1 /*
2 ** The author disclaims copyright to this source code.
3 **
4 *************************************************************************
5 ** Implementation of the "simple" full-text-search tokenizer.
6 */
7 
8 #include <assert.h>
9 #if !defined(__APPLE__)
10 #include <malloc.h>
11 #else
12 #include <stdlib.h>
13 #endif
14 #include <stdio.h>
15 #include <string.h>
16 #include <ctype.h>
17 
18 #include "tokenizer.h"
19 
20 /* Duplicate a string; the caller must free() the returned string.
21  * (We don't use strdup() since it's not part of the standard C library and
22  * may not be available everywhere.) */
23 /* TODO(shess) Copied from fulltext.c, consider util.c for such
24 ** things. */
string_dup(const char * s)25 static char *string_dup(const char *s){
26   char *str = malloc(strlen(s) + 1);
27   strcpy(str, s);
28   return str;
29 }
30 
31 typedef struct simple_tokenizer {
32   sqlite3_tokenizer base;
33   const char *zDelim;          /* token delimiters */
34 } simple_tokenizer;
35 
36 typedef struct simple_tokenizer_cursor {
37   sqlite3_tokenizer_cursor base;
38   const char *pInput;          /* input we are tokenizing */
39   int nBytes;                  /* size of the input */
40   const char *pCurrent;        /* current position in pInput */
41   int iToken;                  /* index of next token to be returned */
42   char *zToken;                /* storage for current token */
43   int nTokenBytes;             /* actual size of current token */
44   int nTokenAllocated;         /* space allocated to zToken buffer */
45 } simple_tokenizer_cursor;
46 
47 static sqlite3_tokenizer_module simpleTokenizerModule;/* forward declaration */
48 
simpleCreate(int argc,const char ** argv,sqlite3_tokenizer ** ppTokenizer)49 static int simpleCreate(
50   int argc, const char **argv,
51   sqlite3_tokenizer **ppTokenizer
52 ){
53   simple_tokenizer *t;
54 
55   t = (simple_tokenizer *) malloc(sizeof(simple_tokenizer));
56   /* TODO(shess) Delimiters need to remain the same from run to run,
57   ** else we need to reindex.  One solution would be a meta-table to
58   ** track such information in the database, then we'd only want this
59   ** information on the initial create.
60   */
61   if( argc>1 ){
62     t->zDelim = string_dup(argv[1]);
63   } else {
64     /* Build a string excluding alphanumeric ASCII characters */
65     char zDelim[0x80];               /* nul-terminated, so nul not a member */
66     int i, j;
67     for(i=1, j=0; i<0x80; i++){
68       if( !isalnum(i) ){
69         zDelim[j++] = i;
70       }
71     }
72     zDelim[j++] = '\0';
73     assert( j<=sizeof(zDelim) );
74     t->zDelim = string_dup(zDelim);
75   }
76 
77   *ppTokenizer = &t->base;
78   return SQLITE_OK;
79 }
80 
simpleDestroy(sqlite3_tokenizer * pTokenizer)81 static int simpleDestroy(sqlite3_tokenizer *pTokenizer){
82   simple_tokenizer *t = (simple_tokenizer *) pTokenizer;
83 
84   free((void *) t->zDelim);
85   free(t);
86 
87   return SQLITE_OK;
88 }
89 
simpleOpen(sqlite3_tokenizer * pTokenizer,const char * pInput,int nBytes,sqlite3_tokenizer_cursor ** ppCursor)90 static int simpleOpen(
91   sqlite3_tokenizer *pTokenizer,
92   const char *pInput, int nBytes,
93   sqlite3_tokenizer_cursor **ppCursor
94 ){
95   simple_tokenizer_cursor *c;
96 
97   c = (simple_tokenizer_cursor *) malloc(sizeof(simple_tokenizer_cursor));
98   c->pInput = pInput;
99   c->nBytes = nBytes<0 ? (int) strlen(pInput) : nBytes;
100   c->pCurrent = c->pInput;        /* start tokenizing at the beginning */
101   c->iToken = 0;
102   c->zToken = NULL;               /* no space allocated, yet. */
103   c->nTokenBytes = 0;
104   c->nTokenAllocated = 0;
105 
106   *ppCursor = &c->base;
107   return SQLITE_OK;
108 }
109 
simpleClose(sqlite3_tokenizer_cursor * pCursor)110 static int simpleClose(sqlite3_tokenizer_cursor *pCursor){
111   simple_tokenizer_cursor *c = (simple_tokenizer_cursor *) pCursor;
112 
113   if( NULL!=c->zToken ){
114     free(c->zToken);
115   }
116   free(c);
117 
118   return SQLITE_OK;
119 }
120 
simpleNext(sqlite3_tokenizer_cursor * pCursor,const char ** ppToken,int * pnBytes,int * piStartOffset,int * piEndOffset,int * piPosition)121 static int simpleNext(
122   sqlite3_tokenizer_cursor *pCursor,
123   const char **ppToken, int *pnBytes,
124   int *piStartOffset, int *piEndOffset, int *piPosition
125 ){
126   simple_tokenizer_cursor *c = (simple_tokenizer_cursor *) pCursor;
127   simple_tokenizer *t = (simple_tokenizer *) pCursor->pTokenizer;
128   int ii;
129 
130   while( c->pCurrent-c->pInput<c->nBytes ){
131     int n = (int) strcspn(c->pCurrent, t->zDelim);
132     if( n>0 ){
133       if( n+1>c->nTokenAllocated ){
134         c->zToken = realloc(c->zToken, n+1);
135       }
136       for(ii=0; ii<n; ii++){
137         /* TODO(shess) This needs expansion to handle UTF-8
138         ** case-insensitivity.
139         */
140         char ch = c->pCurrent[ii];
141         c->zToken[ii] = (unsigned char)ch<0x80 ? tolower((unsigned char)ch):ch;
142       }
143       c->zToken[n] = '\0';
144       *ppToken = c->zToken;
145       *pnBytes = n;
146       *piStartOffset = (int) (c->pCurrent-c->pInput);
147       *piEndOffset = *piStartOffset+n;
148       *piPosition = c->iToken++;
149       c->pCurrent += n + 1;
150 
151       return SQLITE_OK;
152     }
153     c->pCurrent += n + 1;
154     /* TODO(shess) could strspn() to skip delimiters en masse.  Needs
155     ** to happen in two places, though, which is annoying.
156     */
157   }
158   return SQLITE_DONE;
159 }
160 
161 static sqlite3_tokenizer_module simpleTokenizerModule = {
162   0,
163   simpleCreate,
164   simpleDestroy,
165   simpleOpen,
166   simpleClose,
167   simpleNext,
168 };
169 
get_simple_tokenizer_module(sqlite3_tokenizer_module ** ppModule)170 void get_simple_tokenizer_module(
171   sqlite3_tokenizer_module **ppModule
172 ){
173   *ppModule = &simpleTokenizerModule;
174 }
175