Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / python / cpython3 / e822049efcfb3eafede8035609284656671aece4 / . / Lib / test / test_htmlparser.py
blob: 85e60e66373886190ff94d45bfe78149a9a9451e [file] [log] [blame]
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]1"""Tests for HTMLParser.py."""
2
3import HTMLParser
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]4import pprint
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]5import sys
6import test_support
7import unittest
8
9
10class EventCollector(HTMLParser.HTMLParser):
11
12 def __init__(self):
13 self.events = []
14 self.append = self.events.append
15 HTMLParser.HTMLParser.__init__(self)
16
17 def get_events(self):
18 # Normalize the list of events so that buffer artefacts don't
19 # separate runs of contiguous characters.
20 L = []
21 prevtype = None
22 for event in self.events:
23 type = event[0]
24 if type == prevtype == "data":
25 L[-1] = ("data", L[-1][1] + event[1])
26 else:
27 L.append(event)
28 prevtype = type
29 self.events = L
30 return L
31
32 # structure markup
33
34 def handle_starttag(self, tag, attrs):
35 self.append(("starttag", tag, attrs))
36
37 def handle_startendtag(self, tag, attrs):
38 self.append(("startendtag", tag, attrs))
39
40 def handle_endtag(self, tag):
41 self.append(("endtag", tag))
42
43 # all other markup
44
45 def handle_comment(self, data):
46 self.append(("comment", data))
47
48 def handle_charref(self, data):
49 self.append(("charref", data))
50
51 def handle_data(self, data):
52 self.append(("data", data))
53
54 def handle_decl(self, data):
55 self.append(("decl", data))
56
57 def handle_entityref(self, data):
58 self.append(("entityref", data))
59
60 def handle_pi(self, data):
61 self.append(("pi", data))
62
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]63 def unknown_decl(self, decl):
64 self.append(("unknown decl", decl))
65
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]66
67class EventCollectorExtra(EventCollector):
68
69 def handle_starttag(self, tag, attrs):
70 EventCollector.handle_starttag(self, tag, attrs)
71 self.append(("starttag_text", self.get_starttag_text()))
72
73
74class TestCaseBase(unittest.TestCase):
75
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]76 def _run_check(self, source, expected_events, collector=EventCollector):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]77 parser = collector()
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]78 for s in source:
79 parser.feed(s)
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]80 parser.close()
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]81 events = parser.get_events()
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]82 if events != expected_events:
83 self.fail("received events did not match expected events\n"
84 "Expected:\n" + pprint.pformat(expected_events) +
85 "\nReceived:\n" + pprint.pformat(events))
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]86
87 def _run_check_extra(self, source, events):
88 self._run_check(source, events, EventCollectorExtra)
89
90 def _parse_error(self, source):
91 def parse(source=source):
92 parser = HTMLParser.HTMLParser()
93 parser.feed(source)
94 parser.close()
95 self.assertRaises(HTMLParser.HTMLParseError, parse)
96
97
98class HTMLParserTestCase(TestCaseBase):
99
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]100 def test_processing_instruction_only(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]101 self._run_check("<?processing instruction>", [
102 ("pi", "processing instruction"),
103 ])
104
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]105 def test_simple_html(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]106 self._run_check("""
107<!DOCTYPE html PUBLIC 'foo'>
108<HTML>&entity;&#32;
109<!--comment1a
110-></foo><bar>&lt;<?pi?></foo<bar
111comment1b-->
112<Img sRc='Bar' isMAP>sample
113text
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]114&#x201C;
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]115<!--comment2a-- --comment2b-->
116</Html>
117""", [
118 ("data", "\n"),
119 ("decl", "DOCTYPE html PUBLIC 'foo'"),
120 ("data", "\n"),
121 ("starttag", "html", []),
122 ("entityref", "entity"),
123 ("charref", "32"),
124 ("data", "\n"),
125 ("comment", "comment1a\n-></foo><bar>&lt;<?pi?></foo<bar\ncomment1b"),
126 ("data", "\n"),
127 ("starttag", "img", [("src", "Bar"), ("ismap", None)]),
128 ("data", "sample\ntext\n"),
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]129 ("charref", "x201C"),
130 ("data", "\n"),
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]131 ("comment", "comment2a-- --comment2b"),
132 ("data", "\n"),
133 ("endtag", "html"),
134 ("data", "\n"),
135 ])
136
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]137 def test_doctype_decl(self):
138 inside = """\
139DOCTYPE html [
140 <!ELEMENT html - O EMPTY>
141 <!ATTLIST html
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]142 version CDATA #IMPLIED
143 profile CDATA 'DublinCore'>
144 <!NOTATION datatype SYSTEM 'http://xml.python.org/notations/python-module'>
145 <!ENTITY myEntity 'internal parsed entity'>
146 <!ENTITY anEntity SYSTEM 'http://xml.python.org/entities/something.xml'>
147 <!ENTITY % paramEntity 'name|name|name'>
148 %paramEntity;
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]149 <!-- comment -->
150]"""
151 self._run_check("<!%s>" % inside, [
152 ("decl", inside),
153 ])
154
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]155 def test_bad_nesting(self):
156 # Strangely, this *is* supposed to test that overlapping
157 # elements are allowed. HTMLParser is more geared toward
158 # lexing the input that parsing the structure.
