Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / python / cpython2 / 2d679a42bf229692e4400e1efc26eb13d2375c02 / . / Lib / test / test_htmlparser.py
blob: 5dfe466225ebab6ce026f828f96663337678b802 [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 Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]184 def test_illegal_declarations(self):
Fred Drake7cf613d2001-09-04 16:26:03 +0000[diff] [blame]185 self._parse_error('<!spacer type="block" height="25">')
Fred Drakec20a6982001-09-04 15:13:04 +0000[diff] [blame]186
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]187 def test_starttag_end_boundary(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]188 self._run_check("""<a b='<'>""", [("starttag", "a", [("b", "<")])])
189 self._run_check("""<a b='>'>""", [("starttag", "a", [("b", ">")])])
190
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]191 def test_buffer_artefacts(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]192 output = [("starttag", "a", [("b", "<")])]
193 self._run_check(["<a b='<'>"], output)
194 self._run_check(["<a ", "b='<'>"], output)
195 self._run_check(["<a b", "='<'>"], output)
196 self._run_check(["<a b=", "'<'>"], output)
197 self._run_check(["<a b='<", "'>"], output)
198 self._run_check(["<a b='<'", ">"], output)
199
200 output = [("starttag", "a", [("b", ">")])]
201 self._run_check(["<a b='>'>"], output)
202 self._run_check(["<a ", "b='>'>"], output)
203 self._run_check(["<a b", "='>'>"], output)
204 self._run_check(["<a b=", "'>'>"], output)
205 self._run_check(["<a b='>", "'>"], output)
206 self._run_check(["<a b='>'", ">"], output)
207
Fred Drake75d9a622004-09-08 22:57:01 +0000[diff] [blame]208 output = [("comment", "abc")]
209 self._run_check(["", "<!--abc-->"], output)
210 self._run_check(["<", "!--abc-->"], output)
211 self._run_check(["<!", "--abc-->"], output)
212 self._run_check(["<!-", "-abc-->"], output)
213 self._run_check(["<!--", "abc-->"], output)
214 self._run_check(["<!--a", "bc-->"], output)
215 self._run_check(["<!--ab", "c-->"], output)
216 self._run_check(["<!--abc", "-->"], output)
217 self._run_check(["<!--abc-", "->"], output)
218 self._run_check(["<!--abc--", ">"], output)
219 self._run_check(["<!--abc-->", ""], output)
220
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]221 def test_starttag_junk_chars(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]222 self._parse_error("</>")
223 self._parse_error("</$>")
224 self._parse_error("</")
225 self._parse_error("</a")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]226 self._parse_error("<a<a>")
227 self._parse_error("</a<a>")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]228 self._parse_error("<!")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]229 self._parse_error("<a")
230 self._parse_error("<a foo='bar'")
231 self._parse_error("<a foo='bar")
232 self._parse_error("<a foo='>'")
233 self._parse_error("<a foo='>")
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]234
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]235 def test_declaration_junk_chars(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]236 self._parse_error("<!DOCTYPE foo $ >")
237
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]238 def test_startendtag(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]239 self._run_check("<p/>", [
240 ("startendtag", "p", []),
241 ])
242 self._run_check("<p></p>", [
243 ("starttag", "p", []),
244 ("endtag", "p"),
245 ])
246 self._run_check("<p><img src='foo' /></p>", [
247 ("starttag", "p", []),
248 ("startendtag", "img", [("src", "foo")]),
249 ("endtag", "p"),
250 ])
251
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]252 def test_get_starttag_text(self):
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]253 s = """<foo:bar \n one="1"\ttwo=2 >"""
254 self._run_check_extra(s, [
255 ("starttag", "foo:bar", [("one", "1"), ("two", "2")]),
256 ("starttag_text", s)])
257
Fred Drake84bb9d82001-08-03 19:53:01 +0000[diff] [blame]258 def test_cdata_content(self):
Ezio Melotti7e82b272011-11-01 14:09:56 +0200[diff] [blame]259 contents = [
260 '<!-- not a comment --> &not-an-entity-ref;',
261 "<not a='start tag'>",
262 '<a href="" /> <p> <span></span>',
263 'foo = "</scr" + "ipt>";',
264 'foo = "</SCRIPT" + ">";',
265 'foo = <\n/script> ',
266 '<!-- document.write("</scr" + "ipt>"); -->',
267 ('\n//<![CDATA[\n'
268 'document.write(\'<s\'+\'cript type="text/javascript" '
269 'src="http://www.example.org/r=\'+new '
270 'Date().getTime()+\'"><\\/s\'+\'cript>\');\n//]]>'),
271 '\n<!-- //\nvar foo = 3.14;\n// -->\n',
272 'foo = "</sty" + "le>";',
273 u'<!-- \u2603 -->',
274 # these two should be invalid according to the HTML 5 spec,
275 # section 8.1.2.2
276 #'foo = </\nscript>',
277 #'foo = </ script>',
278 ]
279 elements = ['script', 'style', 'SCRIPT', 'STYLE', 'Script', 'Style']
280 for content in contents:
281 for element in elements:
282 element_lower = element.lower()
283 s = u'<{element}>{content}</{element}>'.format(element=element,
284 content=content)
285 self._run_check(s, [("starttag", element_lower, []),
286 ("data", content),
287 ("endtag", element_lower)])
288
Ezio Melotti00dc60b2011-11-18 18:00:40 +0200[diff] [blame]289 def test_cdata_with_closing_tags(self):
290 # see issue #13358
291 # make sure that HTMLParser calls handle_data only once for each CDATA.
