Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / python / cpython2 / eb7113888ef86e917b10ab608791416163ccc544 / . / Lib / test / test_htmlparser.py
blob: f557da1021fd819a387e50af30f843bb9995640e [file] [log] [blame]
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]1"""Tests for HTMLParser.py."""
2
Georg Brandlbcdafa42008-05-20 07:58:42 +0000[diff] [blame]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 unittest
Barry Warsaw04f357c2002-07-23 19:04:11 +0000[diff] [blame]6from test import test_support
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]7
8
Georg Brandlbcdafa42008-05-20 07:58:42 +0000[diff] [blame]9class EventCollector(HTMLParser.HTMLParser):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]10
11 def __init__(self):
12 self.events = []
13 self.append = self.events.append
Georg Brandlbcdafa42008-05-20 07:58:42 +0000[diff] [blame]14 HTMLParser.HTMLParser.__init__(self)
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]15
16 def get_events(self):
17 # Normalize the list of events so that buffer artefacts don't
18 # separate runs of contiguous characters.
19 L = []
20 prevtype = None
21 for event in self.events:
22 type = event[0]
23 if type == prevtype == "data":
24 L[-1] = ("data", L[-1][1] + event[1])
25 else:
26 L.append(event)
27 prevtype = type
28 self.events = L
29 return L
30
31 # structure markup
32
33 def handle_starttag(self, tag, attrs):
34 self.append(("starttag", tag, attrs))
35
36 def handle_startendtag(self, tag, attrs):
37 self.append(("startendtag", tag, attrs))
38
39 def handle_endtag(self, tag):
40 self.append(("endtag", tag))
41
42 # all other markup
43
44 def handle_comment(self, data):
45 self.append(("comment", data))
46
47 def handle_charref(self, data):
48 self.append(("charref", data))
49
50 def handle_data(self, data):
51 self.append(("data", data))
52
53 def handle_decl(self, data):
54 self.append(("decl", data))
55
56 def handle_entityref(self, data):
57 self.append(("entityref", data))
58
59 def handle_pi(self, data):
60 self.append(("pi", data))
61
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]62 def unknown_decl(self, decl):
63 self.append(("unknown decl", decl))
64
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]65
66class EventCollectorExtra(EventCollector):
67
68 def handle_starttag(self, tag, attrs):
69 EventCollector.handle_starttag(self, tag, attrs)
70 self.append(("starttag_text", self.get_starttag_text()))
71
72
73class TestCaseBase(unittest.TestCase):
74
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]75 def _run_check(self, source, expected_events, collector=EventCollector):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]76 parser = collector()
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]77 for s in source:
78 parser.feed(s)
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]79 parser.close()
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]80 events = parser.get_events()
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]81 if events != expected_events:
82 self.fail("received events did not match expected events\n"
83 "Expected:\n" + pprint.pformat(expected_events) +
84 "\nReceived:\n" + pprint.pformat(events))
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]85
86 def _run_check_extra(self, source, events):
87 self._run_check(source, events, EventCollectorExtra)
88
89 def _parse_error(self, source):
90 def parse(source=source):
Georg Brandlbcdafa42008-05-20 07:58:42 +0000[diff] [blame]91 parser = HTMLParser.HTMLParser()
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]92 parser.feed(source)
93 parser.close()
Georg Brandlbcdafa42008-05-20 07:58:42 +0000[diff] [blame]94 self.assertRaises(HTMLParser.HTMLParseError, parse)
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]95
96
97class HTMLParserTestCase(TestCaseBase):
98
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]99 def test_processing_instruction_only(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]100 self._run_check("<?processing instruction>", [
101 ("pi", "processing instruction"),
102 ])
Fred Drakefafd56f2003-04-17 22:19:26 +0000[diff] [blame]103 self._run_check("<?processing instruction ?>", [
