1diff --git a/CMakeLists.txt b/CMakeLists.txt
2index c34734d2daf984075b1bed87f236c2ac33e5ded2..fd48cee9086754409f9cec0ac69cc280720c3188 100644
3--- a/CMakeLists.txt
4+++ b/CMakeLists.txt
5@@ -7,5 +7,7 @@ add_executable(dict ${src})
6
7 target_compile_options(dict PRIVATE -Wall -Wextra -Wpedantic -Werror)
8 target_include_directories(dict PUBLIC "${PROJECT_BINARY_DIR}")
9-target_link_libraries(dict sqlite3 ncursesw)
10+target_link_libraries(dict sqlite3 ncursesw m)
11
12+
13+add_subdirectory(ext)
14diff --git a/data.c b/data.c
15index 81ee9f31356c495fe4754f9105a3211f863e9956..2c3f3bb0bbc811c9a8c3552e84d60f8c50714892 100644
16--- a/data.c
17+++ b/data.c
18@@ -11,14 +11,20 @@ Data* new_data(const char* con) {
19 Data* data = (Data*)malloc(sizeof(Data));
20
21 int v = sqlite3_open(con, &(data->db));
22- if (v == SQLITE_OK) {
23- return data;
24+ if (v != SQLITE_OK) {
25+ print_result_code(v);
26+ return NULL;
27 }
28
29- print_result_code(v);
30- return NULL;
31-}
32+ sqlite3_enable_load_extension(data->db, 1);
33+ v = sqlite3_load_extension(data->db, "ext/libsqlite3ext", "sqlite3_spellfix_init",0);
34+ if (v != SQLITE_OK) {
35+ print_result_code(v);
36+ return NULL;
37+ }
38
39+ return data;
40+}
41
42 void free_data(Data* data) {
43 sqlite3_close(data->db);
44@@ -103,85 +109,5 @@ return list;
45 }
46
47 void print_result_code(int code) {
48- switch(code) {
49- // Primary result code
50- case SQLITE_ABORT:
51- printf("SQLITE_ABORT");
52- break;
53- case SQLITE_AUTH:
54- printf("SQLITE_AUTH");
55- break;
56- case SQLITE_BUSY:
57- printf("SQLITE_BUSY");
58- break;
59- case SQLITE_CANTOPEN:
60- printf("SQLITE_CANTOPEN");
61- break;
62- case SQLITE_CONSTRAINT:
63- printf("SQLITE_CONSTRAINT");
64- break;
65- case SQLITE_CORRUPT:
66- printf("SQLITE_CORRUPT");
67- break;
68- case SQLITE_DONE:
69- printf("SQLITE_DONE");
70- break;
71- case SQLITE_EMPTY:
72- printf("SQLITE_EMPTY");
73- break;
74- case SQLITE_ERROR:
75- printf("SQLITE_ERROR");
76- break;
77- case SQLITE_FORMAT:
78- printf("SQLITE_FORMAT");
79- break;
80- case SQLITE_INTERNAL:
81- printf("SQLITE_INTERNAL");
82- break;
83- case SQLITE_INTERRUPT:
84- printf("SQLITE_INTERRUPT");
85- break;
86- case SQLITE_IOERR:
87- printf("SQLITE_IOERR");
88- break;
89- case SQLITE_LOCKED:
90- printf("SQLITE_LOCKED");
91- break;
92- case SQLITE_MISMATCH:
93- printf("SQLITE_MISMATCH");
94- break;
95- case SQLITE_MISUSE:
96- printf("SQLITE_MISUSE");
97- break;
98- case SQLITE_NOLFS:
99- printf("SQLITE_NOLFS");
100- break;
101- case SQLITE_NOMEM:
102- printf("SQLITE_NOMEM");
103- break;
104- case SQLITE_NOTADB:
105- printf("SQLITE_NOTADB");
106- break;
107- case SQLITE_NOTFOUND:
108- printf("SQLITE_NOTFOUND");
109- break;
110- case SQLITE_NOTICE:
111- printf("SQLITE_NOTICE");
112- break;
113- case SQLITE_OK:
114- printf("SQLITE_OK");
115- break;
116- case SQLITE_PERM:
117- printf("SQLITE_PERM");
118- break;
119- case SQLITE_SCHEMA:
120- printf("SQLITE_SCHEMA");
121- break;
122- case SQLITE_TOOBIG:
123- printf("SQLITE_TOOBIG");
124- break;
125- case SQLITE_WARNING:
126- printf("SQLITE_WARNING");
127- break;
128- }
129+ printf(sqlite3_errstr(code));
130 }
131diff --git a/ext/CMakeLists.txt b/ext/CMakeLists.txt
132new file mode 100644
133index 0000000000000000000000000000000000000000..d107d7f52608d512dd8a4cf800e55a6a0a166c32
134--- /dev/null
135+++ b/ext/CMakeLists.txt
136@@ -0,0 +1,4 @@
137+project(sqlite3ext VERSION 0.1 DESCRIPTION "sqlite3 ext library")
138+
139+file(GLOB src CONFIGURE_DEPENDS "*.c")
140+add_library(sqlite3ext SHARED ${src})
141diff --git a/ext/spellfix.c b/ext/spellfix.c
142new file mode 100644
143index 0000000000000000000000000000000000000000..a0c5aafd1009b6cfe79f5d46c98572f77a335706
144--- /dev/null
145+++ b/ext/spellfix.c
146@@ -0,0 +1,3076 @@
147+/*
148+** 2012 April 10
149+**
150+** The author disclaims copyright to this source code. In place of
151+** a legal notice, here is a blessing:
152+**
153+** May you do good and not evil.
154+** May you find forgiveness for yourself and forgive others.
155+** May you share freely, never taking more than you give.
156+**
157+*************************************************************************
158+**
159+** This module implements the spellfix1 VIRTUAL TABLE that can be used
160+** to search a large vocabulary for close matches. See separate
161+** documentation (http://www.sqlite.org/spellfix1.html) for details.
162+*/
163+#include "sqlite3ext.h"
164+SQLITE_EXTENSION_INIT1
165+
166+#ifndef SQLITE_AMALGAMATION
167+# if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
168+# define NDEBUG 1
169+# endif
170+# if defined(NDEBUG) && defined(SQLITE_DEBUG)
171+# undef NDEBUG
172+# endif
173+# include <string.h>
174+# include <stdio.h>
175+# include <stdlib.h>
176+# include <assert.h>
177+# define ALWAYS(X) 1
178+# define NEVER(X) 0
179+ typedef unsigned char u8;
180+ typedef unsigned short u16;
181+#endif
182+#include <ctype.h>
183+
184+#ifndef SQLITE_OMIT_VIRTUALTABLE
185+
186+/*
187+** Character classes for ASCII characters:
188+**
189+** 0 '' Silent letters: H W
190+** 1 'A' Any vowel: A E I O U (Y)
191+** 2 'B' A bilabeal stop or fricative: B F P V W
192+** 3 'C' Other fricatives or back stops: C G J K Q S X Z
193+** 4 'D' Alveolar stops: D T
194+** 5 'H' Letter H at the beginning of a word
195+** 6 'L' Glide: L
196+** 7 'R' Semivowel: R
197+** 8 'M' Nasals: M N
198+** 9 'Y' Letter Y at the beginning of a word.
199+** 10 '9' Digits: 0 1 2 3 4 5 6 7 8 9
200+** 11 ' ' White space
201+** 12 '?' Other.
202+*/
203+#define CCLASS_SILENT 0
204+#define CCLASS_VOWEL 1
205+#define CCLASS_B 2
206+#define CCLASS_C 3
207+#define CCLASS_D 4
208+#define CCLASS_H 5
209+#define CCLASS_L 6
210+#define CCLASS_R 7
211+#define CCLASS_M 8
212+#define CCLASS_Y 9
213+#define CCLASS_DIGIT 10
214+#define CCLASS_SPACE 11
215+#define CCLASS_OTHER 12
216+
217+/*
218+** The following table gives the character class for non-initial ASCII
219+** characters.
220+*/
221+static const unsigned char midClass[] = {
222+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
223+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
224+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
225+ /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
226+ /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER,
227+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
228+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
229+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
230+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
231+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
232+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE,
233+ /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER,
234+ /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER,
235+ /* ' */ CCLASS_SILENT, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER,
236+ /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER,
237+ /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER,
238+ /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT,
239+ /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT,
240+ /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT,
241+ /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER,
242+ /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER,
243+ /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL,
244+ /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D,
245+ /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C,
246+ /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C,
247+ /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M,
248+ /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B,
249+ /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C,
250+ /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B,
251+ /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_VOWEL,
252+ /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER,
253+ /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER,
254+ /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B,
255+ /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL,
256+ /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT,
257+ /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C,
258+ /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M,
259+ /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C,
260+ /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D,
261+ /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B,
262+ /* x */ CCLASS_C, /* y */ CCLASS_VOWEL, /* z */ CCLASS_C,
263+ /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER,
264+ /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER,
265+};
266+/*
267+** This tables gives the character class for ASCII characters that form the
268+** initial character of a word. The only difference from midClass is with
269+** the letters H, W, and Y.
270+*/
271+static const unsigned char initClass[] = {
272+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
273+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
274+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
275+ /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
276+ /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER,
277+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
278+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
279+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
280+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
281+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER,
282+ /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE,
283+ /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER,
284+ /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER,
285+ /* ' */ CCLASS_OTHER, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER,
286+ /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER,
287+ /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER,
288+ /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT,
289+ /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT,
290+ /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT,
291+ /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER,
292+ /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER,
293+ /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL,
294+ /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D,
295+ /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C,
296+ /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C,
297+ /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M,
298+ /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B,
299+ /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C,
300+ /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B,
301+ /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_Y,
302+ /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER,
303+ /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER,
304+ /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B,
305+ /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL,
306+ /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT,
307+ /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C,
308+ /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M,
309+ /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C,
310+ /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D,
311+ /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B,
312+ /* x */ CCLASS_C, /* y */ CCLASS_Y, /* z */ CCLASS_C,
313+ /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER,
314+ /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER,
315+};
316+
317+/*
318+** Mapping from the character class number (0-13) to a symbol for each
319+** character class. Note that initClass[] can be used to map the class
320+** symbol back into the class number.
321+*/
322+static const unsigned char className[] = ".ABCDHLRMY9 ?";
323+
324+/*
325+** Generate a "phonetic hash" from a string of ASCII characters
326+** in zIn[0..nIn-1].
327+**
328+** * Map characters by character class as defined above.
329+** * Omit double-letters
330+** * Omit vowels beside R and L
331+** * Omit T when followed by CH
332+** * Omit W when followed by R
333+** * Omit D when followed by J or G
334+** * Omit K in KN or G in GN at the beginning of a word
335+**
336+** Space to hold the result is obtained from sqlite3_malloc()
337+**
338+** Return NULL if memory allocation fails.
339+*/
340+static unsigned char *phoneticHash(const unsigned char *zIn, int nIn){
341+ unsigned char *zOut = sqlite3_malloc64( nIn + 1 );
342+ int i;
343+ int nOut = 0;
344+ char cPrev = 0x77;
345+ char cPrevX = 0x77;
346+ const unsigned char *aClass = initClass;
347+
348+ if( zOut==0 ) return 0;
349+ if( nIn>2 ){
350+ switch( zIn[0] ){
351+ case 'g':
352+ case 'k': {
353+ if( zIn[1]=='n' ){ zIn++; nIn--; }
354+ break;
355+ }
356+ }
357+ }
358+ for(i=0; i<nIn; i++){
359+ unsigned char c = zIn[i];
360+ if( i+1<nIn ){
361+ if( c=='w' && zIn[i+1]=='r' ) continue;
362+ if( c=='d' && (zIn[i+1]=='j' || zIn[i+1]=='g') ) continue;
363+ if( i+2<nIn ){
364+ if( c=='t' && zIn[i+1]=='c' && zIn[i+2]=='h' ) continue;
365+ }
366+ }
367+ c = aClass[c&0x7f];
368+ if( c==CCLASS_SPACE ) continue;
369+ if( c==CCLASS_OTHER && cPrev!=CCLASS_DIGIT ) continue;
370+ aClass = midClass;
371+ if( c==CCLASS_VOWEL && (cPrevX==CCLASS_R || cPrevX==CCLASS_L) ){
372+ continue; /* No vowels beside L or R */
373+ }
374+ if( (c==CCLASS_R || c==CCLASS_L) && cPrevX==CCLASS_VOWEL ){
375+ nOut--; /* No vowels beside L or R */
376+ }
377+ cPrev = c;
378+ if( c==CCLASS_SILENT ) continue;
379+ cPrevX = c;
380+ c = className[c];
381+ assert( nOut>=0 );
382+ if( nOut==0 || c!=zOut[nOut-1] ) zOut[nOut++] = c;
383+ }
384+ zOut[nOut] = 0;
385+ return zOut;
386+}
387+
388+/*
389+** This is an SQL function wrapper around phoneticHash(). See
390+** the description of phoneticHash() for additional information.
391+*/
392+static void phoneticHashSqlFunc(
393+ sqlite3_context *context,
394+ int argc,
395+ sqlite3_value **argv
396+){
397+ const unsigned char *zIn;
398+ unsigned char *zOut;
399+
400+ zIn = sqlite3_value_text(argv[0]);
401+ if( zIn==0 ) return;
402+ zOut = phoneticHash(zIn, sqlite3_value_bytes(argv[0]));
403+ if( zOut==0 ){
404+ sqlite3_result_error_nomem(context);
405+ }else{
406+ sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free);
407+ }
408+}
409+
410+/*
411+** Return the character class number for a character given its
412+** context.
413+*/
414+static char characterClass(char cPrev, char c){
415+ return cPrev==0 ? initClass[c&0x7f] : midClass[c&0x7f];
416+}
417+
418+/*
419+** Return the cost of inserting or deleting character c immediately
420+** following character cPrev. If cPrev==0, that means c is the first
421+** character of the word.
422+*/
423+static int insertOrDeleteCost(char cPrev, char c, char cNext){
424+ char classC = characterClass(cPrev, c);
425+ char classCprev;
426+
427+ if( classC==CCLASS_SILENT ){
428+ /* Insert or delete "silent" characters such as H or W */
429+ return 1;
430+ }
431+ if( cPrev==c ){
432+ /* Repeated characters, or miss a repeat */
433+ return 10;
434+ }
435+ if( classC==CCLASS_VOWEL && (cPrev=='r' || cNext=='r') ){
436+ return 20; /* Insert a vowel before or after 'r' */
437+ }
438+ classCprev = characterClass(cPrev, cPrev);
439+ if( classC==classCprev ){
440+ if( classC==CCLASS_VOWEL ){
441+ /* Remove or add a new vowel to a vowel cluster */
442+ return 15;
443+ }else{
444+ /* Remove or add a consonant not in the same class */
445+ return 50;
446+ }
447+ }
448+
449+ /* any other character insertion or deletion */
450+ return 100;
451+}
452+
453+/*
454+** Divide the insertion cost by this factor when appending to the
455+** end of the word.
456+*/
457+#define FINAL_INS_COST_DIV 4
458+
459+/*
460+** Return the cost of substituting cTo in place of cFrom assuming
461+** the previous character is cPrev. If cPrev==0 then cTo is the first
462+** character of the word.
