Mercurial > repos > guerler > springsuite
annotate planemo/lib/python3.7/site-packages/bs4/tests/test_soup.py @ 0:d30785e31577 draft
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
author | guerler |
---|---|
date | Fri, 31 Jul 2020 00:18:57 -0400 |
parents | |
children |
rev | line source |
---|---|
0
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
1 # -*- coding: utf-8 -*- |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
2 """Tests of Beautiful Soup as a whole.""" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
3 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
4 from pdb import set_trace |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
5 import logging |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
6 import unittest |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
7 import sys |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
8 import tempfile |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
9 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
10 from bs4 import ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
11 BeautifulSoup, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
12 BeautifulStoneSoup, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
13 GuessedAtParserWarning, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
14 MarkupResemblesLocatorWarning, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
15 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
16 from bs4.builder import ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
17 TreeBuilder, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
18 ParserRejectedMarkup, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
19 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
20 from bs4.element import ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
21 CharsetMetaAttributeValue, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
22 Comment, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
23 ContentMetaAttributeValue, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
24 SoupStrainer, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
25 NamespacedAttribute, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
26 Tag, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
27 NavigableString, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
28 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
29 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
30 import bs4.dammit |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
31 from bs4.dammit import ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
32 EntitySubstitution, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
33 UnicodeDammit, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
34 EncodingDetector, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
35 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
36 from bs4.testing import ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
37 default_builder, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
38 SoupTest, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
39 skipIf, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
40 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
41 import warnings |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
42 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
43 try: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
44 from bs4.builder import LXMLTreeBuilder, LXMLTreeBuilderForXML |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
45 LXML_PRESENT = True |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
46 except ImportError as e: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
47 LXML_PRESENT = False |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
48 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
49 PYTHON_3_PRE_3_2 = (sys.version_info[0] == 3 and sys.version_info < (3,2)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
50 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
51 class TestConstructor(SoupTest): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
52 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
53 def test_short_unicode_input(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
54 data = "<h1>éé</h1>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
55 soup = self.soup(data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
56 self.assertEqual("éé", soup.h1.string) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
57 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
58 def test_embedded_null(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
59 data = "<h1>foo\0bar</h1>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
60 soup = self.soup(data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
61 self.assertEqual("foo\0bar", soup.h1.string) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
62 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
63 def test_exclude_encodings(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
64 utf8_data = "Räksmörgås".encode("utf-8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
65 soup = self.soup(utf8_data, exclude_encodings=["utf-8"]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
66 self.assertEqual("windows-1252", soup.original_encoding) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
67 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
68 def test_custom_builder_class(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
69 # Verify that you can pass in a custom Builder class and |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
70 # it'll be instantiated with the appropriate keyword arguments. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
71 class Mock(object): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
72 def __init__(self, **kwargs): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
73 self.called_with = kwargs |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
74 self.is_xml = True |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
75 self.store_line_numbers = False |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
76 self.cdata_list_attributes = [] |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
77 self.preserve_whitespace_tags = [] |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
78 self.string_containers = {} |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
79 def initialize_soup(self, soup): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
80 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
81 def feed(self, markup): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
82 self.fed = markup |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
83 def reset(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
84 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
85 def ignore(self, ignore): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
86 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
87 set_up_substitutions = can_be_empty_element = ignore |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
88 def prepare_markup(self, *args, **kwargs): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
89 yield "prepared markup", "original encoding", "declared encoding", "contains replacement characters" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
90 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
91 kwargs = dict( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
92 var="value", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
93 # This is a deprecated BS3-era keyword argument, which |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
94 # will be stripped out. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
95 convertEntities=True, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
96 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
97 with warnings.catch_warnings(record=True): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
98 soup = BeautifulSoup('', builder=Mock, **kwargs) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
