void parse_urls(const char *filename, const url_list_t *elem)
{
TidyDoc tdoc;
int err;
FILE *outfile = NULL;
tdoc = tidyCreate();
tidyOptSetBool(tdoc, TidyForceOutput, yes);
tidyOptSetBool(tdoc, TidyMark, no);
tidyOptSetBool(tdoc, TidyHideEndTags, yes);
tidyOptSetBool(tdoc, TidyDropEmptyParas, no);
tidyOptSetBool(tdoc, TidyJoinStyles, no);
tidyOptSetBool(tdoc, TidyPreserveEntities, yes);
tidyOptSetInt(tdoc, TidyMergeDivs, no);
tidyOptSetInt(tdoc, TidyMergeSpans, no);
tidyOptSetInt(tdoc, TidyWrapLen, 4096);
tidyOptSetValue(tdoc, TidyCharEncoding, "utf8");
tidySetReportFilter(tdoc, filter_cb);
err = tidyParseFile(tdoc, filename);
if (err >= 0)
err = tidyCleanAndRepair(tdoc);
if (err >= 0) {
outfile = option_values.save_relative_links && !option_values.disable_save_tree
? fopen(filename, "w") : NULL;
parse_html(tdoc, tidyGetRoot(tdoc), elem, 1, outfile);
if (outfile)
fclose(outfile);
}
tidyRelease(tdoc);
}
bool TidyReader::openFile (const char * szFilename)
{
UT_DEBUGMSG(("using libtidy to parse HTML...\n"));
m_tidy = tidyCreate ();
if (m_tidy == 0) return false;
if (tidyOptSetBool (m_tidy, TidyXhtmlOut, yes) == 0)
{
UT_DEBUGMSG(("tidyOptSetBool failed!\n"));
closeFile ();
return false;
}
#ifndef DEBUG
tidySetErrorBuffer (m_tidy, &m_errbuf);
#endif
int parse_status;
if (m_buffer && m_length)
{
UT_DEBUGMSG(("parse HTML in buffer...\n"));
UT_Byte * buffer = const_cast<UT_Byte *>(m_buffer); // grr.
TidyBuffer inbuf;
tidyBufInit (&inbuf);
tidyBufAttach (&inbuf, buffer, static_cast<unsigned int>(m_length));
parse_status = tidyParseBuffer (m_tidy, &inbuf);
tidyBufDetach (&inbuf);
}
else
{
UT_DEBUGMSG(("parse HTML in file: %s\n",szFilename));
parse_status = tidyParseFile (m_tidy, szFilename);
}
if (parse_status < 0)
{
UT_DEBUGMSG(("tidyParseBuffer/File failed!\n"));
closeFile ();
return false;
}
parse_status = tidyCleanAndRepair (m_tidy);
if (parse_status < 0)
{
UT_DEBUGMSG(("tidyCleanAndRepair failed!\n"));
closeFile ();
return false;
}
parse_status = tidyRunDiagnostics (m_tidy);
if (parse_status < 0)
{
UT_DEBUGMSG(("tidyRunDiagnostics failed!\n"));
closeFile ();
return false;
}
if (parse_status > 1)
{
parse_status = (tidyOptSetBool (m_tidy, TidyForceOutput, yes) ? parse_status : -1);
}
if (parse_status < 0)
{
UT_DEBUGMSG(("tidyOptSetBool failed!\n"));
closeFile ();
return false;
}
parse_status = tidySaveBuffer (m_tidy, &m_outbuf);
if (parse_status < 0)
{
UT_DEBUGMSG(("tidySaveBuffer failed!\n"));
closeFile ();
return false;
}
UT_DEBUGMSG(("tidy succeeded!\n"));
#ifdef DEBUG
fputs ("================================================================\n", stderr);
fputs ((const char *) m_outbuf.bp, stderr);
fputs ("================================================================\n", stderr);
#endif
m_outbuf.next = 0;
return true;
}
int main( int argc, char** argv )
{
ctmbstr prog = argv[0];
ctmbstr cfgfil = NULL, errfil = NULL, htmlfil = NULL;
TidyDoc tdoc = tidyCreate();
int status = 0;
uint contentErrors = 0;
