// -------------------------------------------------------------
void Webpage::tidy_me()
{
try {
TidyDoc _tdoc = tidyCreate();
//tidyOptSetBool(_tdoc, tidyOptGetIdForName("show-body-only"), (Bool)1);
tidyOptSetBool(_tdoc, tidyOptGetIdForName("output-xhtml"), (Bool)1);
tidyOptSetBool(_tdoc, tidyOptGetIdForName("quote-nbsp"), (Bool)0);
tidyOptSetBool(_tdoc, tidyOptGetIdForName("show-warnings"), (Bool)0);
tidyOptSetValue(_tdoc, tidyOptGetIdForName("char-encoding"), "utf8");
//tidyOptSetBool(_tdoc, tidyOptGetIdForName("ascii-chars"), (Bool)1);
//tidyOptSetBool(_tdoc, tidyOptGetIdForName("markup"), (Bool)1);
//tidyOptSetValue(_tdoc, tidyOptGetIdForName("indent"), "yes");
//tidyOptSetValue(_tdoc, tidyOptGetIdForName("newline"), "\n");
tidyOptSetInt(_tdoc, tidyOptGetIdForName("wrap"), 5000);
tidyParseString( _tdoc, contents.c_str() );
/*
// tidySaveBuffer doesn't seem to work with the makefile for some reason.
TidyBuffer output = {0};
tidySaveBuffer(_tdoc, &output);
cout << "3. TidyBuffer size: " << output.size << endl;
contents = string((char*)output.bp, (size_t)output.size);
*/
// tidySaveString is a tricky beast.
tmbstr buffer = NULL;
uint buflen = 0;
int status;
do {
status = tidySaveString( _tdoc, buffer, &buflen );
if (status == -ENOMEM) {
if(buffer)
free(buffer);
buffer = (tmbstr)malloc(buflen + 1);
}
} while (status == -ENOMEM);
contents = (char*)buffer;
} catch (exception& e) {
throw e.what();
}
}
/*!
* \fn static int TidyHtml(const char *pcSourcePage, string &sDestPage);
* \brief 修补丢失、错误标签
* \param [in]待修补网页字符串
* \param [out]修补后的网页string
* \return 结果码，==0修补正确，<0修补失败
* \date 2011-06-01
* \author nanjunxiao
*/
int Pretreat::TidyHtml(const char *pcSourcePage, std::string &sDestPage)
{
int iReturn = 0;
TidyBuffer errbuf = {0};
TidyDoc tdoc;
tmbstr pBuffer = NULL;
try
{
if ( (pcSourcePage == NULL) || (strlen(pcSourcePage) ==0 ) )
{
//cerr << "TidyHtml 输入页面为空！" << endl;
throw (-1);
}
int iRet = -1;
Bool bOk;
uint uiBufLen;
int iBufSize;
tdoc = tidyCreate();// Initialize "document"
bOk = tidyOptSetBool(tdoc, TidyXhtmlOut, yes);// Convert to XHTML
if (bOk)
{
iRet = tidySetErrorBuffer(tdoc, &errbuf); // Capture diagnostics
}
else
{
throw (-1);
}
if (iRet >= 0)
{
iRet = tidySetCharEncoding(tdoc,"utf8"); //Ensure dealing with gb2312 successfully
}
else
{
throw (-1);
}
if (iRet >= 0)
{
string htmlsrc = pcSourcePage;
iRet = tidyParseString (tdoc, htmlsrc.c_str() ); // Parse the input
}
else
{
throw (-1);
}
if (iRet >= 0)
{
iRet = tidyCleanAndRepair(tdoc); //Tidy it up!
}
else
{
throw (-1);
}
if (iRet >= 0)
{
iRet = tidyRunDiagnostics(tdoc); //Kvetch
}
else
{
throw (-1);
}
if(iRet > 1) // If error, force output.
{
iRet = ( tidyOptSetBool(tdoc, TidyForceOutput, yes) ? iRet : -1 );
}
else if (iRet < 0)
{
throw (-1);
}
if (iRet >= 0)
{
// Pretty Print
iBufSize = 1024 * 1024 * 5;
uiBufLen = iBufSize;
pBuffer = new char [iBufSize];
memset(pBuffer, '\0', iBufSize);
iRet = tidySaveString(tdoc, pBuffer, &uiBufLen);
}
else
{
throw (-1);
}
if (iRet >= 0)
{
sDestPage = pBuffer;
}
else if (iRet == -ENOMEM)
{
//pBuffer 长度不够
//cerr << "TidyHtml pBuffer长度不够！" << endl;
throw (-1);
}
else
{
throw (-1);
}
}
catch(exception &err)
{
//cerr << "TidyHtml HtmlTidy修补页面失败！ " << err.what() << endl;
iReturn = -1;
}
catch(int iThrow)
{
if (iThrow < 0)
{
//cerr << "TidyHtml HtmlTidy修补页面失败！" << endl;
}
iReturn = iThrow;
}
catch(...)
{
//cerr << "TidyHtml HtmlTidy修补页面失败！" << endl;
iReturn = -1;
}
tidyBufFree(&errbuf);
tidyRelease(tdoc);
if (pBuffer != NULL)
{
delete [] pBuffer;
pBuffer = NULL;
}
return iReturn;
}