99 assert isinstance(soup.builder, Mock) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
100 self.assertEqual(dict(var="value"), soup.builder.called_with) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
101 self.assertEqual("prepared markup", soup.builder.fed) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
102 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
103 # You can also instantiate the TreeBuilder yourself. In this |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
104 # case, that specific object is used and any keyword arguments |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
105 # to the BeautifulSoup constructor are ignored. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
106 builder = Mock(**kwargs) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
107 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
108 soup = BeautifulSoup( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
109 '', builder=builder, ignored_value=True, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
110 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
111 msg = str(w[0].message) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
112 assert msg.startswith("Keyword arguments to the BeautifulSoup constructor will be ignored.") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
113 self.assertEqual(builder, soup.builder) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
114 self.assertEqual(kwargs, builder.called_with) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
115 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
116 def test_parser_markup_rejection(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
117 # If markup is completely rejected by the parser, an |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
118 # explanatory ParserRejectedMarkup exception is raised. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
119 class Mock(TreeBuilder): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
120 def feed(self, *args, **kwargs): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
121 raise ParserRejectedMarkup("Nope.") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
122 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
123 def prepare_markup(self, *args, **kwargs): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
124 # We're going to try two different ways of preparing this markup, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
125 # but feed() will reject both of them. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
126 yield markup, None, None, False |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
127 yield markup, None, None, False |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
128 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
129 import re |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
130 self.assertRaisesRegex( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
131 ParserRejectedMarkup, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
132 "The markup you provided was rejected by the parser. Trying a different parser or a different encoding may help.", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
133 BeautifulSoup, '', builder=Mock, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
134 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
135 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
136 def test_cdata_list_attributes(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
137 # Most attribute values are represented as scalars, but the |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
138 # HTML standard says that some attributes, like 'class' have |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
139 # space-separated lists as values. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
140 markup = '<a id=" an id " class=" a class "></a>' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
141 soup = self.soup(markup) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
142 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
143 # Note that the spaces are stripped for 'class' but not for 'id'. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
144 a = soup.a |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
145 self.assertEqual(" an id ", a['id']) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
146 self.assertEqual(["a", "class"], a['class']) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
147 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
148 # TreeBuilder takes an argument called 'mutli_valued_attributes' which lets |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
149 # you customize or disable this. As always, you can customize the TreeBuilder |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
150 # by passing in a keyword argument to the BeautifulSoup constructor. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
151 soup = self.soup(markup, builder=default_builder, multi_valued_attributes=None) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
152 self.assertEqual(" a class ", soup.a['class']) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
153 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
154 # Here are two ways of saying that `id` is a multi-valued |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
155 # attribute in this context, but 'class' is not. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
156 for switcheroo in ({'*': 'id'}, {'a': 'id'}): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
157 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
158 # This will create a warning about not explicitly |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
159 # specifying a parser, but we'll ignore it. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
160 soup = self.soup(markup, builder=None, multi_valued_attributes=switcheroo) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
161 a = soup.a |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
162 self.assertEqual(["an", "id"], a['id']) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
163 self.assertEqual(" a class ", a['class']) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
164 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
165 def test_replacement_classes(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
166 # Test the ability to pass in replacements for element classes |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
167 # which will be used when building the tree. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
168 class TagPlus(Tag): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
169 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
170 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
171 class StringPlus(NavigableString): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
172 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
173 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
174 class CommentPlus(Comment): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
175 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
176 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
177 soup = self.soup( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
178 "<a><b>foo</b>bar</a><!--whee-->", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
179 element_classes = { |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
180 Tag: TagPlus, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
181 NavigableString: StringPlus, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
182 Comment: CommentPlus, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
183 } |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
184 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
185 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
186 # The tree was built with TagPlus, StringPlus, and CommentPlus objects, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
187 # rather than Tag, String, and Comment objects. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
188 assert all( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
