Marc/server
1
0
Fork 0
forked from github/server
Code Issues Pull requests Projects Releases Wiki Activity

Speeding up the hashing in findtoken even more.

Browse source
This commit is contained in:
Enno Rehling 2006-01-25 20:40:19 +00:00
parent 0fe3e9ad4b
commit 6c6c1174d8
2 changed files with 14 additions and 28 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

40
src/common/util/umlaut.c
View file

@ -36,21 +36,6 @@ typedef struct tref {
#define LEAF 1 /* leaf node for a word. always matches */ #define LEAF 1 /* leaf node for a word. always matches */
#define SHARED 2 /* at least two words share the node */ #define SHARED 2 /* at least two words share the node */
#if NODEHASHSIZE == 7
/* lookup table, making c % 7 faster for chars. is this sick or what? */
static int divc7[256] = {
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,0,1,2,3,4,5,6,
0,1,2,3 };
#endif
void void
addtoken(tnode * root, const char* str, variant id) addtoken(tnode * root, const char* str, variant id)
{ {
@ -73,10 +58,10 @@ addtoken(tnode * root, const char* str, variant id)
} else { } else {
tref * next; tref * next;
int index, i = 0; int index, i = 0;
char c = *str; register char c = *str;
if (c<'a' || c>'z') c = (char)tolower((unsigned char)c); if (c<'a' || c>'z') c = (char)tolower((unsigned char)c);
#if NODEHASHSIZE == 7 #if NODEHASHSIZE == 8
index = divc7[(unsigned char)c]; index = c & 7;
#else #else
index = ((unsigned char)c) % NODEHASHSIZE; index = ((unsigned char)c) % NODEHASHSIZE;
#endif #endif
@ -95,7 +80,11 @@ addtoken(tnode * root, const char* str, variant id)
root->next[index] = ref; root->next[index] = ref;
if (u!=c) { if (u!=c) {
index = ((unsigned char)u) % NODEHASHSIZE; #if NODEHASHSIZE == 8
index = u & 7;
#else
index = ((unsigned char)u) % NODEHASHSIZE;
#endif
ref = malloc(sizeof(tref)); ref = malloc(sizeof(tref));
ref->c = u; ref->c = u;
ref->node = node; ref->node = node;
@ -123,18 +112,15 @@ addtoken(tnode * root, const char* str, variant id)
int int
findtoken(const tnode * tk, const char * str, variant* result) findtoken(const tnode * tk, const char * str, variant* result)
{ {
if (!str) return E_TOK_NOMATCH; if (!str || *str==0) return E_TOK_NOMATCH;
if (*str == 0) return E_TOK_NOMATCH;
while (*str) { do {
int index; int index;
const tref * ref; const tref * ref;
char c = *str; char c = *str;
/* if (c<'a' || c>'z') c = (char)tolower((unsigned char)c); */ #if NODEHASHSIZE == 8
index = c & 7;
#if NODEHASHSIZE == 7
index = divc7[(unsigned char)c];
#else #else
index = ((unsigned char)c) % NODEHASHSIZE; index = ((unsigned char)c) % NODEHASHSIZE;
#endif #endif
@ -143,7 +129,7 @@ findtoken(const tnode * tk, const char * str, variant* result)
++str; ++str;
if (!ref) return E_TOK_NOMATCH; if (!ref) return E_TOK_NOMATCH;
tk = ref->node; tk = ref->node;
} } while (*str);
if (tk) { if (tk) {
*result = tk->id; *result = tk->id;
return E_TOK_SUCCESS; return E_TOK_SUCCESS;

2
src/common/util/umlaut.h
View file

@ -23,7 +23,7 @@ extern "C" {
#define E_TOK_NOMATCH (-1) #define E_TOK_NOMATCH (-1)
#define E_TOK_SUCCESS 0 #define E_TOK_SUCCESS 0
#define NODEHASHSIZE 7 #define NODEHASHSIZE 8
struct tref; struct tref;
typedef struct tnode { typedef struct tnode {