def test_HtmlTool_clean_html():
"""Test HTML cleanup algorithm."""
from clair.textprocessing import HtmlTool
html_nice = HtmlTool.clean_html("""<head></head>
<body>Foo</body>""")
print html_nice
html_nice = HtmlTool.clean_html(html_nasty)
# print html_nasty
print html_nice
pytest.xfail("LXML does strange things.") #IGNORE:E1101
assert html_nice.find("Blitzgerät") != -1
def test_HtmlTool_to_nice_text():
"""Test HTML cleanup algorithm."""
from clair.textprocessing import HtmlTool
text_nice = HtmlTool.to_nice_text(html_nasty)
print html_nasty
print text_nice
assert text_nice.find(u"Blitzgerät") != -1
assert text_nice.find(u"Rückseite") != -1
assert len(text_nice.split("\n")) == 8
def test_HtmlTool_remove_html():
"""Test the HTML to pure text conversion."""
from clair.textprocessing import HtmlTool
text = HtmlTool.remove_html(
"This is <b>bold</b> text. <p>Paragraph.</p> 3 > 2")
print text
assert text == "This is bold text. Paragraph. 3 > 2"
text = HtmlTool.remove_html(
"""Integrated style sheet.
<style type="text/css"> p {color:blue;} </style>
Text after style sheet.""")
print text
assert text == "Integrated style sheet. Text after style sheet."
assert HtmlTool.remove_html(None) == ""
assert HtmlTool.remove_html(nan) == ""
text = HtmlTool.remove_html(" 1 < 2 > 0.5 ")
print text
assert text == " 1 < 2 > 0.5 "