189 isinstance(x, (TagPlus, StringPlus, CommentPlus)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
190 for x in soup.recursiveChildGenerator() |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
191 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
192 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
193 def test_alternate_string_containers(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
194 # Test the ability to customize the string containers for |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
195 # different types of tags. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
196 class PString(NavigableString): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
197 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
198 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
199 class BString(NavigableString): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
200 pass |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
201 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
202 soup = self.soup( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
203 "<div>Hello.<p>Here is <b>some <i>bolded</i></b> text", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
204 string_containers = { |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
205 'b': BString, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
206 'p': PString, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
207 } |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
208 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
209 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
210 # The string before the <p> tag is a regular NavigableString. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
211 assert isinstance(soup.div.contents[0], NavigableString) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
212 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
213 # The string inside the <p> tag, but not inside the <i> tag, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
214 # is a PString. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
215 assert isinstance(soup.p.contents[0], PString) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
216 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
217 # Every string inside the <b> tag is a BString, even the one that |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
218 # was also inside an <i> tag. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
219 for s in soup.b.strings: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
220 assert isinstance(s, BString) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
221 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
222 # Now that parsing was complete, the string_container_stack |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
223 # (where this information was kept) has been cleared out. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
224 self.assertEqual([], soup.string_container_stack) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
225 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
226 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
227 class TestWarnings(SoupTest): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
228 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
229 def _assert_warning(self, warnings, cls): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
230 for w in warnings: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
231 if isinstance(w.message, cls): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
232 return w |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
233 raise Exception("%s warning not found in %r" % cls, warnings) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
234 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
235 def _assert_no_parser_specified(self, w): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
236 warning = self._assert_warning(w, GuessedAtParserWarning) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
237 message = str(warning.message) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
238 self.assertTrue( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
239 message.startswith(BeautifulSoup.NO_PARSER_SPECIFIED_WARNING[:60]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
240 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
241 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
242 def test_warning_if_no_parser_specified(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
243 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
244 soup = BeautifulSoup("<a><b></b></a>") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
245 self._assert_no_parser_specified(w) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
246 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
247 def test_warning_if_parser_specified_too_vague(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
248 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
249 soup = BeautifulSoup("<a><b></b></a>", "html") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
250 self._assert_no_parser_specified(w) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
251 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
252 def test_no_warning_if_explicit_parser_specified(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
253 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
254 soup = BeautifulSoup("<a><b></b></a>", "html.parser") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
255 self.assertEqual([], w) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
256 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
257 def test_parseOnlyThese_renamed_to_parse_only(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
258 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
259 soup = self.soup("<a><b></b></a>", parseOnlyThese=SoupStrainer("b")) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
260 msg = str(w[0].message) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
261 self.assertTrue("parseOnlyThese" in msg) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
262 self.assertTrue("parse_only" in msg) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
263 self.assertEqual(b"<b></b>", soup.encode()) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
264 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
265 def test_fromEncoding_renamed_to_from_encoding(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
266 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
267 utf8 = b"\xc3\xa9" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
268 soup = self.soup(utf8, fromEncoding="utf8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
269 msg = str(w[0].message) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
270 self.assertTrue("fromEncoding" in msg) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
271 self.assertTrue("from_encoding" in msg) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
272 self.assertEqual("utf8", soup.original_encoding) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
273 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
274 def test_unrecognized_keyword_argument(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
275 self.assertRaises( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
276 TypeError, self.soup, "<a>", no_such_argument=True) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
277 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
278 def test_disk_file_warning(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
279 filehandle = tempfile.NamedTemporaryFile() |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
280 filename = filehandle.name |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
281 try: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
282 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
283 soup = self.soup(filename) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
284 warning = self._assert_warning(w, MarkupResemblesLocatorWarning) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
