Bug #1755: base64_decode uses inefficient algorithm

From: bfranklin@dct.com Operating system: Sparc Solaris 2.7 PHP version: 4.0 Beta 1 PHP Bug Type: Performance problem Bug description: base64_decode uses inefficient algorithm Base64_decode uses strchr to lookup the location of a character in base64_table[], causing a function call and (on average) 32 comparisons per encoded character. It would be far more efficient to build a static array of reverse lookup values to use. Here's a patch to fix this: --- base64.c.orig Tue Jul 20 10:29:42 1999 +++ base64.c Tue Jul 20 10:51:54 1999 @@ -71,18 +71,34 @@ unsigned char *_php3_base64_decode(const unsigned char *string, int length, int *ret_length) { const unsigned char *current = string; int ch, i = 0, j = 0, k; + static int reverse_table[256]; + static int table_built = 0; unsigned char *result = (unsigned char *)emalloc((length / 4 * 3 + 1) * sizeof(char)); if (result == NULL) { return NULL; } + /* build table of reverse mappings to speed things up */ + if( ! table_built ) + { + char *chp; + for(ch=0;ch<256;ch++) + { + chp = strchr(base64_table, ch); + if( chp != NULL ) + reverse_table[ch] = chp - base64_table; + else + reverse_table[ch] = -1; + } + table_built = 1; + } + /* run through the whole string, converting as we go */ while ((ch = *current++) != '\0') { if (ch == base64_pad) break; - ch = (int)strchr(base64_table, ch); - if (ch == 0) continue; - ch -= (int)base64_table; + ch = reverse_table[ch]; + if (ch < 0) continue; switch(i % 4) { case 0: