同一问题的两种不同解法 : MFC8.0 与 C++ 标准库

前两天写了一个对40M的有格式文本文件的信息抽取,开始使用 MFC的CFile来读取,存储到char*中,然后对每个字符循环读取,判断抽取有用信息.思路很简单,也很笨拙,所以也就没有奢望它的效率好到那里.但实际运行的时候还是非常令人吃惊,两个多小时竟然没有跑完; 又细看了一下代码,只做简单的优化是没有用处了,就改用了C++来重新写,fstream的标准库,string 做为buffer,又调用了string的一些方法,结果非常的好,用了30多秒,就跑完了. 现把代码贴出来 ,达人多指教.

文本文件是按行组织的,每一行的数据以空格分割,共44维数据,每一维是 Key:value的格式.现在要取出其中的 qid,1,13...等几维数据,每行的前后的信息也要保留.
0 qid:1 1:11.512826 2:288.000000 3:195.000000 4:6.000000 5:2.000000 6:0.223063 7:0.000858 8:0.000677 9:0.000018 10:0.000826 11:0.001129 12:0.003413 13:10.546871 14:0.000772 15:-14.725600 16:0.017391 17:-12.865100 18:-9.148330 19:-8.108850 20:0.569967 21:-8.119490 22:-7.974970 23:-7.379150 24:1.563410 25:-7.253540 26:-7.252590 27:10.546905 28:2.000000 29:0.000000 30:0.000000 31:0.000000 32:0.000035 33:0.000000 34:0.000000 35:0.000000 36:0.000249 37:0.143097 38:0.000607 39:0.004101 40:0.001772 41:0.001124 42:19.575701 43:12.622301 44:11.723003 #docid = 1032

第一种写法:

CExtractKeysDlg::CExtractKeysDlg(CWnd* pParent /**//*=NULL*/)
    : CDialog(CExtractKeysDlg::IDD, pParent)
    , openfileedit(_T(""))
    , savefileedit(_T(""))
    , saveDim(_T(""))
...{
    m_hIcon = AfxGetApp()->LoadIcon(IDR_MAINFRAME);
    openfileedit="d:/d.txt";
    savefileedit="d:/e.txt";
    saveDim="qid,1,13,27,28,39,40,41,42,43,44";
    //UpdateData(false);

}

void CExtractKeysDlg::OnBnClickedExtract()

...{
    // TODO: 在此添加控件通知处理程序代码
    const int readlen=1024*64; //读取64k大的文件
    char buf[readlen],wbuf[readlen];
    int truelen(0),wlen(0);
    int wflag(1);//记录需要写入几次的标志
    UpdateData(true);
    CFile cfopen(this->openfileedit,CFile::modeRead);
    CFile cfsave(this->savefileedit,CFile::modeCreate|CFile::modeWrite);
    
    do...{
        truelen=cfopen.Read(buf,readlen);
        int i(0),wbegin(0),wend(0),lastspace(0),wlast(0);

        for(;i<truelen;i++)...{
            switch(buf[i])...{
                case ' ':                       //记录空格位置,完成写任务
                    lastspace=i;
                    if(wflag)...{
                        wflag--;
                        wend=lastspace-1;
                        ArrayPut(buf,wbuf,wbegin,wend,wlast);
                        if (wflag) wbegin=i;

                    }
                    break;
                case ':':                  //检测,置写标记
                    if (IsIn(buf,lastspace+1,i-1))...{
                        wflag++;
                        wbegin=lastspace;
                    }
                    break;
                case 10:                    //换行之前的要写入
                    if(wflag)...{

                        wend=i;
                        ArrayPut(buf,wbuf,wbegin,wend,wlast);
                        wbegin=i+1;

                    }
                    break;
                case '#':                    //docid默认写入
                    wflag += 3;
                    wbegin=i-1;
                    break;

            }//switch
        }//for
        if(wflag && wend>=wbegin) ArrayPut(buf,wbuf,wbegin,wend,wlast);
        if(!wflag)...{
            
            cfopen.Seek(lastspace-truelen,CFile::current);
        }
        cfsave.Write(wbuf,wlast);
        //cfsave.Flush(); //先在内存中缓存,到一定数量再到文件中.
    }while(truelen==readlen);
    cfopen.Close();
    cfsave.Close();
//优化二,把几个小函数声明为内联函数,减小运行代价.
}

inline bool CExtractKeysDlg::ArrayPut(const char* src,char * dst,int begin,int end,int &lastpos)
...{
    //return false;
    for(;begin<=end;begin++)
        dst[lastpos++]=src[begin];
    return true;
}
inline bool CExtractKeysDlg::IsIn(const char* buf,int begin,int end)
...{
    CString substr(',');
    saveDim.Insert(0,',');
    saveDim.AppendChar(',');
    int sub(1);
    for(;begin<=end;begin++)
        substr.AppendChar(buf[begin]);
    substr.AppendChar(',');
    if (saveDim.Find(substr)<0)
        return false;
    else
        return true;
}

第二种写法:

#include "stdafx.h"
#include <fstream>
#include <vector>
#include <algorithm>
#include <string>
#include <sstream>
using namespace std;
/**//*---------------------------------------------------------------------------
                进取维数部分
-----------------------------------------------------------------------------*/
const string token(",1,qid,13,27,28,39,40,41,42,43,44,");
const int NUM=15;
//qid,1,13,27,28,39,40,41,42,43,44
inline bool IsIn(string sbuf)...{
    sbuf.push_back(',');
    sbuf.insert(0,",");
    if (token.find(sbuf)!=string::npos)
        return true;
    else
        return false;

}
/**//*
name:Extracotr
function:从原始数据中提到指定的几维
提取的维数定义在 token中
*/
void Extractor()...{
    string ifile("d:/TD2003.txt");
    cout << "please input file to extract:"<<endl;
    //cin>>ifile;
    ifstream infile(ifile.c_str() );
    string ofile("d:/TD03.txt");
    cout<<"please input the file to save as:"<<endl;
    //cin>>ofile;
    ofstream outfile(ofile.c_str());
    string buf,sbuf;
    //while(getline(infile,buf,' ')){
    //    outfile<<' ';
    //}
    int enter(1);
    string::size_type pos;
    while( infile>>buf )...{
        pos=buf.find(':');
        if (pos==string::npos)            
            outfile<<buf<<(enter++ % NUM? " ":" ");
        else
        ...{
            //buf.copy(sbuf,pos);
            sbuf=buf.substr(0,pos);
            if (IsIn(sbuf))
                outfile<<buf<<(enter++ % NUM? " ":" ");
        }
    }
    infile.close();
    outfile.close();
}