uint contentWarnings = 0;
uint accessWarnings = 0;
errout = stderr; /* initialize to stderr */
status = 0;
#ifdef TIDY_CONFIG_FILE
if ( tidyFileExists( tdoc, TIDY_CONFIG_FILE) )
{
status = tidyLoadConfig( tdoc, TIDY_CONFIG_FILE );
if ( status != 0 )
fprintf(errout, "Loading config file \"%s\" failed, err = %d\n", TIDY_CONFIG_FILE, status);
}
#endif /* TIDY_CONFIG_FILE */
/* look for env var "HTML_TIDY" */
/* then for ~/.tidyrc (on platforms defining $HOME) */
if ( (cfgfil = getenv("HTML_TIDY")) != NULL )
{
status = tidyLoadConfig( tdoc, cfgfil );
if ( status != 0 )
fprintf(errout, "Loading config file \"%s\" failed, err = %d\n", cfgfil, status);
}
#ifdef TIDY_USER_CONFIG_FILE
else if ( tidyFileExists( tdoc, TIDY_USER_CONFIG_FILE) )
{
status = tidyLoadConfig( tdoc, TIDY_USER_CONFIG_FILE );
if ( status != 0 )
fprintf(errout, "Loading config file \"%s\" failed, err = %d\n", TIDY_USER_CONFIG_FILE, status);
}
#endif /* TIDY_USER_CONFIG_FILE */
/* read command line */
while ( argc > 0 )
{
if (argc > 1 && argv[1][0] == '-')
{
/* support -foo and --foo */
ctmbstr arg = argv[1] + 1;
if ( strcasecmp(arg, "xml") == 0)
tidyOptSetBool( tdoc, TidyXmlTags, yes );
else if ( strcasecmp(arg, "asxml") == 0 ||
strcasecmp(arg, "asxhtml") == 0 )
{
tidyOptSetBool( tdoc, TidyXhtmlOut, yes );
}
else if ( strcasecmp(arg, "ashtml") == 0 )
tidyOptSetBool( tdoc, TidyHtmlOut, yes );
else if ( strcasecmp(arg, "indent") == 0 )
{
tidyOptSetInt( tdoc, TidyIndentContent, TidyAutoState );
if ( tidyOptGetInt(tdoc, TidyIndentSpaces) == 0 )
tidyOptResetToDefault( tdoc, TidyIndentSpaces );
}
else if ( strcasecmp(arg, "omit") == 0 )
tidyOptSetBool( tdoc, TidyHideEndTags, yes );
else if ( strcasecmp(arg, "upper") == 0 )
tidyOptSetBool( tdoc, TidyUpperCaseTags, yes );
else if ( strcasecmp(arg, "clean") == 0 )
tidyOptSetBool( tdoc, TidyMakeClean, yes );
else if ( strcasecmp(arg, "bare") == 0 )
tidyOptSetBool( tdoc, TidyMakeBare, yes );
else if ( strcasecmp(arg, "raw") == 0 ||
strcasecmp(arg, "ascii") == 0 ||
strcasecmp(arg, "latin0") == 0 ||
strcasecmp(arg, "latin1") == 0 ||
strcasecmp(arg, "utf8") == 0 ||
#ifndef NO_NATIVE_ISO2022_SUPPORT
strcasecmp(arg, "iso2022") == 0 ||
#endif
#if SUPPORT_UTF16_ENCODINGS
strcasecmp(arg, "utf16le") == 0 ||
strcasecmp(arg, "utf16be") == 0 ||
strcasecmp(arg, "utf16") == 0 ||
#endif
#if SUPPORT_ASIAN_ENCODINGS
strcasecmp(arg, "shiftjis") == 0 ||
strcasecmp(arg, "big5") == 0 ||
#endif
strcasecmp(arg, "mac") == 0 ||
strcasecmp(arg, "win1252") == 0 ||
strcasecmp(arg, "ibm858") == 0 )
{
tidySetCharEncoding( tdoc, arg );
}
else if ( strcasecmp(arg, "numeric") == 0 )
tidyOptSetBool( tdoc, TidyNumEntities, yes );
else if ( strcasecmp(arg, "modify") == 0 ||
strcasecmp(arg, "change") == 0 || /* obsolete */
strcasecmp(arg, "update") == 0 ) /* obsolete */
{
tidyOptSetBool( tdoc, TidyWriteBack, yes );
}
else if ( strcasecmp(arg, "errors") == 0 )
tidyOptSetBool( tdoc, TidyShowMarkup, no );
else if ( strcasecmp(arg, "quiet") == 0 )