463+*/
464+static int substituteCost(char cPrev, char cFrom, char cTo){
465+ char classFrom, classTo;
466+ if( cFrom==cTo ){
467+ /* Exact match */
468+ return 0;
469+ }
470+ if( cFrom==(cTo^0x20) && ((cTo>='A' && cTo<='Z') || (cTo>='a' && cTo<='z')) ){
471+ /* differ only in case */
472+ return 0;
473+ }
474+ classFrom = characterClass(cPrev, cFrom);
475+ classTo = characterClass(cPrev, cTo);
476+ if( classFrom==classTo ){
477+ /* Same character class */
478+ return 40;
479+ }
480+ if( classFrom>=CCLASS_B && classFrom<=CCLASS_Y
481+ && classTo>=CCLASS_B && classTo<=CCLASS_Y ){
482+ /* Convert from one consonant to another, but in a different class */
483+ return 75;
484+ }
485+ /* Any other subsitution */
486+ return 100;
487+}
488+
489+/*
490+** Given two strings zA and zB which are pure ASCII, return the cost
491+** of transforming zA into zB. If zA ends with '*' assume that it is
492+** a prefix of zB and give only minimal penalty for extra characters
493+** on the end of zB.
494+**
495+** Smaller numbers mean a closer match.
496+**
497+** Negative values indicate an error:
498+** -1 One of the inputs is NULL
499+** -2 Non-ASCII characters on input
500+** -3 Unable to allocate memory
501+**
502+** If pnMatch is not NULL, then *pnMatch is set to the number of bytes
503+** of zB that matched the pattern in zA. If zA does not end with a '*',
504+** then this value is always the number of bytes in zB (i.e. strlen(zB)).
505+** If zA does end in a '*', then it is the number of bytes in the prefix
506+** of zB that was deemed to match zA.
507+*/
508+static int editdist1(const char *zA, const char *zB, int *pnMatch){
509+ int nA, nB; /* Number of characters in zA[] and zB[] */
510+ int xA, xB; /* Loop counters for zA[] and zB[] */
511+ char cA = 0, cB; /* Current character of zA and zB */
512+ char cAprev, cBprev; /* Previous character of zA and zB */
513+ char cAnext, cBnext; /* Next character in zA and zB */
514+ int d; /* North-west cost value */
515+ int dc = 0; /* North-west character value */
516+ int res; /* Final result */
517+ int *m; /* The cost matrix */
518+ char *cx; /* Corresponding character values */
519+ int *toFree = 0; /* Malloced space */
520+ int nMatch = 0;
521+ int mStack[60+15]; /* Stack space to use if not too much is needed */
522+
523+ /* Early out if either input is NULL */
524+ if( zA==0 || zB==0 ) return -1;
525+
526+ /* Skip any common prefix */
527+ while( zA[0] && zA[0]==zB[0] ){ dc = zA[0]; zA++; zB++; nMatch++; }
528+ if( pnMatch ) *pnMatch = nMatch;
529+ if( zA[0]==0 && zB[0]==0 ) return 0;
530+
531+#if 0
532+ printf("A=\"%s\" B=\"%s\" dc=%c\n", zA, zB, dc?dc:' ');
533+#endif
534+
535+ /* Verify input strings and measure their lengths */
536+ for(nA=0; zA[nA]; nA++){
537+ if( zA[nA]&0x80 ) return -2;
538+ }
539+ for(nB=0; zB[nB]; nB++){
540+ if( zB[nB]&0x80 ) return -2;
541+ }
542+
543+ /* Special processing if either string is empty */
544+ if( nA==0 ){
545+ cBprev = (char)dc;
546+ for(xB=res=0; (cB = zB[xB])!=0; xB++){
547+ res += insertOrDeleteCost(cBprev, cB, zB[xB+1])/FINAL_INS_COST_DIV;
548+ cBprev = cB;
549+ }
550+ return res;
551+ }
552+ if( nB==0 ){
553+ cAprev = (char)dc;
554+ for(xA=res=0; (cA = zA[xA])!=0; xA++){
555+ res += insertOrDeleteCost(cAprev, cA, zA[xA+1]);
556+ cAprev = cA;
557+ }
558+ return res;
559+ }
560+
561+ /* A is a prefix of B */
562+ if( zA[0]=='*' && zA[1]==0 ) return 0;
563+
564+ /* Allocate and initialize the Wagner matrix */
565+ if( nB<(sizeof(mStack)*4)/(sizeof(mStack[0])*5) ){
566+ m = mStack;
567+ }else{
568+ m = toFree = sqlite3_malloc64( (nB+1)*5*sizeof(m[0])/4 );
569+ if( m==0 ) return -3;
570+ }
571+ cx = (char*)&m[nB+1];
572+
573+ /* Compute the Wagner edit distance */
574+ m[0] = 0;
575+ cx[0] = (char)dc;
576+ cBprev = (char)dc;
577+ for(xB=1; xB<=nB; xB++){
578+ cBnext = zB[xB];
579+ cB = zB[xB-1];
580+ cx[xB] = cB;
581+ m[xB] = m[xB-1] + insertOrDeleteCost(cBprev, cB, cBnext);
582+ cBprev = cB;
583+ }
584+ cAprev = (char)dc;
585+ for(xA=1; xA<=nA; xA++){
586+ int lastA = (xA==nA);
587+ cA = zA[xA-1];
588+ cAnext = zA[xA];
589+ if( cA=='*' && lastA ) break;
590+ d = m[0];
591+ dc = cx[0];
592+ m[0] = d + insertOrDeleteCost(cAprev, cA, cAnext);
593+ cBprev = 0;
594+ for(xB=1; xB<=nB; xB++){
595+ int totalCost, insCost, delCost, subCost, ncx;
596+ cB = zB[xB-1];
597+ cBnext = zB[xB];
598+
599+ /* Cost to insert cB */
600+ insCost = insertOrDeleteCost(cx[xB-1], cB, cBnext);
601+ if( lastA ) insCost /= FINAL_INS_COST_DIV;
602+
603+ /* Cost to delete cA */
604+ delCost = insertOrDeleteCost(cx[xB], cA, cBnext);
605+
606+ /* Cost to substitute cA->cB */
607+ subCost = substituteCost(cx[xB-1], cA, cB);
608+
609+ /* Best cost */
610+ totalCost = insCost + m[xB-1];
611+ ncx = cB;
612+ if( (delCost + m[xB])<totalCost ){
613+ totalCost = delCost + m[xB];
614+ ncx = cA;
615+ }
616+ if( (subCost + d)<totalCost ){
617+ totalCost = subCost + d;
618+ }
619+
620+#if 0
621+ printf("%d,%d d=%4d u=%4d r=%4d dc=%c cA=%c cB=%c"
622+ " ins=%4d del=%4d sub=%4d t=%4d ncx=%c\n",
623+ xA, xB, d, m[xB], m[xB-1], dc?dc:' ', cA, cB,
624+ insCost, delCost, subCost, totalCost, ncx?ncx:' ');
625+#endif
626+
627+ /* Update the matrix */
628+ d = m[xB];
629+ dc = cx[xB];
630+ m[xB] = totalCost;
631+ cx[xB] = (char)ncx;
632+ cBprev = cB;
633+ }
634+ cAprev = cA;
635+ }
636+
637+ /* Free the wagner matrix and return the result */
638+ if( cA=='*' ){
639+ res = m[1];
640+ for(xB=1; xB<=nB; xB++){
641+ if( m[xB]<res ){
642+ res = m[xB];
643+ if( pnMatch ) *pnMatch = xB+nMatch;
644+ }
645+ }
646+ }else{
647+ res = m[nB];
648+ /* In the current implementation, pnMatch is always NULL if zA does
649+ ** not end in "*" */
650+ assert( pnMatch==0 );
651+ }
652+ sqlite3_free(toFree);
653+ return res;
654+}
655+
656+/*
657+** Function: editdist(A,B)
658+**
659+** Return the cost of transforming string A into string B. Both strings
660+** must be pure ASCII text. If A ends with '*' then it is assumed to be
661+** a prefix of B and extra characters on the end of B have minimal additional
662+** cost.
663+*/
664+static void editdistSqlFunc(
665+ sqlite3_context *context,
666+ int argc,
667+ sqlite3_value **argv
668+){
669+ int res = editdist1(
670+ (const char*)sqlite3_value_text(argv[0]),
671+ (const char*)sqlite3_value_text(argv[1]),
672+ 0);
673+ if( res<0 ){
674+ if( res==(-3) ){
675+ sqlite3_result_error_nomem(context);
676+ }else if( res==(-2) ){
677+ sqlite3_result_error(context, "non-ASCII input to editdist()", -1);
678+ }else{
679+ sqlite3_result_error(context, "NULL input to editdist()", -1);
680+ }
681+ }else{
682+ sqlite3_result_int(context, res);
683+ }
684+}
685+
686+/* End of the fixed-cost edit distance implementation
687+******************************************************************************
688+*****************************************************************************
689+** Begin: Configurable cost unicode edit distance routines
690+*/
691+/* Forward declaration of structures */
692+typedef struct EditDist3Cost EditDist3Cost;
693+typedef struct EditDist3Config EditDist3Config;
694+typedef struct EditDist3Point EditDist3Point;
695+typedef struct EditDist3From EditDist3From;
696+typedef struct EditDist3FromString EditDist3FromString;
697+typedef struct EditDist3To EditDist3To;
698+typedef struct EditDist3ToString EditDist3ToString;
699+typedef struct EditDist3Lang EditDist3Lang;
700+
701+
702+/*
703+** An entry in the edit cost table
704+*/
705+struct EditDist3Cost {
706+ EditDist3Cost *pNext; /* Next cost element */
707+ u8 nFrom; /* Number of bytes in aFrom */
708+ u8 nTo; /* Number of bytes in aTo */
709+ u16 iCost; /* Cost of this transformation */
710+ char a[4] ; /* FROM string followed by TO string */
711+ /* Additional TO and FROM string bytes appended as necessary */
712+};
713+
714+/*
715+** Edit costs for a particular language ID
716+*/
717+struct EditDist3Lang {
718+ int iLang; /* Language ID */
719+ int iInsCost; /* Default insertion cost */
720+ int iDelCost; /* Default deletion cost */
721+ int iSubCost; /* Default substitution cost */
722+ EditDist3Cost *pCost; /* Costs */
723+};
724+
725+
726+/*
727+** The default EditDist3Lang object, with default costs.
728+*/
729+static const EditDist3Lang editDist3Lang = { 0, 100, 100, 150, 0 };
730+
731+/*
732+** Complete configuration
733+*/
734+struct EditDist3Config {
735+ int nLang; /* Number of language IDs. Size of a[] */
736+ EditDist3Lang *a; /* One for each distinct language ID */
737+};
738+
739+/*
740+** Extra information about each character in the FROM string.
741+*/
742+struct EditDist3From {
743+ int nSubst; /* Number of substitution cost entries */
744+ int nDel; /* Number of deletion cost entries */
745+ int nByte; /* Number of bytes in this character */
746+ EditDist3Cost **apSubst; /* Array of substitution costs for this element */
747+ EditDist3Cost **apDel; /* Array of deletion cost entries */
748+};
749+
750+/*
751+** A precompiled FROM string.
752+*
753+** In the common case we expect the FROM string to be reused multiple times.
754+** In other words, the common case will be to measure the edit distance
755+** from a single origin string to multiple target strings.
756+*/
757+struct EditDist3FromString {
758+ char *z; /* The complete text of the FROM string */
759+ int n; /* Number of characters in the FROM string */
760+ int isPrefix; /* True if ends with '*' character */
761+ EditDist3From *a; /* Extra info about each char of the FROM string */
762+};
763+
764+/*
765+** Extra information about each character in the TO string.
766+*/
767+struct EditDist3To {
768+ int nIns; /* Number of insertion cost entries */
769+ int nByte; /* Number of bytes in this character */
770+ EditDist3Cost **apIns; /* Array of deletion cost entries */
771+};
772+
773+/*
774+** A precompiled FROM string
775+*/
776+struct EditDist3ToString {
777+ char *z; /* The complete text of the TO string */
778+ int n; /* Number of characters in the TO string */
779+ EditDist3To *a; /* Extra info about each char of the TO string */
780+};
781+
782+/*
783+** Clear or delete an instance of the object that records all edit-distance
784+** weights.
785+*/
786+static void editDist3ConfigClear(EditDist3Config *p){
787+ int i;
788+ if( p==0 ) return;
789+ for(i=0; i<p->nLang; i++){
790+ EditDist3Cost *pCost, *pNext;
791+ pCost = p->a[i].pCost;
792+ while( pCost ){
793+ pNext = pCost->pNext;
794+ sqlite3_free(pCost);
795+ pCost = pNext;
796+ }
797+ }
798+ sqlite3_free(p->a);
799+ memset(p, 0, sizeof(*p));
800+}
801+static void editDist3ConfigDelete(void *pIn){
802+ EditDist3Config *p = (EditDist3Config*)pIn;
803+ editDist3ConfigClear(p);
804+ sqlite3_free(p);
805+}
806+
807+/* Compare the FROM values of two EditDist3Cost objects, for sorting.
808+** Return negative, zero, or positive if the A is less than, equal to,
809+** or greater than B.
810+*/
811+static int editDist3CostCompare(EditDist3Cost *pA, EditDist3Cost *pB){
812+ int n = pA->nFrom;
813+ int rc;
814+ if( n>pB->nFrom ) n = pB->nFrom;
815+ rc = strncmp(pA->a, pB->a, n);
816+ if( rc==0 ) rc = pA->nFrom - pB->nFrom;
817+ return rc;
818+}
819+
820+/*
821+** Merge together two sorted lists of EditDist3Cost objects, in order
822+** of increasing FROM.
823+*/
824+static EditDist3Cost *editDist3CostMerge(
825+ EditDist3Cost *pA,
826+ EditDist3Cost *pB
827+){
828+ EditDist3Cost *pHead = 0;
829+ EditDist3Cost **ppTail = &pHead;
830+ EditDist3Cost *p;
831+ while( pA && pB ){
832+ if( editDist3CostCompare(pA,pB)<=0 ){
833+ p = pA;
834+ pA = pA->pNext;
835+ }else{
836+ p = pB;
837+ pB = pB->pNext;
838+ }
839+ *ppTail = p;
840+ ppTail = &p->pNext;
841+ }
842+ if( pA ){
843+ *ppTail = pA;
844+ }else{
845+ *ppTail = pB;
846+ }
847+ return pHead;
848+}
849+
850+/*
851+** Sort a list of EditDist3Cost objects into order of increasing FROM
852+*/
853+static EditDist3Cost *editDist3CostSort(EditDist3Cost *pList){
854+ EditDist3Cost *ap[60], *p;
855+ int i;
856+ int mx = 0;
857+ ap[0] = 0;
858+ ap[1] = 0;
859+ while( pList ){
860+ p = pList;
861+ pList = p->pNext;
862+ p->pNext = 0;
863+ for(i=0; ap[i]; i++){
864+ p = editDist3CostMerge(ap[i],p);
865+ ap[i] = 0;
866+ }
867+ ap[i] = p;
868+ if( i>mx ){
869+ mx = i;
870+ ap[i+1] = 0;
871+ }
872+ }
873+ p = 0;
874+ for(i=0; i<=mx; i++){
875+ if( ap[i] ) p = editDist3CostMerge(p,ap[i]);
876+ }
877+ return p;
878+}
879+
880+/*
881+** Load all edit-distance weights from a table.