292 # The normal event collector normalizes the events in get_events,
293 # so we override it to return the original list of events.
294 class Collector(EventCollector):
295 def get_events(self):
296 return self.events
297
298 content = """<!-- not a comment --> &not-an-entity-ref;
299 <a href="" /> </p><p> &amp; <span></span></style>
300 '</script' + '>' </html> </head> </scripter>!"""
301 for element in [' script', 'script ', ' script ',
302 '\nscript', 'script\n', '\nscript\n']:
303 s = u'<script>{content}</{element}>'.format(element=element,
304 content=content)
305 self._run_check(s, [("starttag", "script", []),
306 ("data", content),
307 ("endtag", "script")],
308 collector=Collector)
309
Victor Stinner554a3b82010-05-24 21:33:24 +0000[diff] [blame]310 def test_malformatted_charref(self):
311 self._run_check("<p>&#bad;</p>", [
312 ("starttag", "p", []),
313 ("data", "&#bad;"),
314 ("endtag", "p"),
315 ])
316
Senthil Kumaran3f60f092010-12-28 16:05:07 +0000[diff] [blame]317 def test_unescape_function(self):
318 parser = HTMLParser.HTMLParser()
319 self.assertEqual(parser.unescape('&#bad;'),'&#bad;')
320 self.assertEqual(parser.unescape('&#0038;'),'&')
321
Fred Drakebd3090d2001-05-18 15:32:59 +0000[diff] [blame]322
Ezio Melotti74592912011-11-08 02:07:18 +0200[diff] [blame]323
324class AttributesTestCase(TestCaseBase):
325
326 def test_attr_syntax(self):
327 output = [
328 ("starttag", "a", [("b", "v"), ("c", "v"), ("d", "v"), ("e", None)])
329 ]
330 self._run_check("""<a b='v' c="v" d=v e>""", output)
331 self._run_check("""<a b = 'v' c = "v" d = v e>""", output)
332 self._run_check("""<a\nb\n=\n'v'\nc\n=\n"v"\nd\n=\nv\ne>""", output)
333 self._run_check("""<a\tb\t=\t'v'\tc\t=\t"v"\td\t=\tv\te>""", output)
334
335 def test_attr_values(self):
336 self._run_check("""<a b='xxx\n\txxx' c="yyy\t\nyyy" d='\txyz\n'>""",
337 [("starttag", "a", [("b", "xxx\n\txxx"),
338 ("c", "yyy\t\nyyy"),
339 ("d", "\txyz\n")])])
340 self._run_check("""<a b='' c="">""",
341 [("starttag", "a", [("b", ""), ("c", "")])])
342 # Regression test for SF patch #669683.
343 self._run_check("<e a=rgb(1,2,3)>",
344 [("starttag", "e", [("a", "rgb(1,2,3)")])])
345 # Regression test for SF bug #921657.
