"""HTML Parsers
This file is a part of Bookmarks database and Internet robot.
+
"""
-__version__ = "$Revision$"[11:-2]
-__revision__ = "$Id$"[5:-2]
-__date__ = "$Date$"[7:-2]
__author__ = "Oleg Broytman <phd@phdru.name>"
-__copyright__ = "Copyright (C) 1997-2011 PhiloSoft Design"
+__copyright__ = "Copyright (C) 1997-2017 PhiloSoft Design"
__license__ = "GNU GPL"
-__all__ = ['parse_html', 'universal_charset']
+__all__ = ['parse_html', 'parse_filename', 'universal_charset']
import codecs
import os
BKMK_DEBUG_HTML_PARSERS = os.environ.get("BKMK_DEBUG_HTML_PARSERS")
-def parse_html(filename, charset=None, log=None):
+def parse_html(html_text, charset=None, log=None):
if not parsers:
return None
parser = None
for c in charsets:
try:
- parser = p(filename, c, log)
+ parser = p(html_text, c, log)
except UnicodeError:
pass
else:
parser.charset = parser.charset.lower().replace("windows-", "cp")
if title and parser.charset and (
- (parser.charset <> universal_charset) or
- ((not charset) or (charset <> parser.charset))):
+ (parser.charset != universal_charset) or
+ ((not charset) or (charset != parser.charset))):
try:
if parser.meta_charset:
if log: log(" META charset : %s" % parser.charset)
- elif (not charset) or (charset <> parser.charset):
+ elif (not charset) or (charset != parser.charset):
if log: log(" guessed charset: %s" % parser.charset)
#if log: log(" current charset: %s" % universal_charset)
if log: log(" title : %s" % title)
- if parser.charset <> universal_charset:
+ if parser.charset != universal_charset:
try:
converted_title = unicode(title, parser.charset).encode(universal_charset)
except UnicodeError:
if log: log(" incorrect conversion from %s, converting from %s" % (parser.charset, DEFAULT_CHARSET))
converted_title = unicode(title, DEFAULT_CHARSET, "replace").encode(universal_charset, "replace")
parser.charset = DEFAULT_CHARSET
- if log and (converted_title <> title): log(" converted title: %s" % converted_title)
+ if log and (converted_title != title): log(" converted title: %s" % converted_title)
except LookupError:
if log: log(" unknown charset: '%s'" % parser.charset)
else:
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)
+ if log and (final_title != converted_title): log(" final title : %s" % final_title)
parser.title = final_title
icon = parser.icon
if parser.charset:
parser.icon = icon.encode(parser.charset)
return parser
+
+def parse_filename(filename, charset=None, log=None):
+ fp = open(filename, 'r')
+ try:
+ parser = parse_html(fp.read(), charset=charset, log=log)
+ finally:
+ fp.close()
+ return parser