"""
HTML Parsers wrapper
- Written by BroytMann. Copyright (C) 1997-2008 PhiloSoft Design
+ Written by Broytman. Copyright (C) 1997-2008 PhiloSoft Design
"""
import codecs
def recode_entities(title, charset):
output = []
for part in entity_re.split(title):
- if part not in ("&", "<", ">", ""e;") and \
+ if part not in ("&", "<", ">", """) and \
entity_re.match(part):
part = unichr(name2codepoint.get(part[1:-1], part)).encode(charset)
output.append(part)
if log: log("Parser %s.%s failed, trying next one." % (p.__module__, p.__name__))
converted_title = title = parser.title
- if not parser.charset:
+ if title and (not parser.charset):
try:
unicode(title, "ascii")
except UnicodeDecodeError:
if parser.charset:
parser.charset = parser.charset.lower().replace("windows-", "cp")
- if parser.charset and (
+ if title and parser.charset and (
(parser.charset <> universal_charset) or
((not charset) or (charset <> parser.charset))):
try:
else:
if log: log(" title : %s" % title)
- final_title = recode_entities(converted_title, universal_charset)
- parts = [s.strip() for s in final_title.replace('\r', '').split('\n')]
- final_title = ' '.join([s for s in parts if s])
- if log and (final_title <> converted_title): log(" final title : %s" % final_title)
- parser.title = final_title
+ if title:
+ final_title = recode_entities(converted_title, universal_charset)
+ parts = [s.strip() for s in final_title.replace('\r', '').split('\n')]
+ final_title = ' '.join([s for s in parts if s])
+ if log and (final_title <> converted_title): log(" final title : %s" % final_title)
+ parser.title = final_title
return parser
if __name__ == '__main__':
import sys
- from m_lib.defenc import default_encoding
- current_charset = default_encoding.replace("windows-", "cp")
- parser = parse_html(sys.argv[1], universal_charset,
- log=lambda s: sys.stdout.write(s + '\n'))
+ l = len(sys.argv)
+ if l == 3:
+ filename = sys.argv[1]
+ charset = sys.argv[2]
+ elif l == 2:
+ filename = sys.argv[1]
+ charset = universal_charset
+ else:
+ sys.exit("Usage: %s filename [charset]" % sys.argv[0])
+
+ parser = parse_html(filename, charset, log=lambda s: sys.stdout.write(s + '\n'))
print " refresh:", parser.refresh
print " icon :", parser.icon