X-Git-Url: https://git.phdru.name/?a=blobdiff_plain;ds=sidebyside;f=parse_html%2Fbkmk_parse_html.py;h=e1d4710f4da4851009202bb89ef028a714bb59b7;hb=eeffe06ce42763abfff7797ab0ceeedbac4b1669;hp=f42dab8d4de79b040b4638b5d8d7806aa519b5d2;hpb=338c964afba3651bd8fe6318644c0fcabb66cc3b;p=bookmarks_db.git diff --git a/parse_html/bkmk_parse_html.py b/parse_html/bkmk_parse_html.py index f42dab8..e1d4710 100644 --- a/parse_html/bkmk_parse_html.py +++ b/parse_html/bkmk_parse_html.py @@ -29,25 +29,25 @@ else: parsers.append(bkmk_ph_beautifulsoup.parse_html) try: - from . import bkmk_ph_lxml + from . import bkmk_ph_htmlparser except ImportError: pass else: - parsers.append(bkmk_ph_lxml.parse_html) + parsers.append(bkmk_ph_htmlparser.parse_html) try: - from . import bkmk_ph_htmlparser + from . import bkmk_ph_html5 except ImportError: pass else: - parsers.append(bkmk_ph_htmlparser.parse_html) + parsers.append(bkmk_ph_html5.parse_html) try: - from . import bkmk_ph_html5 + from . import bkmk_ph_lxml except ImportError: pass else: - parsers.append(bkmk_ph_html5.parse_html) + parsers.append(bkmk_ph_lxml.parse_html) # ElementTidy often segfaults #try: @@ -86,6 +86,9 @@ def recode_entities(title, charset): return ''.join(output) +import os +BKMK_DEBUG_HTML_PARSERS = os.environ.get("BKMK_DEBUG_HTML_PARSERS") + def parse_html(filename, charset=None, log=None): if not parsers: return None @@ -103,24 +106,36 @@ def parse_html(filename, charset=None, log=None): charsets.remove(charset) charsets.insert(0, charset) + if BKMK_DEBUG_HTML_PARSERS: + _parsers = [] for p in parsers: parser = None for c in charsets: try: parser = p(filename, c, log) - except UnicodeEncodeError: + except UnicodeError: pass else: - break - if parser: - break + if parser: + if BKMK_DEBUG_HTML_PARSERS: + if log: log(" Parser %s: ok" % p.__module__) + _parsers.append((p, parser)) + break else: - if log: log("Parser %s.%s failed, trying next one." % (p.__module__, p.__name__)) + if log: log(" Parser %s: fail" % p.__module__) + if not BKMK_DEBUG_HTML_PARSERS and parser: + break - if not parser: - if log: log("All parser has failed.") + if BKMK_DEBUG_HTML_PARSERS: + if not _parsers: + if log: log(" All parsers have failed") + return None + elif not parser: + if log: log(" All parsers have failed") return None + if BKMK_DEBUG_HTML_PARSERS: + p, parser = _parsers[0] if log: log(" Using %s" % p.__module__) converted_title = title = parser.title