104 ("pi", "processing instruction ?"),
105 ])
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]106
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]107 def test_simple_html(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]108 self._run_check("""
109<!DOCTYPE html PUBLIC 'foo'>
110<HTML>&entity;&#32;
111<!--comment1a
112-></foo><bar>&lt;<?pi?></foo<bar
113comment1b-->
114<Img sRc='Bar' isMAP>sample
115text
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]116&#x201C;
Georg Brandld09def32006-03-09 13:27:14 +0000[diff] [blame]117<!--comment2a-- --comment2b--><!>
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]118</Html>
119""", [
120 ("data", "\n"),
121 ("decl", "DOCTYPE html PUBLIC 'foo'"),
122 ("data", "\n"),
123 ("starttag", "html", []),
124 ("entityref", "entity"),
125 ("charref", "32"),
126 ("data", "\n"),
127 ("comment", "comment1a\n-></foo><bar>&lt;<?pi?></foo<bar\ncomment1b"),
128 ("data", "\n"),
129 ("starttag", "img", [("src", "Bar"), ("ismap", None)]),
130 ("data", "sample\ntext\n"),
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]131 ("charref", "x201C"),
132 ("data", "\n"),
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]133 ("comment", "comment2a-- --comment2b"),
134 ("data", "\n"),
135 ("endtag", "html"),
136 ("data", "\n"),
137 ])
138
Fred Drake073148c2001-12-03 16:44:09 +0000[diff] [blame]139 def test_unclosed_entityref(self):
140 self._run_check("&entityref foo", [
141 ("entityref", "entityref"),
142 ("data", " foo"),
143 ])
144
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]145 def test_doctype_decl(self):
146 inside = """\
147DOCTYPE html [
148 <!ELEMENT html - O EMPTY>
149 <!ATTLIST html
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]150 version CDATA #IMPLIED
151 profile CDATA 'DublinCore'>
152 <!NOTATION datatype SYSTEM 'http://xml.python.org/notations/python-module'>
153 <!ENTITY myEntity 'internal parsed entity'>
154 <!ENTITY anEntity SYSTEM 'http://xml.python.org/entities/something.xml'>
155 <!ENTITY % paramEntity 'name|name|name'>
156 %paramEntity;
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]157 <!-- comment -->
158]"""
159 self._run_check("<!%s>" % inside, [
160 ("decl", inside),
161 ])
162
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]163 def test_bad_nesting(self):
164 # Strangely, this *is* supposed to test that overlapping
165 # elements are allowed. HTMLParser is more geared toward
166 # lexing the input that parsing the structure.
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]167 self._run_check("<a><b></a></b>", [
168 ("starttag", "a", []),
169 ("starttag", "b", []),
170 ("endtag", "a"),
171 ("endtag", "b"),
172 ])
173
Fred Drake029acfb2001-08-20 21:24:19 +0000[diff] [blame]174 def test_bare_ampersands(self):
175 self._run_check("this text & contains & ampersands &", [
176 ("data", "this text & contains & ampersands &"),
177 ])
178
179 def test_bare_pointy_brackets(self):
180 self._run_check("this < text > contains < bare>pointy< brackets", [
181 ("data", "this < text > contains < bare>pointy< brackets"),
182 ])
183
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]184 def test_attr_syntax(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]185 output = [
186 ("starttag", "a", [("b", "v"), ("c", "v"), ("d", "v"), ("e", None)])
187 ]
188 self._run_check("""<a b='v' c="v" d=v e>""", output)
189 self._run_check("""<a b = 'v' c = "v" d = v e>""", output)
190 self._run_check("""<a\nb\n=\n'v'\nc\n=\n"v"\nd\n=\nv\ne>""", output)
191 self._run_check("""<a\tb\t=\t'v'\tc\t=\t"v"\td\t=\tv\te>""", output)
192
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]193 def test_attr_values(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]194 self._run_check("""<a b='xxx\n\txxx' c="yyy\t\nyyy" d='\txyz\n'>""",
195 [("starttag", "a", [("b", "xxx\n\txxx"),
196 ("c", "yyy\t\nyyy"),
197 ("d", "\txyz\n")])
198 ])
199 self._run_check("""<a b='' c="">""", [
200 ("starttag", "a", [("b", ""), ("c", "")]),
201 ])
Fred Drake0834d772003-03-14 16:21:57 +0000[diff] [blame]202 # Regression test for SF patch #669683.