882+*/
883+static int editDist3ConfigLoad(
884+ EditDist3Config *p, /* The edit distance configuration to load */
885+ sqlite3 *db, /* Load from this database */
886+ const char *zTable /* Name of the table from which to load */
887+){
888+ sqlite3_stmt *pStmt;
889+ int rc, rc2;
890+ char *zSql;
891+ int iLangPrev = -9999;
892+ EditDist3Lang *pLang = 0;
893+
894+ zSql = sqlite3_mprintf("SELECT iLang, cFrom, cTo, iCost"
895+ " FROM \"%w\" WHERE iLang>=0 ORDER BY iLang", zTable);
896+ if( zSql==0 ) return SQLITE_NOMEM;
897+ rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0);
898+ sqlite3_free(zSql);
899+ if( rc ) return rc;
900+ editDist3ConfigClear(p);
901+ while( sqlite3_step(pStmt)==SQLITE_ROW ){
902+ int iLang = sqlite3_column_int(pStmt, 0);
903+ const char *zFrom = (const char*)sqlite3_column_text(pStmt, 1);
904+ int nFrom = zFrom ? sqlite3_column_bytes(pStmt, 1) : 0;
905+ const char *zTo = (const char*)sqlite3_column_text(pStmt, 2);
906+ int nTo = zTo ? sqlite3_column_bytes(pStmt, 2) : 0;
907+ int iCost = sqlite3_column_int(pStmt, 3);
908+
909+ assert( zFrom!=0 || nFrom==0 );
910+ assert( zTo!=0 || nTo==0 );
911+ if( nFrom>100 || nTo>100 ) continue;
912+ if( iCost<0 ) continue;
913+ if( iCost>=10000 ) continue; /* Costs above 10K are considered infinite */
914+ if( pLang==0 || iLang!=iLangPrev ){
915+ EditDist3Lang *pNew;
916+ pNew = sqlite3_realloc64(p->a, (p->nLang+1)*sizeof(p->a[0]));
917+ if( pNew==0 ){ rc = SQLITE_NOMEM; break; }
918+ p->a = pNew;
919+ pLang = &p->a[p->nLang];
920+ p->nLang++;
921+ pLang->iLang = iLang;
922+ pLang->iInsCost = 100;
923+ pLang->iDelCost = 100;
924+ pLang->iSubCost = 150;
925+ pLang->pCost = 0;
926+ iLangPrev = iLang;
927+ }
928+ if( nFrom==1 && zFrom[0]=='?' && nTo==0 ){
929+ pLang->iDelCost = iCost;
930+ }else if( nFrom==0 && nTo==1 && zTo[0]=='?' ){
931+ pLang->iInsCost = iCost;
932+ }else if( nFrom==1 && nTo==1 && zFrom[0]=='?' && zTo[0]=='?' ){
933+ pLang->iSubCost = iCost;
934+ }else{
935+ EditDist3Cost *pCost;
936+ int nExtra = nFrom + nTo - 4;
937+ if( nExtra<0 ) nExtra = 0;
938+ pCost = sqlite3_malloc64( sizeof(*pCost) + nExtra );
939+ if( pCost==0 ){ rc = SQLITE_NOMEM; break; }
940+ pCost->nFrom = (u8)nFrom;
941+ pCost->nTo = (u8)nTo;
942+ pCost->iCost = (u16)iCost;
943+ memcpy(pCost->a, zFrom, nFrom);
944+ memcpy(pCost->a + nFrom, zTo, nTo);
945+ pCost->pNext = pLang->pCost;
946+ pLang->pCost = pCost;
947+ }
948+ }
949+ rc2 = sqlite3_finalize(pStmt);
950+ if( rc==SQLITE_OK ) rc = rc2;
951+ if( rc==SQLITE_OK ){
952+ int iLang;
953+ for(iLang=0; iLang<p->nLang; iLang++){
954+ p->a[iLang].pCost = editDist3CostSort(p->a[iLang].pCost);
955+ }
956+ }
957+ return rc;
958+}
959+
960+/*
961+** Return the length (in bytes) of a utf-8 character. Or return a maximum
962+** of N.
963+*/
964+static int utf8Len(unsigned char c, int N){
965+ int len = 1;
966+ if( c>0x7f ){
967+ if( (c&0xe0)==0xc0 ){
968+ len = 2;
969+ }else if( (c&0xf0)==0xe0 ){
970+ len = 3;
971+ }else{
972+ len = 4;
973+ }
974+ }
975+ if( len>N ) len = N;
976+ return len;
977+}
978+
979+/*
980+** Return TRUE (non-zero) if the To side of the given cost matches
981+** the given string.
982+*/
983+static int matchTo(EditDist3Cost *p, const char *z, int n){
984+ assert( n>0 );
985+ if( p->a[p->nFrom]!=z[0] ) return 0;
986+ if( p->nTo>n ) return 0;
987+ if( strncmp(p->a+p->nFrom, z, p->nTo)!=0 ) return 0;
988+ return 1;
989+}
990+
991+/*
992+** Return TRUE (non-zero) if the From side of the given cost matches
993+** the given string.
994+*/
995+static int matchFrom(EditDist3Cost *p, const char *z, int n){
996+ assert( p->nFrom<=n );
997+ if( p->nFrom ){
998+ if( p->a[0]!=z[0] ) return 0;
999+ if( strncmp(p->a, z, p->nFrom)!=0 ) return 0;
1000+ }
1001+ return 1;
1002+}
1003+
1004+/*
1005+** Return TRUE (non-zero) of the next FROM character and the next TO
1006+** character are the same.
1007+*/
1008+static int matchFromTo(
1009+ EditDist3FromString *pStr, /* Left hand string */
1010+ int n1, /* Index of comparison character on the left */
1011+ const char *z2, /* Right-handl comparison character */
1012+ int n2 /* Bytes remaining in z2[] */
1013+){
1014+ int b1 = pStr->a[n1].nByte;
1015+ if( b1>n2 ) return 0;
1016+ assert( b1>0 );
1017+ if( pStr->z[n1]!=z2[0] ) return 0;
1018+ if( strncmp(pStr->z+n1, z2, b1)!=0 ) return 0;
1019+ return 1;
1020+}
1021+
1022+/*
1023+** Delete an EditDist3FromString objecct
1024+*/
1025+static void editDist3FromStringDelete(EditDist3FromString *p){
1026+ int i;
1027+ if( p ){
1028+ for(i=0; i<p->n; i++){
1029+ sqlite3_free(p->a[i].apDel);
1030+ sqlite3_free(p->a[i].apSubst);
1031+ }
1032+ sqlite3_free(p);
1033+ }
1034+}
1035+
1036+/*
1037+** Create a EditDist3FromString object.
1038+*/
1039+static EditDist3FromString *editDist3FromStringNew(
1040+ const EditDist3Lang *pLang,
1041+ const char *z,
1042+ int n
1043+){
1044+ EditDist3FromString *pStr;
1045+ EditDist3Cost *p;
1046+ int i;
1047+
1048+ if( z==0 ) return 0;
1049+ if( n<0 ) n = (int)strlen(z);
1050+ pStr = sqlite3_malloc64( sizeof(*pStr) + sizeof(pStr->a[0])*n + n + 1 );
1051+ if( pStr==0 ) return 0;
1052+ pStr->a = (EditDist3From*)&pStr[1];
1053+ memset(pStr->a, 0, sizeof(pStr->a[0])*n);
1054+ pStr->n = n;
1055+ pStr->z = (char*)&pStr->a[n];
1056+ memcpy(pStr->z, z, n+1);
1057+ if( n && z[n-1]=='*' ){
1058+ pStr->isPrefix = 1;
1059+ n--;
1060+ pStr->n--;
1061+ pStr->z[n] = 0;
1062+ }else{
1063+ pStr->isPrefix = 0;
1064+ }
1065+
1066+ for(i=0; i<n; i++){
1067+ EditDist3From *pFrom = &pStr->a[i];
1068+ memset(pFrom, 0, sizeof(*pFrom));
1069+ pFrom->nByte = utf8Len((unsigned char)z[i], n-i);
1070+ for(p=pLang->pCost; p; p=p->pNext){
1071+ EditDist3Cost **apNew;
1072+ if( i+p->nFrom>n ) continue;
1073+ if( matchFrom(p, z+i, n-i)==0 ) continue;
1074+ if( p->nTo==0 ){
1075+ apNew = sqlite3_realloc64(pFrom->apDel,
1076+ sizeof(*apNew)*(pFrom->nDel+1));
1077+ if( apNew==0 ) break;
1078+ pFrom->apDel = apNew;
1079+ apNew[pFrom->nDel++] = p;
1080+ }else{
1081+ apNew = sqlite3_realloc64(pFrom->apSubst,
1082+ sizeof(*apNew)*(pFrom->nSubst+1));
1083+ if( apNew==0 ) break;
1084+ pFrom->apSubst = apNew;
1085+ apNew[pFrom->nSubst++] = p;
1086+ }
1087+ }
1088+ if( p ){
1089+ editDist3FromStringDelete(pStr);
1090+ pStr = 0;
1091+ break;
1092+ }
1093+ }
1094+ return pStr;
1095+}
1096+
1097+/*
1098+** Update entry m[i] such that it is the minimum of its current value
1099+** and m[j]+iCost.
1100+*/
1101+static void updateCost(
1102+ unsigned int *m,
1103+ int i,
1104+ int j,
1105+ int iCost
1106+){
1107+ unsigned int b;
1108+ assert( iCost>=0 );
1109+ assert( iCost<10000 );
1110+ b = m[j] + iCost;
1111+ if( b<m[i] ) m[i] = b;
1112+}
1113+
1114+/*
1115+** How much stack space (int bytes) to use for Wagner matrix in
1116+** editDist3Core(). If more space than this is required, the entire
1117+** matrix is taken from the heap. To reduce the load on the memory
1118+** allocator, make this value as large as practical for the
1119+** architecture in use.
1120+*/
1121+#ifndef SQLITE_SPELLFIX_STACKALLOC_SZ
1122+# define SQLITE_SPELLFIX_STACKALLOC_SZ (1024)
1123+#endif
1124+
1125+/* Compute the edit distance between two strings.
1126+**
1127+** If an error occurs, return a negative number which is the error code.
1128+**
1129+** If pnMatch is not NULL, then *pnMatch is set to the number of characters
1130+** (not bytes) in z2 that matched the search pattern in *pFrom. If pFrom does
1131+** not contain the pattern for a prefix-search, then this is always the number
1132+** of characters in z2. If pFrom does contain a prefix search pattern, then
1133+** it is the number of characters in the prefix of z2 that was deemed to
1134+** match pFrom.
1135+*/
1136+static int editDist3Core(
1137+ EditDist3FromString *pFrom, /* The FROM string */
1138+ const char *z2, /* The TO string */
1139+ int n2, /* Length of the TO string */
1140+ const EditDist3Lang *pLang, /* Edit weights for a particular language ID */
1141+ int *pnMatch /* OUT: Characters in matched prefix */
1142+){
1143+ int k, n;
1144+ int i1, b1;
1145+ int i2, b2;
1146+ EditDist3FromString f = *pFrom;
1147+ EditDist3To *a2;
1148+ unsigned int *m;
1149+ unsigned int *pToFree;
1150+ int szRow;
1151+ EditDist3Cost *p;
1152+ int res;
1153+ sqlite3_uint64 nByte;
1154+ unsigned int stackSpace[SQLITE_SPELLFIX_STACKALLOC_SZ/sizeof(unsigned int)];
1155+
1156+ /* allocate the Wagner matrix and the aTo[] array for the TO string */
1157+ n = (f.n+1)*(n2+1);
1158+ n = (n+1)&~1;
1159+ nByte = n*sizeof(m[0]) + sizeof(a2[0])*n2;
1160+ if( nByte<=sizeof(stackSpace) ){
1161+ m = stackSpace;
1162+ pToFree = 0;
1163+ }else{
1164+ m = pToFree = sqlite3_malloc64( nByte );
1165+ if( m==0 ) return -1; /* Out of memory */
1166+ }
1167+ a2 = (EditDist3To*)&m[n];
1168+ memset(a2, 0, sizeof(a2[0])*n2);
1169+
1170+ /* Fill in the a1[] matrix for all characters of the TO string */
1171+ for(i2=0; i2<n2; i2++){
1172+ a2[i2].nByte = utf8Len((unsigned char)z2[i2], n2-i2);
1173+ for(p=pLang->pCost; p; p=p->pNext){
1174+ EditDist3Cost **apNew;
1175+ if( p->nFrom>0 ) break;
1176+ if( i2+p->nTo>n2 ) continue;
1177+ if( p->a[0]>z2[i2] ) break;
1178+ if( matchTo(p, z2+i2, n2-i2)==0 ) continue;
1179+ a2[i2].nIns++;
1180+ apNew = sqlite3_realloc64(a2[i2].apIns, sizeof(*apNew)*a2[i2].nIns);
1181+ if( apNew==0 ){
1182+ res = -1; /* Out of memory */
1183+ goto editDist3Abort;
1184+ }
1185+ a2[i2].apIns = apNew;
1186+ a2[i2].apIns[a2[i2].nIns-1] = p;
1187+ }
1188+ }
1189+
1190+ /* Prepare to compute the minimum edit distance */
1191+ szRow = f.n+1;
1192+ memset(m, 0x01, (n2+1)*szRow*sizeof(m[0]));
1193+ m[0] = 0;
1194+
1195+ /* First fill in the top-row of the matrix with FROM deletion costs */
1196+ for(i1=0; i1<f.n; i1 += b1){
1197+ b1 = f.a[i1].nByte;
1198+ updateCost(m, i1+b1, i1, pLang->iDelCost);
1199+ for(k=0; k<f.a[i1].nDel; k++){
1200+ p = f.a[i1].apDel[k];
1201+ updateCost(m, i1+p->nFrom, i1, p->iCost);
1202+ }
1203+ }
1204+
1205+ /* Fill in all subsequent rows, top-to-bottom, left-to-right */
1206+ for(i2=0; i2<n2; i2 += b2){
1207+ int rx; /* Starting index for current row */
1208+ int rxp; /* Starting index for previous row */
1209+ b2 = a2[i2].nByte;
1210+ rx = szRow*(i2+b2);
1211+ rxp = szRow*i2;
1212+ updateCost(m, rx, rxp, pLang->iInsCost);
1213+ for(k=0; k<a2[i2].nIns; k++){
1214+ p = a2[i2].apIns[k];
1215+ updateCost(m, szRow*(i2+p->nTo), rxp, p->iCost);
1216+ }
1217+ for(i1=0; i1<f.n; i1+=b1){
1218+ int cx; /* Index of current cell */
1219+ int cxp; /* Index of cell immediately to the left */
1220+ int cxd; /* Index of cell to the left and one row above */
1221+ int cxu; /* Index of cell immediately above */
1222+ b1 = f.a[i1].nByte;
1223+ cxp = rx + i1;
1224+ cx = cxp + b1;
1225+ cxd = rxp + i1;
1226+ cxu = cxd + b1;
1227+ updateCost(m, cx, cxp, pLang->iDelCost);
1228+ for(k=0; k<f.a[i1].nDel; k++){
1229+ p = f.a[i1].apDel[k];
1230+ updateCost(m, cxp+p->nFrom, cxp, p->iCost);
1231+ }
1232+ updateCost(m, cx, cxu, pLang->iInsCost);
1233+ if( matchFromTo(&f, i1, z2+i2, n2-i2) ){
1234+ updateCost(m, cx, cxd, 0);
1235+ }
1236+ updateCost(m, cx, cxd, pLang->iSubCost);
1237+ for(k=0; k<f.a[i1].nSubst; k++){
1238+ p = f.a[i1].apSubst[k];
1239+ if( matchTo(p, z2+i2, n2-i2) ){
1240+ updateCost(m, cxd+p->nFrom+szRow*p->nTo, cxd, p->iCost);
1241+ }
1242+ }
1243+ }
1244+ }
1245+
1246+#if 0 /* Enable for debugging */
1247+ printf(" ^");
1248+ for(i1=0; i1<f.n; i1++) printf(" %c-%2x", f.z[i1], f.z[i1]&0xff);
1249+ printf("\n ^:");
1250+ for(i1=0; i1<szRow; i1++){
1251+ int v = m[i1];
1252+ if( v>9999 ) printf(" ****");
1253+ else printf(" %4d", v);
1254+ }
1255+ printf("\n");
1256+ for(i2=0; i2<n2; i2++){
1257+ printf("%c-%02x:", z2[i2], z2[i2]&0xff);
1258+ for(i1=0; i1<szRow; i1++){
1259+ int v = m[(i2+1)*szRow+i1];
1260+ if( v>9999 ) printf(" ****");
1261+ else printf(" %4d", v);
1262+ }
1263+ printf("\n");
1264+ }
1265+#endif
1266+
1267+ /* Free memory allocations and return the result */
1268+ res = (int)m[szRow*(n2+1)-1];
1269+ n = n2;
1270+ if( f.isPrefix ){
1271+ for(i2=1; i2<=n2; i2++){
1272+ int b = m[szRow*i2-1];
1273+ if( b<=res ){
1274+ res = b;
1275+ n = i2 - 1;
1276+ }
1277+ }
1278+ }
1279+ if( pnMatch ){
1280+ int nExtra = 0;
1281+ for(k=0; k<n; k++){
1282+ if( (z2[k] & 0xc0)==0x80 ) nExtra++;
1283+ }
1284+ *pnMatch = n - nExtra;
1285+ }
1286+
1287+editDist3Abort:
1288+ for(i2=0; i2<n2; i2++) sqlite3_free(a2[i2].apIns);
1289+ sqlite3_free(pToFree);
1290+ return res;
1291+}
1292+
1293+/*
1294+** Get an appropriate EditDist3Lang object.
1295+*/
1296+static const EditDist3Lang *editDist3FindLang(
1297+ EditDist3Config *pConfig,
1298+ int iLang
1299+){
1300+ int i;
1301+ for(i=0; i<pConfig->nLang; i++){
1302+ if( pConfig->a[i].iLang==iLang ) return &pConfig->a[i];
1303+ }
1304+ return &editDist3Lang;
1305+}
1306+
1307+/*
1308+** Function: editdist3(A,B,iLang)
1309+** editdist3(tablename)
1310+**
1311+** Return the cost of transforming string A into string B using edit
1312+** weights for iLang.
1313+**
1314+** The second form loads edit weights into memory from a table.
1315+*/
1316+static void editDist3SqlFunc(
1317+ sqlite3_context *context,
1318+ int argc,
1319+ sqlite3_value **argv
1320+){
1321+ EditDist3Config *pConfig = (EditDist3Config*)sqlite3_user_data(context);
1322+ sqlite3 *db = sqlite3_context_db_handle(context);
1323+ int rc;
1324+ if( argc==1 ){
1325+ const char *zTable = (const char*)sqlite3_value_text(argv[0]);
1326+ rc = editDist3ConfigLoad(pConfig, db, zTable);
1327+ if( rc ) sqlite3_result_error_code(context, rc);
1328+ }else{
1329+ const char *zA = (const char*)sqlite3_value_text(argv[0]);
1330+ const char *zB = (const char*)sqlite3_value_text(argv[1]);
1331+ int nA = sqlite3_value_bytes(argv[0]);
1332+ int nB = sqlite3_value_bytes(argv[1]);
1333+ int iLang = argc==3 ? sqlite3_value_int(argv[2]) : 0;
1334+ const EditDist3Lang *pLang = editDist3FindLang(pConfig, iLang);
1335+ EditDist3FromString *pFrom;
1336+ int dist;
1337+
1338+ pFrom = editDist3FromStringNew(pLang, zA, nA);
1339+ if( pFrom==0 ){
1340+ sqlite3_result_error_nomem(context);
1341+ return;
1342+ }
1343+ dist = editDist3Core(pFrom, zB, nB, pLang, 0);
1344+ editDist3FromStringDelete(pFrom);
1345+ if( dist==(-1) ){
1346+ sqlite3_result_error_nomem(context);
1347+ }else{
1348+ sqlite3_result_int(context, dist);
1349+ }
1350+ }
1351+}
1352+
1353+/*
1354+** Register the editDist3 function with SQLite
1355+*/
1356+static int editDist3Install(sqlite3 *db){
1357+ int rc;
1358+ EditDist3Config *pConfig = sqlite3_malloc64( sizeof(*pConfig) );
1359+ if( pConfig==0 ) return SQLITE_NOMEM;
1360+ memset(pConfig, 0, sizeof(*pConfig));
1361+ rc = sqlite3_create_function_v2(db, "editdist3",
1362+ 2, SQLITE_UTF8|SQLITE_DETERMINISTIC, pConfig,
1363+ editDist3SqlFunc, 0, 0, 0);
1364+ if( rc==SQLITE_OK ){
1365+ rc = sqlite3_create_function_v2(db, "editdist3",
1366+ 3, SQLITE_UTF8|SQLITE_DETERMINISTIC, pConfig,
1367+ editDist3SqlFunc, 0, 0, 0);
1368+ }
1369+ if( rc==SQLITE_OK ){
1370+ rc = sqlite3_create_function_v2(db, "editdist3",
1371+ 1, SQLITE_UTF8|SQLITE_DETERMINISTIC, pConfig,
1372+ editDist3SqlFunc, 0, 0, editDist3ConfigDelete);
1373+ }else{
1374+ sqlite3_free(pConfig);
1375+ }
1376+ return rc;
1377+}
1378+/* End configurable cost unicode edit distance routines
1379+******************************************************************************
1380+******************************************************************************
1381+** Begin transliterate unicode-to-ascii implementation
1382+*/
1383+
1384+#if !SQLITE_AMALGAMATION
1385+/*
1386+** This lookup table is used to help decode the first byte of
1387+** a multi-byte UTF8 character.
1388+*/
1389+static const unsigned char sqlite3Utf8Trans1[] = {
1390+ 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
1391+ 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
1392+ 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
1393+ 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
1394+ 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
1395+ 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
1396+ 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
1397+ 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00,
1398+};
1399+#endif
1400+
1401+/*
1402+** Return the value of the first UTF-8 character in the string.
1403+*/
1404+static int utf8Read(const unsigned char *z, int n, int *pSize){
1405+ int c, i;
1406+
1407+ /* All callers to this routine (in the current implementation)
1408+ ** always have n>0. */
1409+ if( NEVER(n==0) ){
1410+ c = i = 0;
1411+ }else{
1412+ c = z[0];
1413+ i = 1;
1414+ if( c>=0xc0 ){
1415+ c = sqlite3Utf8Trans1[c-0xc0];
1416+ while( i<n && (z[i] & 0xc0)==0x80 ){
1417+ c = (c<<6) + (0x3f & z[i++]);
1418+ }
1419+ }
1420+ }
1421+ *pSize = i;
1422+ return c;
1423+}
1424+
1425+/*
1426+** Return the number of characters in the utf-8 string in the nIn byte
1427+** buffer pointed to by zIn.
1428+*/
1429+static int utf8Charlen(const char *zIn, int nIn){
1430+ int i;
1431+ int nChar = 0;
1432+ for(i=0; i<nIn; nChar++){
1433+ int sz;
1434+ utf8Read((const unsigned char *)&zIn[i], nIn-i, &sz);
1435+ i += sz;
1436+ }
1437+ return nChar;
1438+}
1439+
1440+typedef struct Transliteration Transliteration;
1441+struct Transliteration {
1442+ unsigned short int cFrom;
1443+ unsigned char cTo0, cTo1, cTo2, cTo3;
1444+#ifdef SQLITE_SPELLFIX_5BYTE_MAPPINGS
1445+ unsigned char cTo4;
1446+#endif
1447+};
1448+
1449+/*
1450+** Table of translations from unicode characters into ASCII.
1451+*/
1452+static const Transliteration translit[] = {
1453+ { 0x00A0, 0x20, 0x00, 0x00, 0x00 }, /* to */
1454+ { 0x00B5, 0x75, 0x00, 0x00, 0x00 }, /* µ to u */
1455+ { 0x00C0, 0x41, 0x00, 0x00, 0x00 }, /* À to A */
1456+ { 0x00C1, 0x41, 0x00, 0x00, 0x00 }, /* Á to A */
1457+ { 0x00C2, 0x41, 0x00, 0x00, 0x00 }, /* Â to A */
1458+ { 0x00C3, 0x41, 0x00, 0x00, 0x00 }, /* Ã to A */
1459+ { 0x00C4, 0x41, 0x65, 0x00, 0x00 }, /* Ä to Ae */
1460+ { 0x00C5, 0x41, 0x61, 0x00, 0x00 }, /* Å to Aa */
1461+ { 0x00C6, 0x41, 0x45, 0x00, 0x00 }, /* Æ to AE */
1462+ { 0x00C7, 0x43, 0x00, 0x00, 0x00 }, /* Ç to C */
1463+ { 0x00C8, 0x45, 0x00, 0x00, 0x00 }, /* È to E */
1464+ { 0x00C9, 0x45, 0x00, 0x00, 0x00 }, /* É to E */
1465+ { 0x00CA, 0x45, 0x00, 0x00, 0x00 }, /* Ê to E */
1466+ { 0x00CB, 0x45, 0x00, 0x00, 0x00 }, /* Ë to E */
1467+ { 0x00CC, 0x49, 0x00, 0x00, 0x00 }, /* Ì to I */
1468+ { 0x00CD, 0x49, 0x00, 0x00, 0x00 }, /* Í to I */
1469+ { 0x00CE, 0x49, 0x00, 0x00, 0x00 }, /* Î to I */
1470+ { 0x00CF, 0x49, 0x00, 0x00, 0x00 }, /* Ï to I */
1471+ { 0x00D0, 0x44, 0x00, 0x00, 0x00 }, /* Ð to D */
1472+ { 0x00D1, 0x4E, 0x00, 0x00, 0x00 }, /* Ñ to N */
1473+ { 0x00D2, 0x4F, 0x00, 0x00, 0x00 }, /* Ò to O */
1474+ { 0x00D3, 0x4F, 0x00, 0x00, 0x00 }, /* Ó to O */
1475+ { 0x00D4, 0x4F, 0x00, 0x00, 0x00 }, /* Ô to O */
1476+ { 0x00D5, 0x4F, 0x00, 0x00, 0x00 }, /* Õ to O */
1477+ { 0x00D6, 0x4F, 0x65, 0x00, 0x00 }, /* Ö to Oe */
1478+ { 0x00D7, 0x78, 0x00, 0x00, 0x00 }, /* × to x */
1479+ { 0x00D8, 0x4F, 0x00, 0x00, 0x00 }, /* Ø to O */
1480+ { 0x00D9, 0x55, 0x00, 0x00, 0x00 }, /* Ù to U */
1481+ { 0x00DA, 0x55, 0x00, 0x00, 0x00 }, /* Ú to U */
1482+ { 0x00DB, 0x55, 0x00, 0x00, 0x00 }, /* Û to U */
1483+ { 0x00DC, 0x55, 0x65, 0x00, 0x00 }, /* Ü to Ue */
1484+ { 0x00DD, 0x59, 0x00, 0x00, 0x00 }, /* Ý to Y */
1485+ { 0x00DE, 0x54, 0x68, 0x00, 0x00 }, /* Þ to Th */
1486+ { 0x00DF, 0x73, 0x73, 0x00, 0x00 }, /* ß to ss */
1487+ { 0x00E0, 0x61, 0x00, 0x00, 0x00 }, /* à to a */
1488+ { 0x00E1, 0x61, 0x00, 0x00, 0x00 }, /* á to a */
1489+ { 0x00E2, 0x61, 0x00, 0x00, 0x00 }, /* â to a */
1490+ { 0x00E3, 0x61, 0x00, 0x00, 0x00 }, /* ã to a */
1491+ { 0x00E4, 0x61, 0x65, 0x00, 0x00 }, /* ä to ae */
1492+ { 0x00E5, 0x61, 0x61, 0x00, 0x00 }, /* å to aa */
1493+ { 0x00E6, 0x61, 0x65, 0x00, 0x00 }, /* æ to ae */
1494+ { 0x00E7, 0x63, 0x00, 0x00, 0x00 }, /* ç to c */
1495+ { 0x00E8, 0x65, 0x00, 0x00, 0x00 }, /* è to e */
1496+ { 0x00E9, 0x65, 0x00, 0x00, 0x00 }, /* é to e */
1497+ { 0x00EA, 0x65, 0x00, 0x00, 0x00 }, /* ê to e */
1498+ { 0x00EB, 0x65, 0x00, 0x00, 0x00 }, /* ë to e */
1499+ { 0x00EC, 0x69, 0x00, 0x00, 0x00 }, /* ì to i */
1500+ { 0x00ED, 0x69, 0x00, 0x00, 0x00 }, /* í to i */
1501+ { 0x00EE, 0x69, 0x00, 0x00, 0x00 }, /* î to i */
1502+ { 0x00EF, 0x69, 0x00, 0x00, 0x00 }, /* ï to i */
1503+ { 0x00F0, 0x64, 0x00, 0x00, 0x00 }, /* ð to d */
1504+ { 0x00F1, 0x6E, 0x00, 0x00, 0x00 }, /* ñ to n */
1505+ { 0x00F2, 0x6F, 0x00, 0x00, 0x00 }, /* ò to o */
1506+ { 0x00F3, 0x6F, 0x00, 0x00, 0x00 }, /* ó to o */
1507+ { 0x00F4, 0x6F, 0x00, 0x00, 0x00 }, /* ô to o */
1508+ { 0x00F5, 0x6F, 0x00, 0x00, 0x00 }, /* õ to o */
1509+ { 0x00F6, 0x6F, 0x65, 0x00, 0x00 }, /* ö to oe */
1510+ { 0x00F7, 0x3A, 0x00, 0x00, 0x00 }, /* ÷ to : */
1511+ { 0x00F8, 0x6F, 0x00, 0x00, 0x00 }, /* ø to o */
1512+ { 0x00F9, 0x75, 0x00, 0x00, 0x00 }, /* ù to u */
1513+ { 0x00FA, 0x75, 0x00, 0x00, 0x00 }, /* ú to u */
1514+ { 0x00FB, 0x75, 0x00, 0x00, 0x00 }, /* û to u */
1515+ { 0x00FC, 0x75, 0x65, 0x00, 0x00 }, /* ü to ue */
1516+ { 0x00FD, 0x79, 0x00, 0x00, 0x00 }, /* ý to y */
1517+ { 0x00FE, 0x74, 0x68, 0x00, 0x00 }, /* þ to th */
1518+ { 0x00FF, 0x79, 0x00, 0x00, 0x00 }, /* ÿ to y */
1519+ { 0x0100, 0x41, 0x00, 0x00, 0x00 }, /* Ā to A */
1520+ { 0x0101, 0x61, 0x00, 0x00, 0x00 }, /* ā to a */
1521+ { 0x0102, 0x41, 0x00, 0x00, 0x00 }, /* Ă to A */
1522+ { 0x0103, 0x61, 0x00, 0x00, 0x00 }, /* ă to a */
1523+ { 0x0104, 0x41, 0x00, 0x00, 0x00 }, /* Ą to A */
