src/lxml/html/tests/test_clean.py

   1 import unittest
   2 from lxml.tests.common_imports import make_doctest
   3
   4 import lxml.html
   5 from lxml.html.clean import Cleaner, clean_html
   6
   7
   8 class CleanerTest(unittest.TestCase):
   9     def test_allow_tags(self):
  10         html = """
  11             <html>
  12             <head>
  13             </head>
  14             <body>
  15             <p>some text</p>
  16             <table>
  17             <tr>
  18             <td>hello</td><td>world</td>
  19             </tr>
  20             <tr>
  21             <td>hello</td><td>world</td>
  22             </tr>
  23             </table>
  24             <img>
  25             </body>
  26             </html>
  27             """
  28
  29         html_root = lxml.html.document_fromstring(html)
  30         cleaner = Cleaner(
  31             remove_unknown_tags = False,
  32             allow_tags = ['table', 'tr', 'td'])
  33         result = cleaner.clean_html(html_root)
  34
  35         self.assertEqual(12-5+1, len(list(result.iter())))
  36
  37     def test_safe_attrs_included(self):
  38         html = """<p><span style="color: #00ffff;">Cyan</span></p>"""
  39
  40         safe_attrs=set(lxml.html.defs.safe_attrs)
  41         safe_attrs.add('style')
  42
  43         cleaner = Cleaner(
  44             safe_attrs_only=True,
  45             safe_attrs=safe_attrs)
  46         result = cleaner.clean_html(html)
  47
  48         self.assertEqual(html, result)
  49
  50     def test_safe_attrs_excluded(self):
  51         html = """<p><span style="color: #00ffff;">Cyan</span></p>"""
  52         expected = """<p><span>Cyan</span></p>"""
  53
  54         safe_attrs=set()
  55
  56         cleaner = Cleaner(
  57             safe_attrs_only=True,
  58             safe_attrs=safe_attrs)
  59         result = cleaner.clean_html(html)
  60
  61         self.assertEqual(expected, result)
  62
  63     def test_clean_invalid_root_tag(self):
  64         # only testing that cleaning with invalid root tags works at all
  65         s = lxml.html.fromstring('parent <invalid tag>child</another>')
  66         self.assertEqual('parent child', clean_html(s).text_content())
  67
  68         s = lxml.html.fromstring('<invalid tag>child</another>')
  69         self.assertEqual('child', clean_html(s).text_content())
  70
  71
  72 def test_suite():
  73     suite = unittest.TestSuite()
  74     suite.addTests([make_doctest('test_clean.txt')])
  75     suite.addTests([make_doctest('test_clean_embed.txt')])
  76     suite.addTests(unittest.makeSuite(CleanerTest))
  77     return suite