Posted on 2010-07-28 21:46
roy 閱讀(2173)
評(píng)論(10) 編輯 收藏 引用 所屬分類(lèi):
C/C++
strcat是C語(yǔ)言一個(gè)基本的字符串操作函數(shù)，它的源代碼一般是這樣的。
char *strcat(char *dest, const char *src)
{
char *tmp = dest;
while (*dest) dest++;
while ((*dest++ = *src++) != '\0');
return tmp;
} 由此可見(jiàn)，strcat調(diào)用時(shí)，先移動(dòng)目標(biāo)字符串的指針到其尾部，再進(jìn)行復(fù)制。這種做法對(duì)于下標(biāo)比較大的數(shù)組重復(fù)調(diào)用時(shí)，效率比較低。想象一下，第一次調(diào)用strcat時(shí)，指針由0數(shù)到100，只不過(guò)復(fù)制了幾個(gè)字符，第二次調(diào)用strcat時(shí)，指針又從0數(shù)到108，無(wú)論調(diào)用多少次，指針總是從0數(shù)起，就會(huì)知道這個(gè)時(shí)候是多么浪費(fèi)系統(tǒng)資源了！
我找到一個(gè)辦法，字符串追加時(shí)，事先給出目標(biāo)字符串結(jié)尾所在的位置，追加時(shí)，也就不用從頭開(kāi)始計(jì)算其長(zhǎng)度了，復(fù)制的過(guò)程中，目標(biāo)字符串的結(jié)尾也隨之移動(dòng)，下一次再追加也就可以使用它了。以下就是優(yōu)化過(guò)的string_append，與strcat相比，增加了一個(gè)整形指針以傳遞目標(biāo)字符串長(zhǎng)度的地址。
/*
* optimizer for strcat when appending to a large array again and again
*/
char *string_append(char *dest, int *end, const char *src) {
if ( *end >= 0 && dest && src ) {
char *p = dest + *end;
while ( *p++ = *src++ ) (*end)++;
}
return dest;
}
經(jīng)試驗(yàn)，string_append在大數(shù)組重復(fù)追加內(nèi)容的情形下，優(yōu)勢(shì)非常明顯。其它情形下，使用原來(lái)的strcat也就足夠了。
#include <stdio.h>
#include <string.h>
#include <time.h>
#define BUFF_SIZE 4096
/*
* optimizer for strcat when appending to a large array again and again
*/
char *string_append(char *dest, int *end, const char *src) {
if ( *end >= 0 && dest && src ) {
char *p = dest + *end;
while ( *p++ = *src++ ) (*end)++;
}
return dest;
}
int main() {
int i = 0, j = 0;
int retry = 100000;
int field = 100;
char output1[BUFF_SIZE], output2[BUFF_SIZE];
time_t time1 = time(NULL);
for ( i = 0; i < retry; i++ ) {
memset(output1, 0, BUFF_SIZE);
int length = 0;
string_append(output1, &length, "header\n");
for ( j = 0; j < field; j++ ) {
string_append(output1, &length, "\tcall detail record ");
char c[8];
sprintf(c, "%d", j);
string_append(output1, &length, c);
string_append(output1, &length, "\n");
}
string_append(output1, &length, "trailer\n");
}
time_t time2 = time(NULL);
printf("It takes %d seconds to show the performance of string_append()\n", time2 - time1);
time1 = time(NULL);
for ( i = 0; i < retry; i++ ) {
memset(output2, 0, BUFF_SIZE);
strcat(output2, "header\n");
for ( j = 0; j < field; j++ ) {
strcat(output2, "\tcall detail record ");
char c[8];
sprintf(c, "%d", j);
strcat(output2, c);
strcat(output2, "\n");
}
strcat(output2, "trailer\n");
}
time2 = time(NULL);
printf("It takes %d seconds to show the performance of strcat()\n", time2 - time1);
if ( strcmp(output1, output2) )
printf("They are NOT equal\n");
else
printf("They are equal\n");
return 0;
}
-bash-3.2$ ./string_append_demo
It takes 2 seconds to show the performance of string_append()
It takes 11 seconds to show the performance of strcat()
They are equal
本文來(lái)自CSDN博客，轉(zhuǎn)載請(qǐng)標(biāo)明出處：http://blog.csdn.net/yui/archive/2010/05/22/5616455.aspx