1524+ { 0x0105, 0x61, 0x00, 0x00, 0x00 }, /* ą to a */
1525+ { 0x0106, 0x43, 0x00, 0x00, 0x00 }, /* Ć to C */
1526+ { 0x0107, 0x63, 0x00, 0x00, 0x00 }, /* ć to c */
1527+ { 0x0108, 0x43, 0x68, 0x00, 0x00 }, /* Ĉ to Ch */
1528+ { 0x0109, 0x63, 0x68, 0x00, 0x00 }, /* ĉ to ch */
1529+ { 0x010A, 0x43, 0x00, 0x00, 0x00 }, /* Ċ to C */
1530+ { 0x010B, 0x63, 0x00, 0x00, 0x00 }, /* ċ to c */
1531+ { 0x010C, 0x43, 0x00, 0x00, 0x00 }, /* Č to C */
1532+ { 0x010D, 0x63, 0x00, 0x00, 0x00 }, /* č to c */
1533+ { 0x010E, 0x44, 0x00, 0x00, 0x00 }, /* Ď to D */
1534+ { 0x010F, 0x64, 0x00, 0x00, 0x00 }, /* ď to d */
1535+ { 0x0110, 0x44, 0x00, 0x00, 0x00 }, /* Đ to D */
1536+ { 0x0111, 0x64, 0x00, 0x00, 0x00 }, /* đ to d */
1537+ { 0x0112, 0x45, 0x00, 0x00, 0x00 }, /* Ē to E */
1538+ { 0x0113, 0x65, 0x00, 0x00, 0x00 }, /* ē to e */
1539+ { 0x0114, 0x45, 0x00, 0x00, 0x00 }, /* Ĕ to E */
1540+ { 0x0115, 0x65, 0x00, 0x00, 0x00 }, /* ĕ to e */
1541+ { 0x0116, 0x45, 0x00, 0x00, 0x00 }, /* Ė to E */
1542+ { 0x0117, 0x65, 0x00, 0x00, 0x00 }, /* ė to e */
1543+ { 0x0118, 0x45, 0x00, 0x00, 0x00 }, /* Ę to E */
1544+ { 0x0119, 0x65, 0x00, 0x00, 0x00 }, /* ę to e */
1545+ { 0x011A, 0x45, 0x00, 0x00, 0x00 }, /* Ě to E */
1546+ { 0x011B, 0x65, 0x00, 0x00, 0x00 }, /* ě to e */
1547+ { 0x011C, 0x47, 0x68, 0x00, 0x00 }, /* Ĝ to Gh */
1548+ { 0x011D, 0x67, 0x68, 0x00, 0x00 }, /* ĝ to gh */
1549+ { 0x011E, 0x47, 0x00, 0x00, 0x00 }, /* Ğ to G */
1550+ { 0x011F, 0x67, 0x00, 0x00, 0x00 }, /* ğ to g */
1551+ { 0x0120, 0x47, 0x00, 0x00, 0x00 }, /* Ġ to G */
1552+ { 0x0121, 0x67, 0x00, 0x00, 0x00 }, /* ġ to g */
1553+ { 0x0122, 0x47, 0x00, 0x00, 0x00 }, /* Ģ to G */
1554+ { 0x0123, 0x67, 0x00, 0x00, 0x00 }, /* ģ to g */
1555+ { 0x0124, 0x48, 0x68, 0x00, 0x00 }, /* Ĥ to Hh */
1556+ { 0x0125, 0x68, 0x68, 0x00, 0x00 }, /* ĥ to hh */
1557+ { 0x0126, 0x48, 0x00, 0x00, 0x00 }, /* Ħ to H */
1558+ { 0x0127, 0x68, 0x00, 0x00, 0x00 }, /* ħ to h */
1559+ { 0x0128, 0x49, 0x00, 0x00, 0x00 }, /* Ĩ to I */
1560+ { 0x0129, 0x69, 0x00, 0x00, 0x00 }, /* ĩ to i */
1561+ { 0x012A, 0x49, 0x00, 0x00, 0x00 }, /* Ī to I */
1562+ { 0x012B, 0x69, 0x00, 0x00, 0x00 }, /* ī to i */
1563+ { 0x012C, 0x49, 0x00, 0x00, 0x00 }, /* Ĭ to I */
1564+ { 0x012D, 0x69, 0x00, 0x00, 0x00 }, /* ĭ to i */
1565+ { 0x012E, 0x49, 0x00, 0x00, 0x00 }, /* Į to I */
1566+ { 0x012F, 0x69, 0x00, 0x00, 0x00 }, /* į to i */
1567+ { 0x0130, 0x49, 0x00, 0x00, 0x00 }, /* İ to I */
1568+ { 0x0131, 0x69, 0x00, 0x00, 0x00 }, /* ı to i */
1569+ { 0x0132, 0x49, 0x4A, 0x00, 0x00 }, /* IJ to IJ */
1570+ { 0x0133, 0x69, 0x6A, 0x00, 0x00 }, /* ij to ij */
1571+ { 0x0134, 0x4A, 0x68, 0x00, 0x00 }, /* Ĵ to Jh */
1572+ { 0x0135, 0x6A, 0x68, 0x00, 0x00 }, /* ĵ to jh */
1573+ { 0x0136, 0x4B, 0x00, 0x00, 0x00 }, /* Ķ to K */
1574+ { 0x0137, 0x6B, 0x00, 0x00, 0x00 }, /* ķ to k */
1575+ { 0x0138, 0x6B, 0x00, 0x00, 0x00 }, /* ĸ to k */
1576+ { 0x0139, 0x4C, 0x00, 0x00, 0x00 }, /* Ĺ to L */
1577+ { 0x013A, 0x6C, 0x00, 0x00, 0x00 }, /* ĺ to l */
1578+ { 0x013B, 0x4C, 0x00, 0x00, 0x00 }, /* Ļ to L */
1579+ { 0x013C, 0x6C, 0x00, 0x00, 0x00 }, /* ļ to l */
1580+ { 0x013D, 0x4C, 0x00, 0x00, 0x00 }, /* Ľ to L */
1581+ { 0x013E, 0x6C, 0x00, 0x00, 0x00 }, /* ľ to l */
1582+ { 0x013F, 0x4C, 0x2E, 0x00, 0x00 }, /* Ŀ to L. */
1583+ { 0x0140, 0x6C, 0x2E, 0x00, 0x00 }, /* ŀ to l. */
1584+ { 0x0141, 0x4C, 0x00, 0x00, 0x00 }, /* Ł to L */
1585+ { 0x0142, 0x6C, 0x00, 0x00, 0x00 }, /* ł to l */
1586+ { 0x0143, 0x4E, 0x00, 0x00, 0x00 }, /* Ń to N */
1587+ { 0x0144, 0x6E, 0x00, 0x00, 0x00 }, /* ń to n */
1588+ { 0x0145, 0x4E, 0x00, 0x00, 0x00 }, /* Ņ to N */
1589+ { 0x0146, 0x6E, 0x00, 0x00, 0x00 }, /* ņ to n */
1590+ { 0x0147, 0x4E, 0x00, 0x00, 0x00 }, /* Ň to N */
1591+ { 0x0148, 0x6E, 0x00, 0x00, 0x00 }, /* ň to n */
1592+ { 0x0149, 0x27, 0x6E, 0x00, 0x00 }, /* ʼn to 'n */
1593+ { 0x014A, 0x4E, 0x47, 0x00, 0x00 }, /* Ŋ to NG */
1594+ { 0x014B, 0x6E, 0x67, 0x00, 0x00 }, /* ŋ to ng */
1595+ { 0x014C, 0x4F, 0x00, 0x00, 0x00 }, /* Ō to O */
1596+ { 0x014D, 0x6F, 0x00, 0x00, 0x00 }, /* ō to o */
1597+ { 0x014E, 0x4F, 0x00, 0x00, 0x00 }, /* Ŏ to O */
1598+ { 0x014F, 0x6F, 0x00, 0x00, 0x00 }, /* ŏ to o */
1599+ { 0x0150, 0x4F, 0x00, 0x00, 0x00 }, /* Ő to O */
1600+ { 0x0151, 0x6F, 0x00, 0x00, 0x00 }, /* ő to o */
1601+ { 0x0152, 0x4F, 0x45, 0x00, 0x00 }, /* Πto OE */
1602+ { 0x0153, 0x6F, 0x65, 0x00, 0x00 }, /* œ to oe */
1603+ { 0x0154, 0x52, 0x00, 0x00, 0x00 }, /* Ŕ to R */
1604+ { 0x0155, 0x72, 0x00, 0x00, 0x00 }, /* ŕ to r */
1605+ { 0x0156, 0x52, 0x00, 0x00, 0x00 }, /* Ŗ to R */
1606+ { 0x0157, 0x72, 0x00, 0x00, 0x00 }, /* ŗ to r */
1607+ { 0x0158, 0x52, 0x00, 0x00, 0x00 }, /* Ř to R */
1608+ { 0x0159, 0x72, 0x00, 0x00, 0x00 }, /* ř to r */
1609+ { 0x015A, 0x53, 0x00, 0x00, 0x00 }, /* Ś to S */
1610+ { 0x015B, 0x73, 0x00, 0x00, 0x00 }, /* ś to s */
1611+ { 0x015C, 0x53, 0x68, 0x00, 0x00 }, /* Ŝ to Sh */
1612+ { 0x015D, 0x73, 0x68, 0x00, 0x00 }, /* ŝ to sh */
1613+ { 0x015E, 0x53, 0x00, 0x00, 0x00 }, /* Ş to S */
1614+ { 0x015F, 0x73, 0x00, 0x00, 0x00 }, /* ş to s */
1615+ { 0x0160, 0x53, 0x00, 0x00, 0x00 }, /* Š to S */
1616+ { 0x0161, 0x73, 0x00, 0x00, 0x00 }, /* š to s */
1617+ { 0x0162, 0x54, 0x00, 0x00, 0x00 }, /* Ţ to T */
1618+ { 0x0163, 0x74, 0x00, 0x00, 0x00 }, /* ţ to t */
1619+ { 0x0164, 0x54, 0x00, 0x00, 0x00 }, /* Ť to T */
1620+ { 0x0165, 0x74, 0x00, 0x00, 0x00 }, /* ť to t */
1621+ { 0x0166, 0x54, 0x00, 0x00, 0x00 }, /* Ŧ to T */
1622+ { 0x0167, 0x74, 0x00, 0x00, 0x00 }, /* ŧ to t */
1623+ { 0x0168, 0x55, 0x00, 0x00, 0x00 }, /* Ũ to U */
1624+ { 0x0169, 0x75, 0x00, 0x00, 0x00 }, /* ũ to u */
1625+ { 0x016A, 0x55, 0x00, 0x00, 0x00 }, /* Ū to U */
1626+ { 0x016B, 0x75, 0x00, 0x00, 0x00 }, /* ū to u */
1627+ { 0x016C, 0x55, 0x00, 0x00, 0x00 }, /* Ŭ to U */
1628+ { 0x016D, 0x75, 0x00, 0x00, 0x00 }, /* ŭ to u */
1629+ { 0x016E, 0x55, 0x00, 0x00, 0x00 }, /* Ů to U */
1630+ { 0x016F, 0x75, 0x00, 0x00, 0x00 }, /* ů to u */
1631+ { 0x0170, 0x55, 0x00, 0x00, 0x00 }, /* Ű to U */
1632+ { 0x0171, 0x75, 0x00, 0x00, 0x00 }, /* ű to u */
1633+ { 0x0172, 0x55, 0x00, 0x00, 0x00 }, /* Ų to U */
1634+ { 0x0173, 0x75, 0x00, 0x00, 0x00 }, /* ų to u */
1635+ { 0x0174, 0x57, 0x00, 0x00, 0x00 }, /* Ŵ to W */
1636+ { 0x0175, 0x77, 0x00, 0x00, 0x00 }, /* ŵ to w */
1637+ { 0x0176, 0x59, 0x00, 0x00, 0x00 }, /* Ŷ to Y */
1638+ { 0x0177, 0x79, 0x00, 0x00, 0x00 }, /* ŷ to y */
1639+ { 0x0178, 0x59, 0x00, 0x00, 0x00 }, /* Ÿ to Y */
1640+ { 0x0179, 0x5A, 0x00, 0x00, 0x00 }, /* Ź to Z */
1641+ { 0x017A, 0x7A, 0x00, 0x00, 0x00 }, /* ź to z */
1642+ { 0x017B, 0x5A, 0x00, 0x00, 0x00 }, /* Ż to Z */
1643+ { 0x017C, 0x7A, 0x00, 0x00, 0x00 }, /* ż to z */
1644+ { 0x017D, 0x5A, 0x00, 0x00, 0x00 }, /* Ž to Z */
1645+ { 0x017E, 0x7A, 0x00, 0x00, 0x00 }, /* ž to z */
1646+ { 0x017F, 0x73, 0x00, 0x00, 0x00 }, /* ſ to s */
1647+ { 0x0192, 0x66, 0x00, 0x00, 0x00 }, /* ƒ to f */
1648+ { 0x0218, 0x53, 0x00, 0x00, 0x00 }, /* Ș to S */
1649+ { 0x0219, 0x73, 0x00, 0x00, 0x00 }, /* ș to s */
1650+ { 0x021A, 0x54, 0x00, 0x00, 0x00 }, /* Ț to T */
1651+ { 0x021B, 0x74, 0x00, 0x00, 0x00 }, /* ț to t */
1652+ { 0x0386, 0x41, 0x00, 0x00, 0x00 }, /* Ά to A */
1653+ { 0x0388, 0x45, 0x00, 0x00, 0x00 }, /* Έ to E */
1654+ { 0x0389, 0x49, 0x00, 0x00, 0x00 }, /* Ή to I */
1655+ { 0x038A, 0x49, 0x00, 0x00, 0x00 }, /* Ί to I */
1656+ { 0x038C, 0x4f, 0x00, 0x00, 0x00 }, /* Ό to O */
1657+ { 0x038E, 0x59, 0x00, 0x00, 0x00 }, /* Ύ to Y */
1658+ { 0x038F, 0x4f, 0x00, 0x00, 0x00 }, /* Ώ to O */
1659+ { 0x0390, 0x69, 0x00, 0x00, 0x00 }, /* ΐ to i */
1660+ { 0x0391, 0x41, 0x00, 0x00, 0x00 }, /* Α to A */
1661+ { 0x0392, 0x42, 0x00, 0x00, 0x00 }, /* Β to B */
1662+ { 0x0393, 0x47, 0x00, 0x00, 0x00 }, /* Γ to G */
1663+ { 0x0394, 0x44, 0x00, 0x00, 0x00 }, /* Δ to D */
1664+ { 0x0395, 0x45, 0x00, 0x00, 0x00 }, /* Ε to E */
1665+ { 0x0396, 0x5a, 0x00, 0x00, 0x00 }, /* Ζ to Z */
1666+ { 0x0397, 0x49, 0x00, 0x00, 0x00 }, /* Η to I */
1667+ { 0x0398, 0x54, 0x68, 0x00, 0x00 }, /* Θ to Th */
1668+ { 0x0399, 0x49, 0x00, 0x00, 0x00 }, /* Ι to I */
1669+ { 0x039A, 0x4b, 0x00, 0x00, 0x00 }, /* Κ to K */
1670+ { 0x039B, 0x4c, 0x00, 0x00, 0x00 }, /* Λ to L */
1671+ { 0x039C, 0x4d, 0x00, 0x00, 0x00 }, /* Μ to M */
1672+ { 0x039D, 0x4e, 0x00, 0x00, 0x00 }, /* Ν to N */
1673+ { 0x039E, 0x58, 0x00, 0x00, 0x00 }, /* Ξ to X */
1674+ { 0x039F, 0x4f, 0x00, 0x00, 0x00 }, /* Ο to O */
1675+ { 0x03A0, 0x50, 0x00, 0x00, 0x00 }, /* Π to P */
1676+ { 0x03A1, 0x52, 0x00, 0x00, 0x00 }, /* Ρ to R */
1677+ { 0x03A3, 0x53, 0x00, 0x00, 0x00 }, /* Σ to S */
1678+ { 0x03A4, 0x54, 0x00, 0x00, 0x00 }, /* Τ to T */
1679+ { 0x03A5, 0x59, 0x00, 0x00, 0x00 }, /* Υ to Y */
1680+ { 0x03A6, 0x46, 0x00, 0x00, 0x00 }, /* Φ to F */
1681+ { 0x03A7, 0x43, 0x68, 0x00, 0x00 }, /* Χ to Ch */
1682+ { 0x03A8, 0x50, 0x73, 0x00, 0x00 }, /* Ψ to Ps */
1683+ { 0x03A9, 0x4f, 0x00, 0x00, 0x00 }, /* Ω to O */
1684+ { 0x03AA, 0x49, 0x00, 0x00, 0x00 }, /* Ϊ to I */
1685+ { 0x03AB, 0x59, 0x00, 0x00, 0x00 }, /* Ϋ to Y */
