CDocument

#ifndef _Document_H_040410_
#define _Document_H_040410_

#include <string>

typedef struct{
        int docid;
        int offset;
}DocIdx;

using namespace std;

class CDocument
{
public:

    int m_nDocId;
    int m_nPos;
    int m_nLength;
    string m_sChecksum;

    string m_sUrl;
    string m_sRecord;    // a record including a HEAD, a header and body
    string m_sHead;
    string m_sHeader;
    string m_sBody;

    string m_sBodyNoTags;

public:
    CDocument();
    ~CDocument();

    bool ParseRecord(string &content) const;
    bool CleanBody(string &body) const;

    void RemoveTags(char *s);
};

#endif /* _Document_H_040410_ */

 

 

/*Document handling
 */

#include "Document.h"

CDocument::CDocument()
{
    m_nDocId = -1;
    m_nPos = -1;
    m_nLength = 0;
    m_sChecksum = "";

    m_sUrl = "";
}

CDocument::~CDocument()
{
}

bool CDocument::ParseRecord(string &content) const
{
    return true;
}

bool CDocument::CleanBody(string &body) const
{
    return true;
}

//把  <...> 删掉
void CDocument::RemoveTags(char *s)
{
    int intag;
    char *p, *q;

    if (!s || !*s)    return;

    for (p=q=s, intag=0; *q; q++) {
        switch (*q){
        case '<':
            intag = 1;
            *p++ = ' ';
            break;
        case '>':
            intag = 0;
            break;
        default:
            if (!intag) {
                *p++ = *q;
            }
            break;
        }
    }

    *p = '\0';

/* second method
    char *d = s;
    while (*s) {
        if (*s == '<') {
            while (*s && *s!='>') s++;
            if( *s == '\0') break;
            s++; 
            continue; 
        } 

        *d++ = *s++; 
    }
    *d = 0;
*/
}