285 self.assertTrue("looks like a filename" in str(warning.message)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
286 finally: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
287 filehandle.close() |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
288 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
289 # The file no longer exists, so Beautiful Soup will no longer issue the warning. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
290 with warnings.catch_warnings(record=True) as w: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
291 soup = self.soup(filename) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
292 self.assertEqual([], w) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
293 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
294 def test_url_warning_with_bytes_url(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
295 with warnings.catch_warnings(record=True) as warning_list: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
296 soup = self.soup(b"http://www.crummybytes.com/") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
297 warning = self._assert_warning( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
298 warning_list, MarkupResemblesLocatorWarning |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
299 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
300 self.assertTrue("looks like a URL" in str(warning.message)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
301 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
302 def test_url_warning_with_unicode_url(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
303 with warnings.catch_warnings(record=True) as warning_list: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
304 # note - this url must differ from the bytes one otherwise |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
305 # python's warnings system swallows the second warning |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
306 soup = self.soup("http://www.crummyunicode.com/") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
307 warning = self._assert_warning( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
308 warning_list, MarkupResemblesLocatorWarning |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
309 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
310 self.assertTrue("looks like a URL" in str(warning.message)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
311 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
312 def test_url_warning_with_bytes_and_space(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
313 # Here the markup contains something besides a URL, so no warning |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
314 # is issued. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
315 with warnings.catch_warnings(record=True) as warning_list: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
316 soup = self.soup(b"http://www.crummybytes.com/ is great") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
317 self.assertFalse(any("looks like a URL" in str(w.message) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
318 for w in warning_list)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
319 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
320 def test_url_warning_with_unicode_and_space(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
321 with warnings.catch_warnings(record=True) as warning_list: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
322 soup = self.soup("http://www.crummyuncode.com/ is great") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
323 self.assertFalse(any("looks like a URL" in str(w.message) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
324 for w in warning_list)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
325 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
326 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
327 class TestSelectiveParsing(SoupTest): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
328 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
329 def test_parse_with_soupstrainer(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
330 markup = "No<b>Yes</b><a>No<b>Yes <c>Yes</c></b>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
331 strainer = SoupStrainer("b") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
332 soup = self.soup(markup, parse_only=strainer) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
333 self.assertEqual(soup.encode(), b"<b>Yes</b><b>Yes <c>Yes</c></b>") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
334 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
335 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
336 class TestEntitySubstitution(unittest.TestCase): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
337 """Standalone tests of the EntitySubstitution class.""" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
338 def setUp(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
339 self.sub = EntitySubstitution |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
340 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
341 def test_simple_html_substitution(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
342 # Unicode characters corresponding to named HTML entites |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
343 # are substituted, and no others. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
344 s = "foo\u2200\N{SNOWMAN}\u00f5bar" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
345 self.assertEqual(self.sub.substitute_html(s), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
346 "foo∀\N{SNOWMAN}õbar") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
347 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
348 def test_smart_quote_substitution(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
349 # MS smart quotes are a common source of frustration, so we |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
350 # give them a special test. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
351 quotes = b"\x91\x92foo\x93\x94" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
352 dammit = UnicodeDammit(quotes) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
353 self.assertEqual(self.sub.substitute_html(dammit.markup), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
354 "‘’foo“”") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
355 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
356 def test_xml_converstion_includes_no_quotes_if_make_quoted_attribute_is_false(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
357 s = 'Welcome to "my bar"' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
358 self.assertEqual(self.sub.substitute_xml(s, False), s) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
359 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
360 def test_xml_attribute_quoting_normally_uses_double_quotes(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
361 self.assertEqual(self.sub.substitute_xml("Welcome", True), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
362 '"Welcome"') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
363 self.assertEqual(self.sub.substitute_xml("Bob's Bar", True), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
364 '"Bob\'s Bar"') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
365 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
366 def test_xml_attribute_quoting_uses_single_quotes_when_value_contains_double_quotes(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
367 s = 'Welcome to "my bar"' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
368 self.assertEqual(self.sub.substitute_xml(s, True), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
369 "'Welcome to \"my bar\"'") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
370 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
371 def test_xml_attribute_quoting_escapes_single_quotes_when_value_contains_both_single_and_double_quotes(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