1686+ { 0x03AC, 0x61, 0x00, 0x00, 0x00 }, /* ά to a */
1687+ { 0x03AD, 0x65, 0x00, 0x00, 0x00 }, /* έ to e */
1688+ { 0x03AE, 0x69, 0x00, 0x00, 0x00 }, /* ή to i */
1689+ { 0x03AF, 0x69, 0x00, 0x00, 0x00 }, /* ί to i */
1690+ { 0x03B1, 0x61, 0x00, 0x00, 0x00 }, /* α to a */
1691+ { 0x03B2, 0x62, 0x00, 0x00, 0x00 }, /* β to b */
1692+ { 0x03B3, 0x67, 0x00, 0x00, 0x00 }, /* γ to g */
1693+ { 0x03B4, 0x64, 0x00, 0x00, 0x00 }, /* δ to d */
1694+ { 0x03B5, 0x65, 0x00, 0x00, 0x00 }, /* ε to e */
1695+ { 0x03B6, 0x7a, 0x00, 0x00, 0x00 }, /* ζ to z */
1696+ { 0x03B7, 0x69, 0x00, 0x00, 0x00 }, /* η to i */
1697+ { 0x03B8, 0x74, 0x68, 0x00, 0x00 }, /* θ to th */
1698+ { 0x03B9, 0x69, 0x00, 0x00, 0x00 }, /* ι to i */
1699+ { 0x03BA, 0x6b, 0x00, 0x00, 0x00 }, /* κ to k */
1700+ { 0x03BB, 0x6c, 0x00, 0x00, 0x00 }, /* λ to l */
1701+ { 0x03BC, 0x6d, 0x00, 0x00, 0x00 }, /* μ to m */
1702+ { 0x03BD, 0x6e, 0x00, 0x00, 0x00 }, /* ν to n */
1703+ { 0x03BE, 0x78, 0x00, 0x00, 0x00 }, /* ξ to x */
1704+ { 0x03BF, 0x6f, 0x00, 0x00, 0x00 }, /* ο to o */
1705+ { 0x03C0, 0x70, 0x00, 0x00, 0x00 }, /* π to p */
1706+ { 0x03C1, 0x72, 0x00, 0x00, 0x00 }, /* ρ to r */
1707+ { 0x03C3, 0x73, 0x00, 0x00, 0x00 }, /* σ to s */
1708+ { 0x03C4, 0x74, 0x00, 0x00, 0x00 }, /* τ to t */
1709+ { 0x03C5, 0x79, 0x00, 0x00, 0x00 }, /* υ to y */
1710+ { 0x03C6, 0x66, 0x00, 0x00, 0x00 }, /* φ to f */
1711+ { 0x03C7, 0x63, 0x68, 0x00, 0x00 }, /* χ to ch */
1712+ { 0x03C8, 0x70, 0x73, 0x00, 0x00 }, /* ψ to ps */
1713+ { 0x03C9, 0x6f, 0x00, 0x00, 0x00 }, /* ω to o */
1714+ { 0x03CA, 0x69, 0x00, 0x00, 0x00 }, /* ϊ to i */
1715+ { 0x03CB, 0x79, 0x00, 0x00, 0x00 }, /* ϋ to y */
1716+ { 0x03CC, 0x6f, 0x00, 0x00, 0x00 }, /* ό to o */
1717+ { 0x03CD, 0x79, 0x00, 0x00, 0x00 }, /* ύ to y */
1718+ { 0x03CE, 0x69, 0x00, 0x00, 0x00 }, /* ώ to i */
1719+ { 0x0400, 0x45, 0x00, 0x00, 0x00 }, /* Ѐ to E */
1720+ { 0x0401, 0x45, 0x00, 0x00, 0x00 }, /* Ё to E */
1721+ { 0x0402, 0x44, 0x00, 0x00, 0x00 }, /* Ђ to D */
1722+ { 0x0403, 0x47, 0x00, 0x00, 0x00 }, /* Ѓ to G */
1723+ { 0x0404, 0x45, 0x00, 0x00, 0x00 }, /* Є to E */
1724+ { 0x0405, 0x5a, 0x00, 0x00, 0x00 }, /* Ѕ to Z */
1725+ { 0x0406, 0x49, 0x00, 0x00, 0x00 }, /* І to I */
1726+ { 0x0407, 0x49, 0x00, 0x00, 0x00 }, /* Ї to I */
1727+ { 0x0408, 0x4a, 0x00, 0x00, 0x00 }, /* Ј to J */
1728+ { 0x0409, 0x49, 0x00, 0x00, 0x00 }, /* Љ to I */
1729+ { 0x040A, 0x4e, 0x00, 0x00, 0x00 }, /* Њ to N */
1730+ { 0x040B, 0x44, 0x00, 0x00, 0x00 }, /* Ћ to D */
1731+ { 0x040C, 0x4b, 0x00, 0x00, 0x00 }, /* Ќ to K */
1732+ { 0x040D, 0x49, 0x00, 0x00, 0x00 }, /* Ѝ to I */
1733+ { 0x040E, 0x55, 0x00, 0x00, 0x00 }, /* Ў to U */
1734+ { 0x040F, 0x44, 0x00, 0x00, 0x00 }, /* Џ to D */
1735+ { 0x0410, 0x41, 0x00, 0x00, 0x00 }, /* А to A */
1736+ { 0x0411, 0x42, 0x00, 0x00, 0x00 }, /* Б to B */
1737+ { 0x0412, 0x56, 0x00, 0x00, 0x00 }, /* В to V */
1738+ { 0x0413, 0x47, 0x00, 0x00, 0x00 }, /* Г to G */
1739+ { 0x0414, 0x44, 0x00, 0x00, 0x00 }, /* Д to D */
1740+ { 0x0415, 0x45, 0x00, 0x00, 0x00 }, /* Е to E */
1741+ { 0x0416, 0x5a, 0x68, 0x00, 0x00 }, /* Ж to Zh */
1742+ { 0x0417, 0x5a, 0x00, 0x00, 0x00 }, /* З to Z */
1743+ { 0x0418, 0x49, 0x00, 0x00, 0x00 }, /* И to I */
1744+ { 0x0419, 0x49, 0x00, 0x00, 0x00 }, /* Й to I */
1745+ { 0x041A, 0x4b, 0x00, 0x00, 0x00 }, /* К to K */
1746+ { 0x041B, 0x4c, 0x00, 0x00, 0x00 }, /* Л to L */
1747+ { 0x041C, 0x4d, 0x00, 0x00, 0x00 }, /* М to M */
1748+ { 0x041D, 0x4e, 0x00, 0x00, 0x00 }, /* Н to N */
1749+ { 0x041E, 0x4f, 0x00, 0x00, 0x00 }, /* О to O */
1750+ { 0x041F, 0x50, 0x00, 0x00, 0x00 }, /* П to P */
1751+ { 0x0420, 0x52, 0x00, 0x00, 0x00 }, /* Р to R */
1752+ { 0x0421, 0x53, 0x00, 0x00, 0x00 }, /* С to S */
1753+ { 0x0422, 0x54, 0x00, 0x00, 0x00 }, /* Т to T */
1754+ { 0x0423, 0x55, 0x00, 0x00, 0x00 }, /* У to U */
1755+ { 0x0424, 0x46, 0x00, 0x00, 0x00 }, /* Ф to F */
1756+ { 0x0425, 0x4b, 0x68, 0x00, 0x00 }, /* Х to Kh */
1757+ { 0x0426, 0x54, 0x63, 0x00, 0x00 }, /* Ц to Tc */
1758+ { 0x0427, 0x43, 0x68, 0x00, 0x00 }, /* Ч to Ch */
1759+ { 0x0428, 0x53, 0x68, 0x00, 0x00 }, /* Ш to Sh */
1760+ { 0x0429, 0x53, 0x68, 0x63, 0x68 }, /* Щ to Shch */
1761+ { 0x042A, 0x61, 0x00, 0x00, 0x00 }, /* to A */
1762+ { 0x042B, 0x59, 0x00, 0x00, 0x00 }, /* Ы to Y */
1763+ { 0x042C, 0x59, 0x00, 0x00, 0x00 }, /* to Y */
1764+ { 0x042D, 0x45, 0x00, 0x00, 0x00 }, /* Э to E */
1765+ { 0x042E, 0x49, 0x75, 0x00, 0x00 }, /* Ю to Iu */
1766+ { 0x042F, 0x49, 0x61, 0x00, 0x00 }, /* Я to Ia */
1767+ { 0x0430, 0x61, 0x00, 0x00, 0x00 }, /* а to a */
1768+ { 0x0431, 0x62, 0x00, 0x00, 0x00 }, /* б to b */
1769+ { 0x0432, 0x76, 0x00, 0x00, 0x00 }, /* в to v */
1770+ { 0x0433, 0x67, 0x00, 0x00, 0x00 }, /* г to g */
1771+ { 0x0434, 0x64, 0x00, 0x00, 0x00 }, /* д to d */
1772+ { 0x0435, 0x65, 0x00, 0x00, 0x00 }, /* е to e */
1773+ { 0x0436, 0x7a, 0x68, 0x00, 0x00 }, /* ж to zh */
1774+ { 0x0437, 0x7a, 0x00, 0x00, 0x00 }, /* з to z */
1775+ { 0x0438, 0x69, 0x00, 0x00, 0x00 }, /* и to i */
1776+ { 0x0439, 0x69, 0x00, 0x00, 0x00 }, /* й to i */
1777+ { 0x043A, 0x6b, 0x00, 0x00, 0x00 }, /* к to k */
1778+ { 0x043B, 0x6c, 0x00, 0x00, 0x00 }, /* л to l */
1779+ { 0x043C, 0x6d, 0x00, 0x00, 0x00 }, /* м to m */
1780+ { 0x043D, 0x6e, 0x00, 0x00, 0x00 }, /* н to n */
1781+ { 0x043E, 0x6f, 0x00, 0x00, 0x00 }, /* о to o */
1782+ { 0x043F, 0x70, 0x00, 0x00, 0x00 }, /* п to p */
1783+ { 0x0440, 0x72, 0x00, 0x00, 0x00 }, /* р to r */
1784+ { 0x0441, 0x73, 0x00, 0x00, 0x00 }, /* с to s */
1785+ { 0x0442, 0x74, 0x00, 0x00, 0x00 }, /* т to t */
1786+ { 0x0443, 0x75, 0x00, 0x00, 0x00 }, /* у to u */
1787+ { 0x0444, 0x66, 0x00, 0x00, 0x00 }, /* ф to f */
1788+ { 0x0445, 0x6b, 0x68, 0x00, 0x00 }, /* х to kh */
1789+ { 0x0446, 0x74, 0x63, 0x00, 0x00 }, /* ц to tc */
1790+ { 0x0447, 0x63, 0x68, 0x00, 0x00 }, /* ч to ch */
1791+ { 0x0448, 0x73, 0x68, 0x00, 0x00 }, /* ш to sh */
1792+ { 0x0449, 0x73, 0x68, 0x63, 0x68 }, /* щ to shch */
1793+ { 0x044A, 0x61, 0x00, 0x00, 0x00 }, /* to a */
1794+ { 0x044B, 0x79, 0x00, 0x00, 0x00 }, /* ы to y */
1795+ { 0x044C, 0x79, 0x00, 0x00, 0x00 }, /* to y */
1796+ { 0x044D, 0x65, 0x00, 0x00, 0x00 }, /* э to e */
1797+ { 0x044E, 0x69, 0x75, 0x00, 0x00 }, /* ю to iu */
1798+ { 0x044F, 0x69, 0x61, 0x00, 0x00 }, /* я to ia */
1799+ { 0x0450, 0x65, 0x00, 0x00, 0x00 }, /* ѐ to e */
1800+ { 0x0451, 0x65, 0x00, 0x00, 0x00 }, /* ё to e */
1801+ { 0x0452, 0x64, 0x00, 0x00, 0x00 }, /* ђ to d */
1802+ { 0x0453, 0x67, 0x00, 0x00, 0x00 }, /* ѓ to g */
1803+ { 0x0454, 0x65, 0x00, 0x00, 0x00 }, /* є to e */
1804+ { 0x0455, 0x7a, 0x00, 0x00, 0x00 }, /* ѕ to z */
1805+ { 0x0456, 0x69, 0x00, 0x00, 0x00 }, /* і to i */
1806+ { 0x0457, 0x69, 0x00, 0x00, 0x00 }, /* ї to i */
1807+ { 0x0458, 0x6a, 0x00, 0x00, 0x00 }, /* ј to j */
1808+ { 0x0459, 0x69, 0x00, 0x00, 0x00 }, /* љ to i */
1809+ { 0x045A, 0x6e, 0x00, 0x00, 0x00 }, /* њ to n */
1810+ { 0x045B, 0x64, 0x00, 0x00, 0x00 }, /* ћ to d */
1811+ { 0x045C, 0x6b, 0x00, 0x00, 0x00 }, /* ќ to k */
1812+ { 0x045D, 0x69, 0x00, 0x00, 0x00 }, /* ѝ to i */
1813+ { 0x045E, 0x75, 0x00, 0x00, 0x00 }, /* ў to u */
1814+ { 0x045F, 0x64, 0x00, 0x00, 0x00 }, /* џ to d */
1815+ { 0x1E02, 0x42, 0x00, 0x00, 0x00 }, /* Ḃ to B */
1816+ { 0x1E03, 0x62, 0x00, 0x00, 0x00 }, /* ḃ to b */
1817+ { 0x1E0A, 0x44, 0x00, 0x00, 0x00 }, /* Ḋ to D */
1818+ { 0x1E0B, 0x64, 0x00, 0x00, 0x00 }, /* ḋ to d */
1819+ { 0x1E1E, 0x46, 0x00, 0x00, 0x00 }, /* Ḟ to F */
1820+ { 0x1E1F, 0x66, 0x00, 0x00, 0x00 }, /* ḟ to f */
1821+ { 0x1E40, 0x4D, 0x00, 0x00, 0x00 }, /* Ṁ to M */
1822+ { 0x1E41, 0x6D, 0x00, 0x00, 0x00 }, /* ṁ to m */
1823+ { 0x1E56, 0x50, 0x00, 0x00, 0x00 }, /* Ṗ to P */
1824+ { 0x1E57, 0x70, 0x00, 0x00, 0x00 }, /* ṗ to p */
1825+ { 0x1E60, 0x53, 0x00, 0x00, 0x00 }, /* Ṡ to S */
1826+ { 0x1E61, 0x73, 0x00, 0x00, 0x00 }, /* ṡ to s */
1827+ { 0x1E6A, 0x54, 0x00, 0x00, 0x00 }, /* Ṫ to T */
1828+ { 0x1E6B, 0x74, 0x00, 0x00, 0x00 }, /* ṫ to t */
1829+ { 0x1E80, 0x57, 0x00, 0x00, 0x00 }, /* Ẁ to W */
1830+ { 0x1E81, 0x77, 0x00, 0x00, 0x00 }, /* ẁ to w */
1831+ { 0x1E82, 0x57, 0x00, 0x00, 0x00 }, /* Ẃ to W */
1832+ { 0x1E83, 0x77, 0x00, 0x00, 0x00 }, /* ẃ to w */
1833+ { 0x1E84, 0x57, 0x00, 0x00, 0x00 }, /* Ẅ to W */
1834+ { 0x1E85, 0x77, 0x00, 0x00, 0x00 }, /* ẅ to w */
1835+ { 0x1EF2, 0x59, 0x00, 0x00, 0x00 }, /* Ỳ to Y */
1836+ { 0x1EF3, 0x79, 0x00, 0x00, 0x00 }, /* ỳ to y */
1837+ { 0xFB00, 0x66, 0x66, 0x00, 0x00 }, /* ff to ff */
1838+ { 0xFB01, 0x66, 0x69, 0x00, 0x00 }, /* fi to fi */
1839+ { 0xFB02, 0x66, 0x6C, 0x00, 0x00 }, /* fl to fl */
1840+ { 0xFB05, 0x73, 0x74, 0x00, 0x00 }, /* ſt to st */
1841+ { 0xFB06, 0x73, 0x74, 0x00, 0x00 }, /* st to st */
1842+};
1843+
1844+static const Transliteration *spellfixFindTranslit(int c, int *pxTop){
1845+ *pxTop = (sizeof(translit)/sizeof(translit[0])) - 1;
1846+ return translit;
1847+}
1848+
1849+/*
1850+** Convert the input string from UTF-8 into pure ASCII by converting
1851+** all non-ASCII characters to some combination of characters in the
1852+** ASCII subset.
