
|
# For each character in 'ucschar' or 'iprivate'
|
||
|
# 1. encode as utf-8
|
||
|
# 2. then %-encode each octet of that utf-8
|
||
|
uri = urlparse.urlunsplit((scheme, authority, path, query, fragment))
|
||
|
uri = "".join([encode(c) for c in uri])
|
||
|
return uri
|
||
|
|
||
|
if __name__ == "__main__":
|
||
|
import unittest
|
||
|
|
||
|
class Test(unittest.TestCase):
|
||
|
|
||
|
def test_uris(self):
|
||
|
"""Test that URIs are invariant under the transformation."""
|
||
|
invariant = [
|
||
|
u"ftp://ftp.is.co.za/rfc/rfc1808.txt",
|
||
|
u"http://www.ietf.org/rfc/rfc2396.txt",
|
||
|
u"ldap://[2001:db8::7]/c=GB?objectClass?one",
|
||
|
u"mailto:John.Doe@example.com",
|
||
|
u"news:comp.infosystems.www.servers.unix",
|
||
|
u"tel:+1-816-555-1212",
|
||
|
u"telnet://192.0.2.16:80/",
|
||
|
u"urn:oasis:names:specification:docbook:dtd:xml:4.1.2" ]
|
||
|
for uri in invariant:
|
||
|
self.assertEqual(uri, iri2uri(uri))
|
||
|
|
||
|
def test_iri(self):
|
||
|
""" Test that the right type of escaping is done for each part of the URI."""
|
||
|
self.assertEqual("http://xn--o3h.com/%E2%98%84", iri2uri(u"http://\N{COMET}.com/\N{COMET}"))
|
||
|
self.assertEqual("http://bitworking.org/?fred=%E2%98%84", iri2uri(u"http://bitworking.org/?fred=\N{COMET}"))
|
||
|
self.assertEqual("http://bitworking.org/#%E2%98%84", iri2uri(u"http://bitworking.org/#\N{COMET}"))
|
||
|
self.assertEqual("#%E2%98%84", iri2uri(u"#\N{COMET}"))
|
||
|
self.assertEqual("/fred?bar=%E2%98%9A#%E2%98%84", iri2uri(u"/fred?bar=\N{BLACK LEFT POINTING INDEX}#\N{COMET}"))
|
||
|
self.assertEqual("/fred?bar=%E2%98%9A#%E2%98%84", iri2uri(iri2uri(u"/fred?bar=\N{BLACK LEFT POINTING INDEX}#\N{COMET}")))
|
||
|
self.assertNotEqual("/fred?bar=%E2%98%9A#%E2%98%84", iri2uri(u"/fred?bar=\N{BLACK LEFT POINTING INDEX}#\N{COMET}".encode('utf-8')))
|
||
|
|
||
|
unittest.main()
|
||
|
|
||
|
|