372 s = 'Welcome to "Bob\'s Bar"' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
373 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
374 self.sub.substitute_xml(s, True), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
375 '"Welcome to "Bob\'s Bar""') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
376 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
377 def test_xml_quotes_arent_escaped_when_value_is_not_being_quoted(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
378 quoted = 'Welcome to "Bob\'s Bar"' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
379 self.assertEqual(self.sub.substitute_xml(quoted), quoted) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
380 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
381 def test_xml_quoting_handles_angle_brackets(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
382 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
383 self.sub.substitute_xml("foo<bar>"), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
384 "foo<bar>") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
385 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
386 def test_xml_quoting_handles_ampersands(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
387 self.assertEqual(self.sub.substitute_xml("AT&T"), "AT&T") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
388 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
389 def test_xml_quoting_including_ampersands_when_they_are_part_of_an_entity(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
390 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
391 self.sub.substitute_xml("ÁT&T"), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
392 "&Aacute;T&T") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
393 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
394 def test_xml_quoting_ignoring_ampersands_when_they_are_part_of_an_entity(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
395 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
396 self.sub.substitute_xml_containing_entities("ÁT&T"), |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
397 "ÁT&T") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
398 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
399 def test_quotes_not_html_substituted(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
400 """There's no need to do this except inside attribute values.""" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
401 text = 'Bob\'s "bar"' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
402 self.assertEqual(self.sub.substitute_html(text), text) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
403 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
404 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
405 class TestEncodingConversion(SoupTest): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
406 # Test Beautiful Soup's ability to decode and encode from various |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
407 # encodings. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
408 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
409 def setUp(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
410 super(TestEncodingConversion, self).setUp() |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
411 self.unicode_data = '<html><head><meta charset="utf-8"/></head><body><foo>Sacr\N{LATIN SMALL LETTER E WITH ACUTE} bleu!</foo></body></html>' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
412 self.utf8_data = self.unicode_data.encode("utf-8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
413 # Just so you know what it looks like. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
414 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
415 self.utf8_data, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
416 b'<html><head><meta charset="utf-8"/></head><body><foo>Sacr\xc3\xa9 bleu!</foo></body></html>') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
417 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
418 def test_ascii_in_unicode_out(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
419 # ASCII input is converted to Unicode. The original_encoding |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
420 # attribute is set to 'utf-8', a superset of ASCII. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
421 chardet = bs4.dammit.chardet_dammit |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
422 logging.disable(logging.WARNING) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
423 try: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
424 def noop(str): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
425 return None |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
426 # Disable chardet, which will realize that the ASCII is ASCII. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
427 bs4.dammit.chardet_dammit = noop |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
428 ascii = b"<foo>a</foo>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
429 soup_from_ascii = self.soup(ascii) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
430 unicode_output = soup_from_ascii.decode() |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
431 self.assertTrue(isinstance(unicode_output, str)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
432 self.assertEqual(unicode_output, self.document_for(ascii.decode())) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
433 self.assertEqual(soup_from_ascii.original_encoding.lower(), "utf-8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
434 finally: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
435 logging.disable(logging.NOTSET) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
436 bs4.dammit.chardet_dammit = chardet |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
437 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
438 def test_unicode_in_unicode_out(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
439 # Unicode input is left alone. The original_encoding attribute |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
440 # is not set. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
441 soup_from_unicode = self.soup(self.unicode_data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
442 self.assertEqual(soup_from_unicode.decode(), self.unicode_data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
443 self.assertEqual(soup_from_unicode.foo.string, 'Sacr\xe9 bleu!') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
444 self.assertEqual(soup_from_unicode.original_encoding, None) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
445 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
446 def test_utf8_in_unicode_out(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
447 # UTF-8 input is converted to Unicode. The original_encoding |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
448 # attribute is set. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
449 soup_from_utf8 = self.soup(self.utf8_data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
450 self.assertEqual(soup_from_utf8.decode(), self.unicode_data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
451 self.assertEqual(soup_from_utf8.foo.string, 'Sacr\xe9 bleu!') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
452 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
453 def test_utf8_out(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
454 # The internal data structures can be encoded as UTF-8. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
455 soup_from_unicode = self.soup(self.unicode_data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
456 self.assertEqual(soup_from_unicode.encode('utf-8'), self.utf8_data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
457 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
458 @skipIf( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
459 PYTHON_3_PRE_3_2, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
460 "Bad HTMLParser detected; skipping test of non-ASCII characters in attribute name.") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
