1# -*- coding: utf-8 -*- 2"""Converts an IRI to a URI.""" 3 4__author__ = "Joe Gregorio (joe@bitworking.org)" 5__copyright__ = "Copyright 2006, Joe Gregorio" 6__contributors__ = [] 7__version__ = "1.0.0" 8__license__ = "MIT" 9 10import urllib.parse 11 12# Convert an IRI to a URI following the rules in RFC 3987 13# 14# The characters we need to enocde and escape are defined in the spec: 15# 16# iprivate = %xE000-F8FF / %xF0000-FFFFD / %x100000-10FFFD 17# ucschar = %xA0-D7FF / %xF900-FDCF / %xFDF0-FFEF 18# / %x10000-1FFFD / %x20000-2FFFD / %x30000-3FFFD 19# / %x40000-4FFFD / %x50000-5FFFD / %x60000-6FFFD 20# / %x70000-7FFFD / %x80000-8FFFD / %x90000-9FFFD 21# / %xA0000-AFFFD / %xB0000-BFFFD / %xC0000-CFFFD 22# / %xD0000-DFFFD / %xE1000-EFFFD 23 24escape_range = [ 25 (0xA0, 0xD7FF), 26 (0xE000, 0xF8FF), 27 (0xF900, 0xFDCF), 28 (0xFDF0, 0xFFEF), 29 (0x10000, 0x1FFFD), 30 (0x20000, 0x2FFFD), 31 (0x30000, 0x3FFFD), 32 (0x40000, 0x4FFFD), 33 (0x50000, 0x5FFFD), 34 (0x60000, 0x6FFFD), 35 (0x70000, 0x7FFFD), 36 (0x80000, 0x8FFFD), 37 (0x90000, 0x9FFFD), 38 (0xA0000, 0xAFFFD), 39 (0xB0000, 0xBFFFD), 40 (0xC0000, 0xCFFFD), 41 (0xD0000, 0xDFFFD), 42 (0xE1000, 0xEFFFD), 43 (0xF0000, 0xFFFFD), 44 (0x100000, 0x10FFFD), 45] 46 47 48def encode(c): 49 retval = c 50 i = ord(c) 51 for low, high in escape_range: 52 if i < low: 53 break 54 if i >= low and i <= high: 55 retval = "".join(["%%%2X" % o for o in c.encode("utf-8")]) 56 break 57 return retval 58 59 60def iri2uri(uri): 61 """Convert an IRI to a URI. Note that IRIs must be 62 passed in a unicode strings. That is, do not utf-8 encode 63 the IRI before passing it into the function.""" 64 if isinstance(uri, str): 65 (scheme, authority, path, query, fragment) = urllib.parse.urlsplit(uri) 66 authority = authority.encode("idna").decode("utf-8") 67 # For each character in 'ucschar' or 'iprivate' 68 # 1. encode as utf-8 69 # 2. then %-encode each octet of that utf-8 70 uri = urllib.parse.urlunsplit((scheme, authority, path, query, fragment)) 71 uri = "".join([encode(c) for c in uri]) 72 return uri 73 74 75if __name__ == "__main__": 76 import unittest 77 78 class Test(unittest.TestCase): 79 def test_uris(self): 80 """Test that URIs are invariant under the transformation.""" 81 invariant = [ 82 "ftp://ftp.is.co.za/rfc/rfc1808.txt", 83 "http://www.ietf.org/rfc/rfc2396.txt", 84 "ldap://[2001:db8::7]/c=GB?objectClass?one", 85 "mailto:John.Doe@example.com", 86 "news:comp.infosystems.www.servers.unix", 87 "tel:+1-816-555-1212", 88 "telnet://192.0.2.16:80/", 89 "urn:oasis:names:specification:docbook:dtd:xml:4.1.2", 90 ] 91 for uri in invariant: 92 self.assertEqual(uri, iri2uri(uri)) 93 94 def test_iri(self): 95 """Test that the right type of escaping is done for each part of the URI.""" 96 self.assertEqual( 97 "http://xn--o3h.com/%E2%98%84", 98 iri2uri("http://\N{COMET}.com/\N{COMET}"), 99 ) 100 self.assertEqual( 101 "http://bitworking.org/?fred=%E2%98%84", 102 iri2uri("http://bitworking.org/?fred=\N{COMET}"), 103 ) 104 self.assertEqual( 105 "http://bitworking.org/#%E2%98%84", 106 iri2uri("http://bitworking.org/#\N{COMET}"), 107 ) 108 self.assertEqual("#%E2%98%84", iri2uri("#\N{COMET}")) 109 self.assertEqual( 110 "/fred?bar=%E2%98%9A#%E2%98%84", 111 iri2uri("/fred?bar=\N{BLACK LEFT POINTING INDEX}#\N{COMET}"), 112 ) 113 self.assertEqual( 114 "/fred?bar=%E2%98%9A#%E2%98%84", 115 iri2uri(iri2uri("/fred?bar=\N{BLACK LEFT POINTING INDEX}#\N{COMET}")), 116 ) 117 self.assertNotEqual( 118 "/fred?bar=%E2%98%9A#%E2%98%84", 119 iri2uri( 120 "/fred?bar=\N{BLACK LEFT POINTING INDEX}#\N{COMET}".encode("utf-8") 121 ), 122 ) 123 124 unittest.main() 125