1853+**
1854+** The returned string might contain more characters than the input.
1855+**
1856+** Space to hold the returned string comes from sqlite3_malloc() and
1857+** should be freed by the caller.
1858+*/
1859+static unsigned char *transliterate(const unsigned char *zIn, int nIn){
1860+#ifdef SQLITE_SPELLFIX_5BYTE_MAPPINGS
1861+ unsigned char *zOut = sqlite3_malloc64( nIn*5 + 1 );
1862+#else
1863+ unsigned char *zOut = sqlite3_malloc64( nIn*4 + 1 );
1864+#endif
1865+ int c, sz, nOut;
1866+ if( zOut==0 ) return 0;
1867+ nOut = 0;
1868+ while( nIn>0 ){
1869+ c = utf8Read(zIn, nIn, &sz);
1870+ zIn += sz;
1871+ nIn -= sz;
1872+ if( c<=127 ){
1873+ zOut[nOut++] = (unsigned char)c;
1874+ }else{
1875+ int xTop, xBtm, x;
1876+ const Transliteration *tbl = spellfixFindTranslit(c, &xTop);
1877+ xBtm = 0;
1878+ while( xTop>=xBtm ){
1879+ x = (xTop + xBtm)/2;
1880+ if( tbl[x].cFrom==c ){
1881+ zOut[nOut++] = tbl[x].cTo0;
1882+ if( tbl[x].cTo1 ){
1883+ zOut[nOut++] = tbl[x].cTo1;
1884+ if( tbl[x].cTo2 ){
1885+ zOut[nOut++] = tbl[x].cTo2;
1886+ if( tbl[x].cTo3 ){
1887+ zOut[nOut++] = tbl[x].cTo3;
1888+#ifdef SQLITE_SPELLFIX_5BYTE_MAPPINGS
1889+ if( tbl[x].cTo4 ){
1890+ zOut[nOut++] = tbl[x].cTo4;
1891+ }
1892+#endif /* SQLITE_SPELLFIX_5BYTE_MAPPINGS */
1893+ }
1894+ }
1895+ }
1896+ c = 0;
1897+ break;
1898+ }else if( tbl[x].cFrom>c ){
1899+ xTop = x-1;
1900+ }else{
1901+ xBtm = x+1;
1902+ }
1903+ }
1904+ if( c ) zOut[nOut++] = '?';
1905+ }
1906+ }
1907+ zOut[nOut] = 0;
1908+ return zOut;
1909+}
1910+
1911+/*
1912+** Return the number of characters in the shortest prefix of the input
1913+** string that transliterates to an ASCII string nTrans bytes or longer.
1914+** Or, if the transliteration of the input string is less than nTrans
1915+** bytes in size, return the number of characters in the input string.
1916+*/
1917+static int translen_to_charlen(const char *zIn, int nIn, int nTrans){
1918+ int i, c, sz, nOut;
1919+ int nChar;
1920+
1921+ i = nOut = 0;
1922+ for(nChar=0; i<nIn && nOut<nTrans; nChar++){
1923+ c = utf8Read((const unsigned char *)&zIn[i], nIn-i, &sz);
1924+ i += sz;
1925+
1926+ nOut++;
1927+ if( c>=128 ){
1928+ int xTop, xBtm, x;
1929+ const Transliteration *tbl = spellfixFindTranslit(c, &xTop);
1930+ xBtm = 0;
1931+ while( xTop>=xBtm ){
1932+ x = (xTop + xBtm)/2;
1933+ if( tbl[x].cFrom==c ){
1934+ if( tbl[x].cTo1 ){
1935+ nOut++;
1936+ if( tbl[x].cTo2 ){
1937+ nOut++;
1938+ if( tbl[x].cTo3 ){
1939+ nOut++;
1940+ }
1941+ }
1942+ }
1943+ break;
1944+ }else if( tbl[x].cFrom>c ){
1945+ xTop = x-1;
1946+ }else{
1947+ xBtm = x+1;
1948+ }
1949+ }
1950+ }
1951+ }
1952+
1953+ return nChar;
1954+}
1955+
1956+
1957+/*
1958+** spellfix1_translit(X)
1959+**
1960+** Convert a string that contains non-ASCII Roman characters into
1961+** pure ASCII.
1962+*/
1963+static void transliterateSqlFunc(
1964+ sqlite3_context *context,
1965+ int argc,
1966+ sqlite3_value **argv
1967+){
1968+ const unsigned char *zIn = sqlite3_value_text(argv[0]);
1969+ int nIn = sqlite3_value_bytes(argv[0]);
1970+ unsigned char *zOut = transliterate(zIn, nIn);
1971+ if( zOut==0 ){
1972+ sqlite3_result_error_nomem(context);
1973+ }else{
1974+ sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free);
1975+ }
1976+}
1977+
1978+/*
1979+** spellfix1_scriptcode(X)
1980+**
1981+** Try to determine the dominant script used by the word X and return
1982+** its ISO 15924 numeric code.
1983+**
1984+** The current implementation only understands the following scripts:
1985+**
1986+** 215 (Latin)
1987+** 220 (Cyrillic)
1988+** 200 (Greek)
1989+**
1990+** This routine will return 998 if the input X contains characters from
1991+** two or more of the above scripts or 999 if X contains no characters
1992+** from any of the above scripts.
1993+*/
1994+static void scriptCodeSqlFunc(
1995+ sqlite3_context *context,
1996+ int argc,
1997+ sqlite3_value **argv
1998+){
1999+ const unsigned char *zIn = sqlite3_value_text(argv[0]);
2000+ int nIn = sqlite3_value_bytes(argv[0]);
2001+ int c, sz;
2002+ int scriptMask = 0;
2003+ int res;
2004+ int seenDigit = 0;
2005+# define SCRIPT_LATIN 0x0001
2006+# define SCRIPT_CYRILLIC 0x0002
2007+# define SCRIPT_GREEK 0x0004
2008+# define SCRIPT_HEBREW 0x0008
2009+# define SCRIPT_ARABIC 0x0010
2010+
2011+ while( nIn>0 ){
2012+ c = utf8Read(zIn, nIn, &sz);
2013+ zIn += sz;
2014+ nIn -= sz;
2015+ if( c<0x02af ){
2016+ if( c>=0x80 || midClass[c&0x7f]<CCLASS_DIGIT ){
2017+ scriptMask |= SCRIPT_LATIN;
2018+ }else if( c>='0' && c<='9' ){
2019+ seenDigit = 1;
2020+ }
2021+ }else if( c>=0x0400 && c<=0x04ff ){
2022+ scriptMask |= SCRIPT_CYRILLIC;
2023+ }else if( c>=0x0386 && c<=0x03ce ){
2024+ scriptMask |= SCRIPT_GREEK;
2025+ }else if( c>=0x0590 && c<=0x05ff ){
2026+ scriptMask |= SCRIPT_HEBREW;
2027+ }else if( c>=0x0600 && c<=0x06ff ){
2028+ scriptMask |= SCRIPT_ARABIC;
2029+ }
2030+ }
2031+ if( scriptMask==0 && seenDigit ) scriptMask = SCRIPT_LATIN;
2032+ switch( scriptMask ){
2033+ case 0: res = 999; break;
2034+ case SCRIPT_LATIN: res = 215; break;
2035+ case SCRIPT_CYRILLIC: res = 220; break;
2036+ case SCRIPT_GREEK: res = 200; break;
2037+ case SCRIPT_HEBREW: res = 125; break;
2038+ case SCRIPT_ARABIC: res = 160; break;
2039+ default: res = 998; break;
2040+ }
2041+ sqlite3_result_int(context, res);
2042+}
2043+
2044+/* End transliterate
2045+******************************************************************************
2046+******************************************************************************
2047+** Begin spellfix1 virtual table.
2048+*/
2049+
2050+/* Maximum length of a phonehash used for querying the shadow table */
2051+#define SPELLFIX_MX_HASH 32
2052+
2053+/* Maximum number of hash strings to examine per query */
2054+#define SPELLFIX_MX_RUN 1
2055+
2056+typedef struct spellfix1_vtab spellfix1_vtab;
2057+typedef struct spellfix1_cursor spellfix1_cursor;
2058+
2059+/* Fuzzy-search virtual table object */
2060+struct spellfix1_vtab {
2061+ sqlite3_vtab base; /* Base class - must be first */
2062+ sqlite3 *db; /* Database connection */
2063+ char *zDbName; /* Name of database holding this table */
2064+ char *zTableName; /* Name of the virtual table */
2065+ char *zCostTable; /* Table holding edit-distance cost numbers */
2066+ EditDist3Config *pConfig3; /* Parsed edit distance costs */
2067+};
2068+
2069+/* Fuzzy-search cursor object */
2070+struct spellfix1_cursor {
2071+ sqlite3_vtab_cursor base; /* Base class - must be first */
2072+ spellfix1_vtab *pVTab; /* The table to which this cursor belongs */
2073+ char *zPattern; /* rhs of MATCH clause */
2074+ int idxNum; /* idxNum value passed to xFilter() */
2075+ int nRow; /* Number of rows of content */
2076+ int nAlloc; /* Number of allocated rows */
2077+ int iRow; /* Current row of content */
2078+ int iLang; /* Value of the langid= constraint */
2079+ int iTop; /* Value of the top= constraint */
2080+ int iScope; /* Value of the scope= constraint */
2081+ int nSearch; /* Number of vocabulary items checked */
2082+ sqlite3_stmt *pFullScan; /* Shadow query for a full table scan */
2083+ struct spellfix1_row { /* For each row of content */
2084+ sqlite3_int64 iRowid; /* Rowid for this row */
2085+ char *zWord; /* Text for this row */
2086+ int iRank; /* Rank for this row */
2087+ int iDistance; /* Distance from pattern for this row */
2088+ int iScore; /* Score for sorting */
2089+ int iMatchlen; /* Value of matchlen column (or -1) */
2090+ char zHash[SPELLFIX_MX_HASH]; /* the phonehash used for this match */
2091+ } *a;
2092+};
2093+
2094+/*
2095+** Construct one or more SQL statements from the format string given
2096+** and then evaluate those statements. The success code is written
2097+** into *pRc.
2098+**
2099+** If *pRc is initially non-zero then this routine is a no-op.
2100+*/
2101+static void spellfix1DbExec(
2102+ int *pRc, /* Success code */
2103+ sqlite3 *db, /* Database in which to run SQL */
2104+ const char *zFormat, /* Format string for SQL */
2105+ ... /* Arguments to the format string */
2106+){
2107+ va_list ap;
2108+ char *zSql;
2109+ if( *pRc ) return;
2110+ va_start(ap, zFormat);
2111+ zSql = sqlite3_vmprintf(zFormat, ap);
2112+ va_end(ap);
2113+ if( zSql==0 ){
2114+ *pRc = SQLITE_NOMEM;
2115+ }else{
2116+ *pRc = sqlite3_exec(db, zSql, 0, 0, 0);
2117+ sqlite3_free(zSql);
2118+ }
2119+}
2120+
2121+/*
2122+** xDisconnect/xDestroy method for the fuzzy-search module.