203 self._run_check("<e a=rgb(1,2,3)>", [
204 ("starttag", "e", [("a", "rgb(1,2,3)")]),
205 ])
Tim Peters27f88362004-07-08 04:22:35 +0000[diff] [blame]206 # Regression test for SF bug #921657.
Andrew M. Kuchlingb7d8ce02004-06-05 15:31:45 +0000[diff] [blame]207 self._run_check("<a href=mailto:xyz@example.com>", [
208 ("starttag", "a", [("href", "mailto:xyz@example.com")]),
209 ])
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]210
Ezio Melotti9f1ffb22011-04-05 20:40:52 +0300[diff] [blame]211 def test_attr_nonascii(self):
212 # see issue 7311
213 self._run_check(u"<img src=/foo/bar.png alt=\u4e2d\u6587>", [
214 ("starttag", "img", [("src", "/foo/bar.png"),
215 ("alt", u"\u4e2d\u6587")]),
216 ])
217 self._run_check(u"<a title='\u30c6\u30b9\u30c8' "
218 u"href='\u30c6\u30b9\u30c8.html'>", [
219 ("starttag", "a", [("title", u"\u30c6\u30b9\u30c8"),
220 ("href", u"\u30c6\u30b9\u30c8.html")]),
221 ])
222 self._run_check(u'<a title="\u30c6\u30b9\u30c8" '
223 u'href="\u30c6\u30b9\u30c8.html">', [
224 ("starttag", "a", [("title", u"\u30c6\u30b9\u30c8"),
225 ("href", u"\u30c6\u30b9\u30c8.html")]),
226 ])
227
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]228 def test_attr_entity_replacement(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]229 self._run_check("""<a b='&amp;&gt;&lt;&quot;&apos;'>""", [
230 ("starttag", "a", [("b", "&><\"'")]),
231 ])
232
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]233 def test_attr_funky_names(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]234 self._run_check("""<a a.b='v' c:d=v e-f=v>""", [
235 ("starttag", "a", [("a.b", "v"), ("c:d", "v"), ("e-f", "v")]),
236 ])
237
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]238 def test_illegal_declarations(self):
Fred Drake7cf613d2001-09-04 16:26:03 +0000[diff] [blame]239 self._parse_error('<!spacer type="block" height="25">')
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]240
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]241 def test_starttag_end_boundary(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]242 self._run_check("""<a b='<'>""", [("starttag", "a", [("b", "<")])])
243 self._run_check("""<a b='>'>""", [("starttag", "a", [("b", ">")])])
244
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]245 def test_buffer_artefacts(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]246 output = [("starttag", "a", [("b", "<")])]
247 self._run_check(["<a b='<'>"], output)
248 self._run_check(["<a ", "b='<'>"], output)
249 self._run_check(["<a b", "='<'>"], output)
250 self._run_check(["<a b=", "'<'>"], output)
251 self._run_check(["<a b='<", "'>"], output)
252 self._run_check(["<a b='<'", ">"], output)
253
254 output = [("starttag", "a", [("b", ">")])]
255 self._run_check(["<a b='>'>"], output)
256 self._run_check(["<a ", "b='>'>"], output)
257 self._run_check(["<a b", "='>'>"], output)
258 self._run_check(["<a b=", "'>'>"], output)
259 self._run_check(["<a b='>", "'>"], output)
260 self._run_check(["<a b='>'", ">"], output)
261
Fred Drake75d9a622004-09-08 22:57:01 +0000[diff] [blame]262 output = [("comment", "abc")]
263 self._run_check(["", "<!--abc-->"], output)
264 self._run_check(["<", "!--abc-->"], output)
265 self._run_check(["<!", "--abc-->"], output)
266 self._run_check(["<!-", "-abc-->"], output)
267 self._run_check(["<!--", "abc-->"], output)
268 self._run_check(["<!--a", "bc-->"], output)
269 self._run_check(["<!--ab", "c-->"], output)
270 self._run_check(["<!--abc", "-->"], output)
271 self._run_check(["<!--abc-", "->"], output)
272 self._run_check(["<!--abc--", ">"], output)
273 self._run_check(["<!--abc-->", ""], output)
274
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]275 def test_starttag_junk_chars(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]276 self._parse_error("</>")
277 self._parse_error("</$>")
278 self._parse_error("</")
279 self._parse_error("</a")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]280 self._parse_error("<a<a>")
281 self._parse_error("</a<a>")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]282 self._parse_error("<!")
