为了探录c++ 风格的fstream与 C 风格(例如fread 和 fwrite )两种读写文件的方法的效率,我特意做了两个实验。
我的机器是Windows XP, Visual Studio 2008
1. 测试写文件速度
程序设计思路: 将TEST_SIZE个字符用两种方式写入文件,记录两种方式的耗时。
实验代码:
[cpp] view plain copy void test_write() { const int TEST_SIZE = 10000000 ; const char* c_plus_write_file = "H://c_plus_write_file.txt"; const char* c_write_file = "H://c_write_file.txt"; cout<<"Test size :" << TEST_SIZE <<endl; //c++ style writing file ofstream of(c_plus_write_file); assert(of); time_t start, end; start = clock(); for(int i=0; i < TEST_SIZE; ++i) { char tmp[1]; tmp[0] = char(i); of << tmp[0]; } end = clock(); of.close(); cout<<"C++ style: "<<end - start <<" ms"<<endl; //c style writing file FILE* fp = fopen(c_write_file, "w"); start = clock(); for(int i=0; i < TEST_SIZE; ++i) { char tmp[1]; tmp[0] = char(i); fwrite( tmp, 1, 1, fp); } end = clock(); fclose(fp); cout<<"C style: "<<end - start <<" ms"<<endl; cin.get(); }
实验结果:
图1
图2
图3
**从图1、2、3,可以看出, ofstream 的 << 运算符 所耗时 是fwrite()的近三倍
把of<<的代码改成了: of.write(tmp,1); 后结果:
实验代码:
[cpp] view plain copy void test_write() { const int TEST_SIZE = 1000000 ; const char* c_plus_write_file = "H://c_plus_write_file.txt"; const char* c_write_file = "H://c_write_file.txt"; cout<<"Test size :" << TEST_SIZE <<endl; //c++ style writing file ofstream of(c_plus_write_file); assert(of); time_t start, end; start = clock(); for(int i=0; i < TEST_SIZE; ++i) { char tmp[1]; tmp[0] = char(i); of.write(tmp,1); } end = clock(); of.close(); cout<<"C++ style: "<<end - start <<" ms"<<endl; //c style writing file FILE* fp = fopen(c_write_file, "w"); start = clock(); for(int i=0; i < TEST_SIZE; ++i) { char tmp[1]; tmp[0] = char(i); fwrite( tmp, 1, 1, fp); } end = clock(); fclose(fp); cout<<"C style: "<<end - start <<" ms"<<endl; cin.get(); }
实验结果:
图4
图5
图6
对比图4 和 图1、 图5 和 图2、图6 和 图3, 可以看到 << 运算符没有 ofstream.write(), 快, 但两者还是没有 fwrite() 快
结论: 效率 fwrite() > ofstream.operator<<() > ofstream.write()
3. 下面做读文件的比较:
程序设计思路: 用两种方法去读一个近100M的文本,记录时间。
实验代码:
[cpp] view plain copy void test_read() { const char* read_file = "H://read4.txt"; const int BUF_SIZE = 1024 ; char buf[BUF_SIZE]; //c++ style writing file ifstream ifs(read_file,<a href="http://lib.csdn.net/base/ios" class='replace_word' title="iOS知识库" target='_blank' style='color:#df3434; font-weight:bold;'>iOS</a>::binary); assert(ifs); time_t start, end; start = clock(); while(!ifs.eof()) { ifs.read(buf,BUF_SIZE); } end = clock(); ifs.close(); cout<<"C++ style: "<<end - start <<" ms"<<endl; //c style writing file FILE* fp = fopen(read_file, "rb"); start = clock(); int len = 0; do { len = fread(buf,1,BUF_SIZE,fp); //cout<<len<<endl; }while(len != 0); end = clock(); fclose(fp); cout<<"C style: "<<end - start <<" ms"<<endl; cin.get(); }
实验结果:
图7
结论: 读取一个 100M 的文件, fread() 的效率 是 ifstream.read()的将近十倍! (此结论惊人!)