461 def test_attribute_name_containing_unicode_characters(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
462 markup = '<div><a \N{SNOWMAN}="snowman"></a></div>' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
463 self.assertEqual(self.soup(markup).div.encode("utf8"), markup.encode("utf8")) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
464 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
465 class TestUnicodeDammit(unittest.TestCase): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
466 """Standalone tests of UnicodeDammit.""" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
467 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
468 def test_unicode_input(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
469 markup = "I'm already Unicode! \N{SNOWMAN}" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
470 dammit = UnicodeDammit(markup) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
471 self.assertEqual(dammit.unicode_markup, markup) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
472 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
473 def test_smart_quotes_to_unicode(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
474 markup = b"<foo>\x91\x92\x93\x94</foo>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
475 dammit = UnicodeDammit(markup) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
476 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
477 dammit.unicode_markup, "<foo>\u2018\u2019\u201c\u201d</foo>") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
478 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
479 def test_smart_quotes_to_xml_entities(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
480 markup = b"<foo>\x91\x92\x93\x94</foo>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
481 dammit = UnicodeDammit(markup, smart_quotes_to="xml") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
482 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
483 dammit.unicode_markup, "<foo>‘’“”</foo>") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
484 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
485 def test_smart_quotes_to_html_entities(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
486 markup = b"<foo>\x91\x92\x93\x94</foo>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
487 dammit = UnicodeDammit(markup, smart_quotes_to="html") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
488 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
489 dammit.unicode_markup, "<foo>‘’“”</foo>") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
490 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
491 def test_smart_quotes_to_ascii(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
492 markup = b"<foo>\x91\x92\x93\x94</foo>" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
493 dammit = UnicodeDammit(markup, smart_quotes_to="ascii") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
494 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
495 dammit.unicode_markup, """<foo>''""</foo>""") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
496 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
497 def test_detect_utf8(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
498 utf8 = b"Sacr\xc3\xa9 bleu! \xe2\x98\x83" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
499 dammit = UnicodeDammit(utf8) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
500 self.assertEqual(dammit.original_encoding.lower(), 'utf-8') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
501 self.assertEqual(dammit.unicode_markup, 'Sacr\xe9 bleu! \N{SNOWMAN}') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
502 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
503 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
504 def test_convert_hebrew(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
505 hebrew = b"\xed\xe5\xec\xf9" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
506 dammit = UnicodeDammit(hebrew, ["iso-8859-8"]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
507 self.assertEqual(dammit.original_encoding.lower(), 'iso-8859-8') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
508 self.assertEqual(dammit.unicode_markup, '\u05dd\u05d5\u05dc\u05e9') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
509 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
510 def test_dont_see_smart_quotes_where_there_are_none(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
511 utf_8 = b"\343\202\261\343\203\274\343\202\277\343\202\244 Watch" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
512 dammit = UnicodeDammit(utf_8) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
513 self.assertEqual(dammit.original_encoding.lower(), 'utf-8') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
514 self.assertEqual(dammit.unicode_markup.encode("utf-8"), utf_8) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
515 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
516 def test_ignore_inappropriate_codecs(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
517 utf8_data = "Räksmörgås".encode("utf-8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
518 dammit = UnicodeDammit(utf8_data, ["iso-8859-8"]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
519 self.assertEqual(dammit.original_encoding.lower(), 'utf-8') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
520 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
521 def test_ignore_invalid_codecs(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
522 utf8_data = "Räksmörgås".encode("utf-8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
523 for bad_encoding in ['.utf8', '...', 'utF---16.!']: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
524 dammit = UnicodeDammit(utf8_data, [bad_encoding]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
525 self.assertEqual(dammit.original_encoding.lower(), 'utf-8') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
526 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
527 def test_exclude_encodings(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
528 # This is UTF-8. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
529 utf8_data = "Räksmörgås".encode("utf-8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
530 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
531 # But if we exclude UTF-8 from consideration, the guess is |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
532 # Windows-1252. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
533 dammit = UnicodeDammit(utf8_data, exclude_encodings=["utf-8"]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
534 self.assertEqual(dammit.original_encoding.lower(), 'windows-1252') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
535 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
536 # And if we exclude that, there is no valid guess at all. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
537 dammit = UnicodeDammit( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
538 utf8_data, exclude_encodings=["utf-8", "windows-1252"]) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
539 self.assertEqual(dammit.original_encoding, None) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
540 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
541 def test_encoding_detector_replaces_junk_in_encoding_name_with_replacement_character(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
542 detected = EncodingDetector( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
543 b'<?xml version="1.0" encoding="UTF-\xdb" ?>') |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
544 encodings = list(detected.encodings) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
545 assert 'utf-\N{REPLACEMENT CHARACTER}' in encodings |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
546 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
547 def test_detect_html5_style_meta_tag(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
548 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
549 for data in ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