tidyOptSetBool( tdoc, TidyQuiet, yes );
else if ( strcasecmp(arg, "help") == 0 ||
strcasecmp(arg, "h") == 0 || *arg == '?' )
{
help( prog );
tidyRelease( tdoc );
return 0; /* success */
}
else if ( strcasecmp(arg, "xml-help") == 0)
{
xml_help( );
tidyRelease( tdoc );
return 0; /* success */
}
else if ( strcasecmp(arg, "help-config") == 0 )
{
optionhelp( tdoc );
tidyRelease( tdoc );
return 0; /* success */
}
else if ( strcasecmp(arg, "xml-config") == 0 )
{
XMLoptionhelp( tdoc );
tidyRelease( tdoc );
return 0; /* success */
}
else if ( strcasecmp(arg, "show-config") == 0 )
{
optionvalues( tdoc );
tidyRelease( tdoc );
return 0; /* success */
}
else if ( strcasecmp(arg, "config") == 0 )
{
if ( argc >= 3 )
{
ctmbstr post;
tidyLoadConfig( tdoc, argv[2] );
/* Set new error output stream if setting changed */
post = tidyOptGetValue( tdoc, TidyErrFile );
if ( post && (!errfil || !samefile(errfil, post)) )
{
errfil = post;
errout = tidySetErrorFile( tdoc, post );
}
--argc;
++argv;
}
}
#if SUPPORT_ASIAN_ENCODINGS
else if ( strcasecmp(arg, "language") == 0 ||
strcasecmp(arg, "lang") == 0 )
{
if ( argc >= 3 )
{
tidyOptSetValue( tdoc, TidyLanguage, argv[2] );
--argc;
++argv;
}
}
#endif
else if ( strcasecmp(arg, "output") == 0 ||
strcasecmp(arg, "-output-file") == 0 ||
strcasecmp(arg, "o") == 0 )
{
if ( argc >= 3 )
{
tidyOptSetValue( tdoc, TidyOutFile, argv[2] );
--argc;
++argv;
}
}
else if ( strcasecmp(arg, "file") == 0 ||
strcasecmp(arg, "-file") == 0 ||
strcasecmp(arg, "f") == 0 )
{
if ( argc >= 3 )
{
errfil = argv[2];
errout = tidySetErrorFile( tdoc, errfil );
--argc;
++argv;
}
}
else if ( strcasecmp(arg, "wrap") == 0 ||
strcasecmp(arg, "-wrap") == 0 ||
strcasecmp(arg, "w") == 0 )
{
if ( argc >= 3 )
{
uint wraplen = 0;
int nfields = sscanf( argv[2], "%u", &wraplen );
tidyOptSetInt( tdoc, TidyWrapLen, wraplen );
if (nfields > 0)
{
--argc;
++argv;
}
}
}
else if ( strcasecmp(arg, "version") == 0 ||
strcasecmp(arg, "-version") == 0 ||
strcasecmp(arg, "v") == 0 )
{
version();
tidyRelease( tdoc );
return 0; /* success */
}
else if ( strncmp(argv[1], "--", 2 ) == 0)
{
if ( tidyOptParseValue(tdoc, argv[1]+2, argv[2]) )
{
/* Set new error output stream if setting changed */
ctmbstr post = tidyOptGetValue( tdoc, TidyErrFile );
if ( post && (!errfil || !samefile(errfil, post)) )
{
errfil = post;
errout = tidySetErrorFile( tdoc, post );
}
++argv;
--argc;
}
}
#if SUPPORT_ACCESSIBILITY_CHECKS
else if ( strcasecmp(arg, "access") == 0 )
{
if ( argc >= 3 )
{
uint acclvl = 0;
int nfields = sscanf( argv[2], "%u", &acclvl );
tidyOptSetInt( tdoc, TidyAccessibilityCheckLevel, acclvl );
if (nfields > 0)
{
--argc;
++argv;
}
}
}
#endif
else
{
uint c;
ctmbstr s = argv[1];
while ( (c = *++s) != '\0' )
{
switch ( c )
{
case 'i':
tidyOptSetInt( tdoc, TidyIndentContent, TidyAutoState );
if ( tidyOptGetInt(tdoc, TidyIndentSpaces) == 0 )
tidyOptResetToDefault( tdoc, TidyIndentSpaces );
break;
/* Usurp -o for output file. Anyone hiding end tags?