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]159 self._run_check("<a><b></a></b>", [
160 ("starttag", "a", []),
161 ("starttag", "b", []),
162 ("endtag", "a"),
163 ("endtag", "b"),
164 ])
165
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]166 def test_bare_ampersands(self):
167 self._run_check("this text & contains & ampersands &", [
168 ("data", "this text & contains & ampersands &"),
169 ])
170
171 def test_bare_pointy_brackets(self):
172 self._run_check("this < text > contains < bare>pointy< brackets", [
173 ("data", "this < text > contains < bare>pointy< brackets"),
174 ])
175
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]176 def test_attr_syntax(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]177 output = [
178 ("starttag", "a", [("b", "v"), ("c", "v"), ("d", "v"), ("e", None)])
179 ]
180 self._run_check("""<a b='v' c="v" d=v e>""", output)
181 self._run_check("""<a b = 'v' c = "v" d = v e>""", output)
182 self._run_check("""<a\nb\n=\n'v'\nc\n=\n"v"\nd\n=\nv\ne>""", output)
183 self._run_check("""<a\tb\t=\t'v'\tc\t=\t"v"\td\t=\tv\te>""", output)
184
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]185 def test_attr_values(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]186 self._run_check("""<a b='xxx\n\txxx' c="yyy\t\nyyy" d='\txyz\n'>""",
187 [("starttag", "a", [("b", "xxx\n\txxx"),
188 ("c", "yyy\t\nyyy"),
189 ("d", "\txyz\n")])
190 ])
191 self._run_check("""<a b='' c="">""", [
192 ("starttag", "a", [("b", ""), ("c", "")]),
193 ])
194
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]195 def test_attr_entity_replacement(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]196 self._run_check("""<a b='&amp;&gt;&lt;&quot;&apos;'>""", [
197 ("starttag", "a", [("b", "&><\"'")]),
198 ])
199
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]200 def test_attr_funky_names(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]201 self._run_check("""<a a.b='v' c:d=v e-f=v>""", [
202 ("starttag", "a", [("a.b", "v"), ("c:d", "v"), ("e-f", "v")]),
203 ])
204
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]205 def test_illegal_declarations(self):
Fred Drake7cf613d2001-09-04 16:26:03 +0000[diff] [blame]206 self._parse_error('<!spacer type="block" height="25">')
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]207
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]208 def test_starttag_end_boundary(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]209 self._run_check("""<a b='<'>""", [("starttag", "a", [("b", "<")])])
210 self._run_check("""<a b='>'>""", [("starttag", "a", [("b", ">")])])
211
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]212 def test_buffer_artefacts(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]213 output = [("starttag", "a", [("b", "<")])]
214 self._run_check(["<a b='<'>"], output)
215 self._run_check(["<a ", "b='<'>"], output)
216 self._run_check(["<a b", "='<'>"], output)
217 self._run_check(["<a b=", "'<'>"], output)
218 self._run_check(["<a b='<", "'>"], output)
219 self._run_check(["<a b='<'", ">"], output)
220
221 output = [("starttag", "a", [("b", ">")])]
222 self._run_check(["<a b='>'>"], output)
223 self._run_check(["<a ", "b='>'>"], output)
224 self._run_check(["<a b", "='>'>"], output)
225 self._run_check(["<a b=", "'>'>"], output)
226 self._run_check(["<a b='>", "'>"], output)
227 self._run_check(["<a b='>'", ">"], output)
228
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]229 def test_starttag_junk_chars(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]230 self._parse_error("</>")
231 self._parse_error("</$>")
232 self._parse_error("</")
233 self._parse_error("</a")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]234 self._parse_error("<a<a>")
235 self._parse_error("</a<a>")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]236 self._parse_error("<!")
237 self._parse_error("<a $>")
238 self._parse_error("<a")
239 self._parse_error("<a foo='bar'")
240 self._parse_error("<a foo='bar")
241 self._parse_error("<a foo='>'")
242 self._parse_error("<a foo='>")
243 self._parse_error("<a foo=>")
244
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]245 def test_declaration_junk_chars(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]246 self._parse_error("<!DOCTYPE foo $ >")
247
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]248 def test_startendtag(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]249 self._run_check("<p/>", [
250 ("startendtag", "p", []),
251 ])
252 self._run_check("<p></p>", [
253 ("starttag", "p", []),
254 ("endtag", "p"),
255 ])
256 self._run_check("<p><img src='foo' /></p>", [
257 ("starttag", "p", []),
258 ("startendtag", "img", [("src", "foo")]),
259 ("endtag", "p"),
260 ])
261
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]262 def test_get_starttag_text(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]263 s = """<foo:bar \n one="1"\ttwo=2 >"""
264 self._run_check_extra(s, [
265 ("starttag", "foo:bar", [("one", "1"), ("two", "2")]),
266 ("starttag_text", s)])
267
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]268 def test_cdata_content(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]269 s = """<script> <!-- not a comment --> &not-an-entity-ref; </script>"""
270 self._run_check(s, [
271 ("starttag", "script", []),
272 ("data", " <!-- not a comment --> &not-an-entity-ref; "),
273 ("endtag", "script"),
274 ])
275 s = """<script> <not a='start tag'> </script>"""
276 self._run_check(s, [
277 ("starttag", "script", []),
278 ("data", " <not a='start tag'> "),
279 ("endtag", "script"),
280 ])
281
282
Fred Drakee8220492001-09-24 20:19:08 +0000[diff] [blame^]283def test_main():
284 test_support.run_unittest(HTMLParserTestCase)
285
286
287if __name__ == "__main__":
288 test_main()