550 b'<html><meta charset="euc-jp" /></html>', |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
551 b"<html><meta charset='euc-jp' /></html>", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
552 b"<html><meta charset=euc-jp /></html>", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
553 b"<html><meta charset=euc-jp/></html>"): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
554 dammit = UnicodeDammit(data, is_html=True) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
555 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
556 "euc-jp", dammit.original_encoding) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
557 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
558 def test_last_ditch_entity_replacement(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
559 # This is a UTF-8 document that contains bytestrings |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
560 # completely incompatible with UTF-8 (ie. encoded with some other |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
561 # encoding). |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
562 # |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
563 # Since there is no consistent encoding for the document, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
564 # Unicode, Dammit will eventually encode the document as UTF-8 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
565 # and encode the incompatible characters as REPLACEMENT |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
566 # CHARACTER. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
567 # |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
568 # If chardet is installed, it will detect that the document |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
569 # can be converted into ISO-8859-1 without errors. This happens |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
570 # to be the wrong encoding, but it is a consistent encoding, so the |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
571 # code we're testing here won't run. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
572 # |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
573 # So we temporarily disable chardet if it's present. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
574 doc = b"""\357\273\277<?xml version="1.0" encoding="UTF-8"?> |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
575 <html><b>\330\250\330\252\330\261</b> |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
576 <i>\310\322\321\220\312\321\355\344</i></html>""" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
577 chardet = bs4.dammit.chardet_dammit |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
578 logging.disable(logging.WARNING) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
579 try: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
580 def noop(str): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
581 return None |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
582 bs4.dammit.chardet_dammit = noop |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
583 dammit = UnicodeDammit(doc) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
584 self.assertEqual(True, dammit.contains_replacement_characters) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
585 self.assertTrue("\ufffd" in dammit.unicode_markup) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
586 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
587 soup = BeautifulSoup(doc, "html.parser") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
588 self.assertTrue(soup.contains_replacement_characters) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
589 finally: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
590 logging.disable(logging.NOTSET) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
591 bs4.dammit.chardet_dammit = chardet |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
592 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
593 def test_byte_order_mark_removed(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
594 # A document written in UTF-16LE will have its byte order marker stripped. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
595 data = b'\xff\xfe<\x00a\x00>\x00\xe1\x00\xe9\x00<\x00/\x00a\x00>\x00' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
596 dammit = UnicodeDammit(data) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
597 self.assertEqual("<a>áé</a>", dammit.unicode_markup) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
598 self.assertEqual("utf-16le", dammit.original_encoding) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
599 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
600 def test_detwingle(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
601 # Here's a UTF8 document. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
602 utf8 = ("\N{SNOWMAN}" * 3).encode("utf8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
603 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
604 # Here's a Windows-1252 document. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
605 windows_1252 = ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
606 "\N{LEFT DOUBLE QUOTATION MARK}Hi, I like Windows!" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
607 "\N{RIGHT DOUBLE QUOTATION MARK}").encode("windows_1252") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
608 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
609 # Through some unholy alchemy, they've been stuck together. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
610 doc = utf8 + windows_1252 + utf8 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
611 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
612 # The document can't be turned into UTF-8: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
613 self.assertRaises(UnicodeDecodeError, doc.decode, "utf8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
614 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
615 # Unicode, Dammit thinks the whole document is Windows-1252, |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
616 # and decodes it into "☃☃☃“Hi, I like Windows!”☃☃☃" |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
617 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
618 # But if we run it through fix_embedded_windows_1252, it's fixed: |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
619 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
620 fixed = UnicodeDammit.detwingle(doc) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
621 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
622 "☃☃☃“Hi, I like Windows!”☃☃☃", fixed.decode("utf8")) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
623 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
624 def test_detwingle_ignores_multibyte_characters(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
625 # Each of these characters has a UTF-8 representation ending |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
626 # in \x93. \x93 is a smart quote if interpreted as |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
627 # Windows-1252. But our code knows to skip over multibyte |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
628 # UTF-8 characters, so they'll survive the process unscathed. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
629 for tricky_unicode_char in ( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
630 "\N{LATIN SMALL LIGATURE OE}", # 2-byte char '\xc5\x93' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
631 "\N{LATIN SUBSCRIPT SMALL LETTER X}", # 3-byte char '\xe2\x82\x93' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
632 "\xf0\x90\x90\x93", # This is a CJK character, not sure which one. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
633 ): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
634 input = tricky_unicode_char.encode("utf8") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
635 self.assertTrue(input.endswith(b'\x93')) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