case 'o':
tidyOptSetBool( tdoc, TidyHideEndTags, yes );
break;
*/
case 'u':
tidyOptSetBool( tdoc, TidyUpperCaseTags, yes );
break;
case 'c':
tidyOptSetBool( tdoc, TidyMakeClean, yes );
break;
case 'b':
tidyOptSetBool( tdoc, TidyMakeBare, yes );
break;
case 'n':
tidyOptSetBool( tdoc, TidyNumEntities, yes );
break;
case 'm':
tidyOptSetBool( tdoc, TidyWriteBack, yes );
break;
case 'e':
tidyOptSetBool( tdoc, TidyShowMarkup, no );
break;
case 'q':
tidyOptSetBool( tdoc, TidyQuiet, yes );
break;
default:
unknownOption( c );
break;
}
}
}
--argc;
++argv;
continue;
}
if ( argc > 1 )
{
htmlfil = argv[1];
if ( tidyOptGetBool(tdoc, TidyEmacs) )
tidyOptSetValue( tdoc, TidyEmacsFile, htmlfil );
status = tidyParseFile( tdoc, htmlfil );
}
else
{
htmlfil = "stdin";
status = tidyParseStdin( tdoc );
}
if ( status >= 0 )
status = tidyCleanAndRepair( tdoc );
if ( status >= 0 )
status = tidyRunDiagnostics( tdoc );
if ( status > 1 ) /* If errors, do we want to force output? */
status = ( tidyOptGetBool(tdoc, TidyForceOutput) ? status : -1 );
if ( status >= 0 && tidyOptGetBool(tdoc, TidyShowMarkup) )
{
if ( tidyOptGetBool(tdoc, TidyWriteBack) && argc > 1 )
status = tidySaveFile( tdoc, htmlfil );
else
{
ctmbstr outfil = tidyOptGetValue( tdoc, TidyOutFile );
if ( outfil )
status = tidySaveFile( tdoc, outfil );
else
status = tidySaveStdout( tdoc );
}
}
contentErrors += tidyErrorCount( tdoc );
contentWarnings += tidyWarningCount( tdoc );
accessWarnings += tidyAccessWarningCount( tdoc );
--argc;
++argv;
if ( argc <= 1 )
break;
}
if (!tidyOptGetBool(tdoc, TidyQuiet) &&
errout == stderr && !contentErrors)
fprintf(errout, "\n");
if (contentErrors + contentWarnings > 0 &&
!tidyOptGetBool(tdoc, TidyQuiet))
tidyErrorSummary(tdoc);
if (!tidyOptGetBool(tdoc, TidyQuiet))
tidyGeneralInfo(tdoc);
/* called to free hash tables etc. */
tidyRelease( tdoc );
/* return status can be used by scripts */
if ( contentErrors > 0 )
return 2;
if ( contentWarnings > 0 )
return 1;
/* 0 signifies all is ok */
return 0;
}
