简单的语法解析器(附带fread的性能分析)

自己写的一个读取资源列表的源码,思想来自lua的单词分割解析模块,但是没有像lua一样通过buf来储存每个WORD,采用0x00字符串结束符来分割各个WORD,有点类似投机取巧的感觉,不是绝对安全的方法,所以并不推荐这样写,只是希望能汲取里面解析文件的思路

/*****************************   
*        ResManager.h        *   
*****************************/
#pragma once
  
struct img_member
{
    wchar_t *name;
    wchar_t *path;
    bool mask;
};
  
struct aimg_member
{
    wchar_t *name;
    wchar_t *path;
    int hnum;
    int vnum;
};
  
struct music_member
{
    char *name;
    char *path;
};
  
class ResManagerClass
{
public:
    std::list<img_member*> imglist;
    std::list<aimg_member*> aimglist;
    std::list<music_member*> musiclist;
  
    bool LoadResArray(const char *resfile);
  
private:
    FILE *f;
  
    char *buf;
    unsigned long filelen;
};
  
extern ResManagerClass ResManager;                                                                                                         
                                                                                                  
                                                  
/*****************************   
*       ResManager.cpp       *   
*****************************/
#include "StdAfx.h"
#include "resmanager.h"
using namespace std;
  
//每种格式的个数
#define MEMBER_NUM_IMG 3
#define MEMBER_NUM_AIMG 4
#define MEMBER_NUM_MUSIC 2
enum ResStyle {img,aimg,music,none};
  
ResManagerClass ResManager;
  
  
bool ResManagerClass::LoadResArray(const char *resfile)
{
    if(fopen_s(&f,resfile,"rb"))
        return false;
  
    fseek(f,0,2); filelen = ftell(f); fseek(f,0,0);
    buf = new char[filelen];
    fread(buf,filelen,1,f);
  
    //初始化各种mark
    bool in_note = false, in_string = false;
    int string_strat_pos = -1, mn_img = 0, mn_aimg = 0, mn_music = 0;
    char *string = 0;
    ResStyle style_entry = none;
    char *imgmember_str[MEMBER_NUM_IMG], *aimgmember_str[MEMBER_NUM_AIMG], *musicmember_str[MEMBER_NUM_MUSIC];
  
    for(unsigned long p=0;p<filelen;p++)
    {
        switch(buf[p])
        {
        case ' ': case '\f': case '\t': case '\v':      //跳过空格制表符
        case '\n': case '\r':                           //跳过换行
            if(in_note) break;
            if(in_string)
            {
                buf[p] = 0;
  
                int slen = p-string_strat_pos + 1;
                string = new char[slen];
                strcpy_s(string,slen,buf+string_strat_pos);
  
                string_strat_pos = -1;
                in_string = false;
            }
            break;
  
        //注释
        case '/':
            if(buf[p+1]=='*') {p++; in_note = true;}
            break;
        case '*':
            if(buf[p+1]=='/')
                if(in_note) {p++; in_note = false;};
            break;
  
        //类型入口
        case ':':
            if(buf[p+1]=='i'&&buf[p+2]=='m'&&buf[p+3]=='g') {style_entry=img;p+=3;}
            else if(buf[p+1]=='a'&&buf[p+2]=='i'&&buf[p+3]=='m'&&buf[p+4]=='g') {
                style_entry=aimg;p+=4;}
            else if(buf[p+1]=='m'&&buf[p+2]=='u'&&buf[p+3]=='s'&&buf[p+4]=='i'&&buf[p+5]=='c') {
                style_entry=music;p+=5;}
            else style_entry=none;
            break;
  
  
        //String
        default:
            if(in_note) break;
  
            if(!in_string)
            {
                string_strat_pos = p;
                in_string = true;
            }
            else
                //到达文件尾
                if(p==filelen-1)
                {
                    int slen = p-string_strat_pos + 2;
                    string = new char[slen];
                    for(int i=0;i<slen+1;i++)
                        string[i] = buf[string_strat_pos++];
                    string[slen-1]=0;
                }
  
            break;
        }
  
        //接收到string的话
        if(string)
        {
            //确定资源类型
            switch(style_entry)
            {
            case img:
                //将获得的string放入resarray_member
                if(mn_img < MEMBER_NUM_IMG)
                {
                    imgmember_str[mn_img] = string;
                    string = 0;
                }
                //string达到数量时,放入一个pic_member中,并push list
                if(++mn_img == MEMBER_NUM_IMG)
                {
                    img_member *im = new img_member;
                    im->name = ctow(imgmember_str[0]);
                    im->path = ctow(imgmember_str[1]);
                    im->mask = !strcmp("yes",imgmember_str[2]);
  
                    imglist.push_back(im);
  
                    delete[] imgmember_str[0];
                    delete[] imgmember_str[1];
                    delete[] imgmember_str[2];
  
                    mn_img=0;
                }
                break;
            case aimg:
                if(mn_aimg < MEMBER_NUM_AIMG)
                {
                    aimgmember_str[mn_aimg] = string;
                    string = 0;
                }
                if(++mn_aimg == MEMBER_NUM_AIMG)
                {
                    aimg_member *aim = new aimg_member;
                    aim->name = ctow(aimgmember_str[0]);
                    aim->path = ctow(aimgmember_str[1]);
                    aim->hnum= atoi(aimgmember_str[2]);
                    aim->vnum= atoi(aimgmember_str[3]);
  
                    aimglist.push_back(aim);
  
                    delete[] aimgmember_str[0];
                    delete[] aimgmember_str[1];
                    delete[] aimgmember_str[2];
                    delete[] aimgmember_str[3];
  
                    mn_aimg=0;
                }
                break;
            case music:
                if(mn_music < MEMBER_NUM_MUSIC)
                {
                    musicmember_str[mn_music] = string;
                    string = 0;
                }
                if(++mn_music == MEMBER_NUM_IMG)
                {
                    music_member *micm = new music_member;
                    micm->name = musicmember_str[0];
                    micm->path = musicmember_str[1];
  
                    musiclist.push_back(micm);
  
                    mn_music=0;
                }
                break;
            }
        }
    }
  
    return true;
}

这里要说的主要是提出一个意见,在需要高效解析文件时应该使用c语言风格的fread,而不应该使用iostream或者fstream的格式化输入,因为那样会消耗将近几倍的时间,但是在数据量比较小的情况下fstream是不错的选择,省去了自己写解析分割WORD的实现,所以在面对需要从文件读取信息时需要谨慎考虑应该选用的方法.

各种输入函数在不同平台下读取一千万个随机数所用的时间

方法/平台/时间(秒)	Linux gcc	Windows mingw	Windows VC2008
scanf	2.010	3.704	3.425
cin	6.380	64.003	19.208
cin取消同步	2.050	6.004	19.616
fread	0.290	0.241	0.304
read	0.290	0.398	不支持
mmap	0.250	不支持	不支持
Pascal read	2.160	4.668

来自:http://www.byvoid.com/blog/fast-readfile/

转载于:https://my.oschina.net/gal/blog/141389