346 self._run_check(
347 "<a href=mailto:xyz@example.com>",
348 [("starttag", "a", [("href", "mailto:xyz@example.com")])])
349
350 def test_attr_nonascii(self):
351 # see issue 7311
352 self._run_check(
353 u"<img src=/foo/bar.png alt=\u4e2d\u6587>",
354 [("starttag", "img", [("src", "/foo/bar.png"),
355 ("alt", u"\u4e2d\u6587")])])
356 self._run_check(
357 u"<a title='\u30c6\u30b9\u30c8' href='\u30c6\u30b9\u30c8.html'>",
358 [("starttag", "a", [("title", u"\u30c6\u30b9\u30c8"),
359 ("href", u"\u30c6\u30b9\u30c8.html")])])
360 self._run_check(
361 u'<a title="\u30c6\u30b9\u30c8" href="\u30c6\u30b9\u30c8.html">',
362 [("starttag", "a", [("title", u"\u30c6\u30b9\u30c8"),
363 ("href", u"\u30c6\u30b9\u30c8.html")])])
364
365 def test_attr_entity_replacement(self):
366 self._run_check(
367 "<a b='&amp;&gt;&lt;&quot;&apos;'>",
368 [("starttag", "a", [("b", "&><\"'")])])
369
370 def test_attr_funky_names(self):
371 self._run_check(
372 "<a a.b='v' c:d=v e-f=v>",
373 [("starttag", "a", [("a.b", "v"), ("c:d", "v"), ("e-f", "v")])])
Ezio Melotti0f1571c2011-11-14 18:04:05 +0200[diff] [blame]374 self._run_check(
375 "<a $><b $=%><c \=/>",
376 [("starttag", "a", [("$", None)]),
377 ("starttag", "b", [("$", "%")]),
378 ("starttag", "c", [("\\", "/")])])
Ezio Melotti74592912011-11-08 02:07:18 +0200[diff] [blame]379
380 def test_entityrefs_in_attributes(self):
381 self._run_check(
382 "<html foo='&euro;&amp;&#97;&#x61;&unsupported;'>",
383 [("starttag", "html", [("foo", u"\u20AC&aa&unsupported;")])])
384
Ezio Melotti0f1571c2011-11-14 18:04:05 +0200[diff] [blame]385 def test_entities_in_attribute_value(self):
386 # see #1200313
387 for entity in ['&', '&amp;', '&#38;', '&#x26;']:
388 self._run_check('<a href="%s">' % entity,
389 [("starttag", "a", [("href", "&")])])
390 self._run_check("<a href='%s'>" % entity,
391 [("starttag", "a", [("href", "&")])])
392 self._run_check("<a href=%s>" % entity,
393 [("starttag", "a", [("href", "&")])])
394
395 def test_malformed_attributes(self):
396 # see #13357
397 html = (
398 "<a href=test'style='color:red;bad1'>test - bad1</a>"
399 "<a href=test'+style='color:red;ba2'>test - bad2</a>"
400 "<a href=test'&nbsp;style='color:red;bad3'>test - bad3</a>"
401 "<a href = test'&nbsp;style='color:red;bad4' >test - bad4</a>"
402 )
403 expected = [
404 ('starttag', 'a', [('href', "test'style='color:red;bad1'")]),
405 ('data', 'test - bad1'), ('endtag', 'a'),
406 ('starttag', 'a', [('href', "test'+style='color:red;ba2'")]),
407 ('data', 'test - bad2'), ('endtag', 'a'),
408 ('starttag', 'a', [('href', u"test'\xa0style='color:red;bad3'")]),
409 ('data', 'test - bad3'), ('endtag', 'a'),
410 ('starttag', 'a', [('href', u"test'\xa0style='color:red;bad4'")]),
411 ('data', 'test - bad4'), ('endtag', 'a')
412 ]
413 self._run_check(html, expected)
414
415 def test_malformed_adjacent_attributes(self):
416 # see #12629
417 self._run_check('<x><y z=""o"" /></x>',
418 [('starttag', 'x', []),
419 ('startendtag', 'y', [('z', ''), ('o""', None)]),
420 ('endtag', 'x')])
421 self._run_check('<x><y z="""" /></x>',
422 [('starttag', 'x', []),
423 ('startendtag', 'y', [('z', ''), ('""', None)]),
424 ('endtag', 'x')])
425
426 # see #755670 for the following 3 tests
427 def test_adjacent_attributes(self):
428 self._run_check('<a width="100%"cellspacing=0>',
429 [("starttag", "a",
430 [("width", "100%"), ("cellspacing","0")])])
431
432 self._run_check('<a id="foo"class="bar">',
433 [("starttag", "a",
434 [("id", "foo"), ("class","bar")])])
435
436 def test_missing_attribute_value(self):
437 self._run_check('<a v=>',
438 [("starttag", "a", [("v", "")])])
439
440 def test_javascript_attribute_value(self):
441 self._run_check("<a href=javascript:popup('/popup/help.html')>",
442 [("starttag", "a",
443 [("href", "javascript:popup('/popup/help.html')")])])
444
445 def test_end_tag_in_attribute_value(self):
446 # see #1745761
447 self._run_check("<a href='http://www.example.org/\">;'>spam</a>",
448 [("starttag", "a",
449 [("href", "http://www.example.org/\">;")]),
450 ("data", "spam"), ("endtag", "a")])
451
452
Fred Drakee8220492001-09-24 20:19:08 +0000[diff] [blame]453def test_main():
Ezio Melotti74592912011-11-08 02:07:18 +0200[diff] [blame]454 test_support.run_unittest(HTMLParserTestCase, AttributesTestCase)
Fred Drakee8220492001-09-24 20:19:08 +0000[diff] [blame]455
456
457if __name__ == "__main__":
458 test_main()