636 output = UnicodeDammit.detwingle(input) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
637 self.assertEqual(output, input) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
638 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
639 def test_find_declared_encoding(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
640 # Test our ability to find a declared encoding inside an |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
641 # XML or HTML document. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
642 # |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
643 # Even if the document comes in as Unicode, it may be |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
644 # interesting to know what encoding was claimed |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
645 # originally. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
646 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
647 html_unicode = '<html><head><meta charset="utf-8"></head></html>' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
648 html_bytes = html_unicode.encode("ascii") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
649 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
650 xml_unicode= '<?xml version="1.0" encoding="ISO-8859-1" ?>' |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
651 xml_bytes = xml_unicode.encode("ascii") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
652 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
653 m = EncodingDetector.find_declared_encoding |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
654 self.assertEqual(None, m(html_unicode, is_html=False)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
655 self.assertEqual("utf-8", m(html_unicode, is_html=True)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
656 self.assertEqual("utf-8", m(html_bytes, is_html=True)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
657 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
658 self.assertEqual("iso-8859-1", m(xml_unicode)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
659 self.assertEqual("iso-8859-1", m(xml_bytes)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
660 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
661 # Normally, only the first few kilobytes of a document are checked for |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
662 # an encoding. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
663 spacer = b' ' * 5000 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
664 self.assertEqual(None, m(spacer + html_bytes)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
665 self.assertEqual(None, m(spacer + xml_bytes)) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
666 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
667 # But you can tell find_declared_encoding to search an entire |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
668 # HTML document. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
669 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
670 "utf-8", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
671 m(spacer + html_bytes, is_html=True, search_entire_document=True) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
672 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
673 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
674 # The XML encoding declaration has to be the very first thing |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
675 # in the document. We'll allow whitespace before the document |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
676 # starts, but nothing else. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
677 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
678 "iso-8859-1", |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
679 m(xml_bytes, search_entire_document=True) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
680 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
681 self.assertEqual( |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
682 None, m(b'a' + xml_bytes, search_entire_document=True) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
683 ) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
684 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
685 class TestNamedspacedAttribute(SoupTest): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
686 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
687 def test_name_may_be_none_or_missing(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
688 a = NamespacedAttribute("xmlns", None) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
689 self.assertEqual(a, "xmlns") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
690 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
691 a = NamespacedAttribute("xmlns") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
692 self.assertEqual(a, "xmlns") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
693 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
694 def test_attribute_is_equivalent_to_colon_separated_string(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
695 a = NamespacedAttribute("a", "b") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
696 self.assertEqual("a:b", a) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
697 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
698 def test_attributes_are_equivalent_if_prefix_and_name_identical(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
699 a = NamespacedAttribute("a", "b", "c") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
700 b = NamespacedAttribute("a", "b", "c") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
701 self.assertEqual(a, b) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
702 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
703 # The actual namespace is not considered. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
704 c = NamespacedAttribute("a", "b", None) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
705 self.assertEqual(a, c) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
706 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
707 # But name and prefix are important. |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
708 d = NamespacedAttribute("a", "z", "c") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
709 self.assertNotEqual(a, d) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
710 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
711 e = NamespacedAttribute("z", "b", "c") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
712 self.assertNotEqual(a, e) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
713 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
714 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
715 class TestAttributeValueWithCharsetSubstitution(unittest.TestCase): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
716 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
717 def test_content_meta_attribute_value(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
718 value = CharsetMetaAttributeValue("euc-jp") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
719 self.assertEqual("euc-jp", value) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
720 self.assertEqual("euc-jp", value.original_value) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
721 self.assertEqual("utf8", value.encode("utf8")) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
722 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
723 |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
724 def test_content_meta_attribute_value(self): |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
725 value = ContentMetaAttributeValue("text/html; charset=euc-jp") |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
726 self.assertEqual("text/html; charset=euc-jp", value) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
727 self.assertEqual("text/html; charset=euc-jp", value.original_value) |
d30785e31577
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
728 self.assertEqual("text/html; charset=utf8", value.encode("utf8")) |