283 self._parse_error("<a $>")
284 self._parse_error("<a")
285 self._parse_error("<a foo='bar'")
286 self._parse_error("<a foo='bar")
287 self._parse_error("<a foo='>'")
288 self._parse_error("<a foo='>")
289 self._parse_error("<a foo=>")
290
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]291 def test_declaration_junk_chars(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]292 self._parse_error("<!DOCTYPE foo $ >")
293
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]294 def test_startendtag(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]295 self._run_check("<p/>", [
296 ("startendtag", "p", []),
297 ])
298 self._run_check("<p></p>", [
299 ("starttag", "p", []),
300 ("endtag", "p"),
301 ])
302 self._run_check("<p><img src='foo' /></p>", [
303 ("starttag", "p", []),
304 ("startendtag", "img", [("src", "foo")]),
305 ("endtag", "p"),
306 ])
307
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]308 def test_get_starttag_text(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]309 s = """<foo:bar \n one="1"\ttwo=2 >"""
310 self._run_check_extra(s, [
311 ("starttag", "foo:bar", [("one", "1"), ("two", "2")]),
312 ("starttag_text", s)])
313
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]314 def test_cdata_content(self):
Ezio Melotti7e82b272011-11-01 14:09:56 +0200[diff] [blame]315 contents = [
316 '<!-- not a comment --> &not-an-entity-ref;',
317 "<not a='start tag'>",
318 '<a href="" /> <p> <span></span>',
319 'foo = "</scr" + "ipt>";',
320 'foo = "</SCRIPT" + ">";',
321 'foo = <\n/script> ',
322 '<!-- document.write("</scr" + "ipt>"); -->',
323 ('\n//<![CDATA[\n'
324 'document.write(\'<s\'+\'cript type="text/javascript" '
325 'src="http://www.example.org/r=\'+new '
326 'Date().getTime()+\'"><\\/s\'+\'cript>\');\n//]]>'),
327 '\n<!-- //\nvar foo = 3.14;\n// -->\n',
328 'foo = "</sty" + "le>";',
329 u'<!-- \u2603 -->',
330 # these two should be invalid according to the HTML 5 spec,
331 # section 8.1.2.2
332 #'foo = </\nscript>',
333 #'foo = </ script>',
334 ]
335 elements = ['script', 'style', 'SCRIPT', 'STYLE', 'Script', 'Style']
336 for content in contents:
337 for element in elements:
338 element_lower = element.lower()
339 s = u'<{element}>{content}</{element}>'.format(element=element,
340 content=content)
341 self._run_check(s, [("starttag", element_lower, []),
342 ("data", content),
343 ("endtag", element_lower)])
344
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]345
Martin v. Löwisab8a6bb2007-03-06 14:43:00 +0000[diff] [blame]346 def test_entityrefs_in_attributes(self):
347 self._run_check("<html foo='&euro;&amp;&#97;&#x61;&unsupported;'>", [
348 ("starttag", "html", [("foo", u"\u20AC&aa&unsupported;")])
349 ])
350
Victor Stinner554a3b82010-05-24 21:33:24 +0000[diff] [blame]351 def test_malformatted_charref(self):
352 self._run_check("<p>&#bad;</p>", [
353 ("starttag", "p", []),
354 ("data", "&#bad;"),
355 ("endtag", "p"),
356 ])
357
Senthil Kumaran3f60f092010-12-28 16:05:07 +0000[diff] [blame]358 def test_unescape_function(self):
359 parser = HTMLParser.HTMLParser()
360 self.assertEqual(parser.unescape('&#bad;'),'&#bad;')
361 self.assertEqual(parser.unescape('&#0038;'),'&')
362
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]363
Fred Drakee8220492001-09-24 20:19:08 +0000[diff] [blame]364def test_main():
365 test_support.run_unittest(HTMLParserTestCase)
366
367
368if __name__ == "__main__":
369 test_main()