2123+*/
2124+static int spellfix1Uninit(int isDestroy, sqlite3_vtab *pVTab){
2125+ spellfix1_vtab *p = (spellfix1_vtab*)pVTab;
2126+ int rc = SQLITE_OK;
2127+ if( isDestroy ){
2128+ sqlite3 *db = p->db;
2129+ spellfix1DbExec(&rc, db, "DROP TABLE IF EXISTS \"%w\".\"%w_vocab\"",
2130+ p->zDbName, p->zTableName);
2131+ }
2132+ if( rc==SQLITE_OK ){
2133+ sqlite3_free(p->zTableName);
2134+ editDist3ConfigDelete(p->pConfig3);
2135+ sqlite3_free(p->zCostTable);
2136+ sqlite3_free(p);
2137+ }
2138+ return rc;
2139+}
2140+static int spellfix1Disconnect(sqlite3_vtab *pVTab){
2141+ return spellfix1Uninit(0, pVTab);
2142+}
2143+static int spellfix1Destroy(sqlite3_vtab *pVTab){
2144+ return spellfix1Uninit(1, pVTab);
2145+}
2146+
2147+/*
2148+** Make a copy of a string. Remove leading and trailing whitespace
2149+** and dequote it.
2150+*/
2151+static char *spellfix1Dequote(const char *zIn){
2152+ char *zOut;
2153+ int i, j;
2154+ char c;
2155+ while( isspace((unsigned char)zIn[0]) ) zIn++;
2156+ zOut = sqlite3_mprintf("%s", zIn);
2157+ if( zOut==0 ) return 0;
2158+ i = (int)strlen(zOut);
2159+#if 0 /* The parser will never leave spaces at the end */
2160+ while( i>0 && isspace(zOut[i-1]) ){ i--; }
2161+#endif
2162+ zOut[i] = 0;
2163+ c = zOut[0];
2164+ if( c=='\'' || c=='"' ){
2165+ for(i=1, j=0; ALWAYS(zOut[i]); i++){
2166+ zOut[j++] = zOut[i];
2167+ if( zOut[i]==c ){
2168+ if( zOut[i+1]==c ){
2169+ i++;
2170+ }else{
2171+ zOut[j-1] = 0;
2172+ break;
2173+ }
2174+ }
2175+ }
2176+ }
2177+ return zOut;
2178+}
2179+
2180+
2181+/*
2182+** xConnect/xCreate method for the spellfix1 module. Arguments are:
2183+**
2184+** argv[0] -> module name ("spellfix1")
2185+** argv[1] -> database name
2186+** argv[2] -> table name
2187+** argv[3].. -> optional arguments (i.e. "edit_cost_table" parameter)
2188+*/
2189+static int spellfix1Init(
2190+ int isCreate,
2191+ sqlite3 *db,
2192+ void *pAux,
2193+ int argc, const char *const*argv,
2194+ sqlite3_vtab **ppVTab,
2195+ char **pzErr
2196+){
2197+ spellfix1_vtab *pNew = 0;
2198+ /* const char *zModule = argv[0]; // not used */
2199+ const char *zDbName = argv[1];
2200+ const char *zTableName = argv[2];
2201+ int nDbName;
2202+ int rc = SQLITE_OK;
2203+ int i;
2204+
2205+ nDbName = (int)strlen(zDbName);
2206+ pNew = sqlite3_malloc64( sizeof(*pNew) + nDbName + 1);
2207+ if( pNew==0 ){
2208+ rc = SQLITE_NOMEM;
2209+ }else{
2210+ memset(pNew, 0, sizeof(*pNew));
2211+ pNew->zDbName = (char*)&pNew[1];
2212+ memcpy(pNew->zDbName, zDbName, nDbName+1);
2213+ pNew->zTableName = sqlite3_mprintf("%s", zTableName);
2214+ pNew->db = db;
2215+ if( pNew->zTableName==0 ){
2216+ rc = SQLITE_NOMEM;
2217+ }else{
2218+ sqlite3_vtab_config(db, SQLITE_VTAB_INNOCUOUS);
2219+ rc = sqlite3_declare_vtab(db,
2220+ "CREATE TABLE x(word,rank,distance,langid, "
2221+ "score, matchlen, phonehash HIDDEN, "
2222+ "top HIDDEN, scope HIDDEN, srchcnt HIDDEN, "
2223+ "soundslike HIDDEN, command HIDDEN)"
2224+ );
2225+#define SPELLFIX_COL_WORD 0
2226+#define SPELLFIX_COL_RANK 1
2227+#define SPELLFIX_COL_DISTANCE 2
2228+#define SPELLFIX_COL_LANGID 3
2229+#define SPELLFIX_COL_SCORE 4
2230+#define SPELLFIX_COL_MATCHLEN 5
2231+#define SPELLFIX_COL_PHONEHASH 6
2232+#define SPELLFIX_COL_TOP 7
2233+#define SPELLFIX_COL_SCOPE 8
2234+#define SPELLFIX_COL_SRCHCNT 9
2235+#define SPELLFIX_COL_SOUNDSLIKE 10
2236+#define SPELLFIX_COL_COMMAND 11
2237+ }
2238+ if( rc==SQLITE_OK && isCreate ){
2239+ spellfix1DbExec(&rc, db,
2240+ "CREATE TABLE IF NOT EXISTS \"%w\".\"%w_vocab\"(\n"
2241+ " id INTEGER PRIMARY KEY,\n"
2242+ " rank INT,\n"
2243+ " langid INT,\n"
2244+ " word TEXT,\n"
2245+ " k1 TEXT,\n"
2246+ " k2 TEXT\n"
2247+ ");\n",
2248+ zDbName, zTableName
2249+ );
2250+ spellfix1DbExec(&rc, db,
2251+ "CREATE INDEX IF NOT EXISTS \"%w\".\"%w_vocab_index_langid_k2\" "
2252+ "ON \"%w_vocab\"(langid,k2);",
2253+ zDbName, zTableName, zTableName
2254+ );
2255+ }
2256+ for(i=3; rc==SQLITE_OK && i<argc; i++){
2257+ if( strncmp(argv[i],"edit_cost_table=",16)==0 && pNew->zCostTable==0 ){
2258+ pNew->zCostTable = spellfix1Dequote(&argv[i][16]);
2259+ if( pNew->zCostTable==0 ) rc = SQLITE_NOMEM;
2260+ continue;
2261+ }
2262+ *pzErr = sqlite3_mprintf("bad argument to spellfix1(): \"%s\"", argv[i]);
2263+ rc = SQLITE_ERROR;
2264+ }
2265+ }
2266+
2267+ if( rc && pNew ){
2268+ *ppVTab = 0;
2269+ spellfix1Uninit(0, &pNew->base);
2270+ }else{
2271+ *ppVTab = (sqlite3_vtab *)pNew;
2272+ }
2273+ return rc;
2274+}
2275+
2276+/*
2277+** The xConnect and xCreate methods
2278+*/
2279+static int spellfix1Connect(
2280+ sqlite3 *db,
2281+ void *pAux,
2282+ int argc, const char *const*argv,
2283+ sqlite3_vtab **ppVTab,
2284+ char **pzErr
2285+){
2286+ return spellfix1Init(0, db, pAux, argc, argv, ppVTab, pzErr);
2287+}
2288+static int spellfix1Create(
2289+ sqlite3 *db,
2290+ void *pAux,
2291+ int argc, const char *const*argv,
2292+ sqlite3_vtab **ppVTab,
2293+ char **pzErr
2294+){
2295+ return spellfix1Init(1, db, pAux, argc, argv, ppVTab, pzErr);
2296+}
2297+
2298+/*
2299+** Clear all of the content from a cursor.
2300+*/
2301+static void spellfix1ResetCursor(spellfix1_cursor *pCur){
2302+ int i;
2303+ for(i=0; i<pCur->nRow; i++){
2304+ sqlite3_free(pCur->a[i].zWord);
2305+ }
2306+ pCur->nRow = 0;
2307+ pCur->iRow = 0;
2308+ pCur->nSearch = 0;
2309+ if( pCur->pFullScan ){
2310+ sqlite3_finalize(pCur->pFullScan);
2311+ pCur->pFullScan = 0;
2312+ }
2313+}
2314+
2315+/*
2316+** Resize the cursor to hold up to N rows of content
2317+*/
2318+static void spellfix1ResizeCursor(spellfix1_cursor *pCur, int N){
2319+ struct spellfix1_row *aNew;
2320+ assert( N>=pCur->nRow );
2321+ aNew = sqlite3_realloc64(pCur->a, sizeof(pCur->a[0])*N);
2322+ if( aNew==0 && N>0 ){
2323+ spellfix1ResetCursor(pCur);
2324+ sqlite3_free(pCur->a);
2325+ pCur->nAlloc = 0;
2326+ pCur->a = 0;
2327+ }else{
2328+ pCur->nAlloc = N;
2329+ pCur->a = aNew;
2330+ }
2331+}
2332+
2333+
2334+/*
2335+** Close a fuzzy-search cursor.
2336+*/
2337+static int spellfix1Close(sqlite3_vtab_cursor *cur){
2338+ spellfix1_cursor *pCur = (spellfix1_cursor *)cur;
2339+ spellfix1ResetCursor(pCur);
2340+ spellfix1ResizeCursor(pCur, 0);
2341+ sqlite3_free(pCur->zPattern);
2342+ sqlite3_free(pCur);
2343+ return SQLITE_OK;
2344+}
2345+
2346+#define SPELLFIX_IDXNUM_MATCH 0x01 /* word MATCH $str */
2347+#define SPELLFIX_IDXNUM_LANGID 0x02 /* langid == $langid */
2348+#define SPELLFIX_IDXNUM_TOP 0x04 /* top = $top */
2349+#define SPELLFIX_IDXNUM_SCOPE 0x08 /* scope = $scope */
2350+#define SPELLFIX_IDXNUM_DISTLT 0x10 /* distance < $distance */
2351+#define SPELLFIX_IDXNUM_DISTLE 0x20 /* distance <= $distance */
2352+#define SPELLFIX_IDXNUM_ROWID 0x40 /* rowid = $rowid */
2353+#define SPELLFIX_IDXNUM_DIST (0x10|0x20) /* DISTLT and DISTLE */
2354+
2355+/*
2356+**
2357+** The plan number is a bitmask of the SPELLFIX_IDXNUM_* values defined
2358+** above.
2359+**
2360+** filter.argv[*] values contains $str, $langid, $top, $scope and $rowid
2361+** if specified and in that order.
2362+*/
2363+static int spellfix1BestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){
2364+ int iPlan = 0;
2365+ int iLangTerm = -1;
2366+ int iTopTerm = -1;
2367+ int iScopeTerm = -1;
2368+ int iDistTerm = -1;
2369+ int iRowidTerm = -1;
2370+ int i;
2371+ const struct sqlite3_index_constraint *pConstraint;
2372+ pConstraint = pIdxInfo->aConstraint;
2373+ for(i=0; i<pIdxInfo->nConstraint; i++, pConstraint++){
2374+ if( pConstraint->usable==0 ) continue;
2375+
2376+ /* Terms of the form: word MATCH $str */
2377+ if( (iPlan & SPELLFIX_IDXNUM_MATCH)==0
2378+ && pConstraint->iColumn==SPELLFIX_COL_WORD
2379+ && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH
2380+ ){
2381+ iPlan |= SPELLFIX_IDXNUM_MATCH;
2382+ pIdxInfo->aConstraintUsage[i].argvIndex = 1;
2383+ pIdxInfo->aConstraintUsage[i].omit = 1;
2384+ }
2385+
2386+ /* Terms of the form: langid = $langid */
2387+ if( (iPlan & SPELLFIX_IDXNUM_LANGID)==0
2388+ && pConstraint->iColumn==SPELLFIX_COL_LANGID
2389+ && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ
2390+ ){
2391+ iPlan |= SPELLFIX_IDXNUM_LANGID;
2392+ iLangTerm = i;
2393+ }
2394+
2395+ /* Terms of the form: top = $top */
2396+ if( (iPlan & SPELLFIX_IDXNUM_TOP)==0
2397+ && pConstraint->iColumn==SPELLFIX_COL_TOP
2398+ && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ
2399+ ){
2400+ iPlan |= SPELLFIX_IDXNUM_TOP;
2401+ iTopTerm = i;
2402+ }
2403+
2404+ /* Terms of the form: scope = $scope */
2405+ if( (iPlan & SPELLFIX_IDXNUM_SCOPE)==0
2406+ && pConstraint->iColumn==SPELLFIX_COL_SCOPE
2407+ && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ
2408+ ){
2409+ iPlan |= SPELLFIX_IDXNUM_SCOPE;
2410+ iScopeTerm = i;
2411+ }
2412+
2413+ /* Terms of the form: distance < $dist or distance <= $dist */
2414+ if( (iPlan & SPELLFIX_IDXNUM_DIST)==0
2415+ && pConstraint->iColumn==SPELLFIX_COL_DISTANCE
2416+ && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT
2417+ || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE)
2418+ ){
2419+ if( pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ){
2420+ iPlan |= SPELLFIX_IDXNUM_DISTLT;
2421+ }else{
2422+ iPlan |= SPELLFIX_IDXNUM_DISTLE;
2423+ }
2424+ iDistTerm = i;
2425+ }
2426+
2427+ /* Terms of the form: distance < $dist or distance <= $dist */
2428+ if( (iPlan & SPELLFIX_IDXNUM_ROWID)==0
2429+ && pConstraint->iColumn<0
2430+ && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ
2431+ ){
2432+ iPlan |= SPELLFIX_IDXNUM_ROWID;
2433+ iRowidTerm = i;
2434+ }
2435+ }
2436+ if( iPlan&SPELLFIX_IDXNUM_MATCH ){
2437+ int idx = 2;
2438+ pIdxInfo->idxNum = iPlan;
2439+ if( pIdxInfo->nOrderBy==1
2440+ && pIdxInfo->aOrderBy[0].iColumn==SPELLFIX_COL_SCORE
2441+ && pIdxInfo->aOrderBy[0].desc==0
2442+ ){
2443+ pIdxInfo->orderByConsumed = 1; /* Default order by iScore */
2444+ }
2445+ if( iPlan&SPELLFIX_IDXNUM_LANGID ){
2446+ pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx++;
2447+ pIdxInfo->aConstraintUsage[iLangTerm].omit = 1;
2448+ }
2449+ if( iPlan&SPELLFIX_IDXNUM_TOP ){
2450+ pIdxInfo->aConstraintUsage[iTopTerm].argvIndex = idx++;
2451+ pIdxInfo->aConstraintUsage[iTopTerm].omit = 1;
2452+ }
2453+ if( iPlan&SPELLFIX_IDXNUM_SCOPE ){
2454+ pIdxInfo->aConstraintUsage[iScopeTerm].argvIndex = idx++;
2455+ pIdxInfo->aConstraintUsage[iScopeTerm].omit = 1;
2456+ }
2457+ if( iPlan&SPELLFIX_IDXNUM_DIST ){
2458+ pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = idx++;
2459+ pIdxInfo->aConstraintUsage[iDistTerm].omit = 1;
2460+ }
2461+ pIdxInfo->estimatedCost = 1e5;
2462+ }else if( (iPlan & SPELLFIX_IDXNUM_ROWID) ){
2463+ pIdxInfo->idxNum = SPELLFIX_IDXNUM_ROWID;
2464+ pIdxInfo->aConstraintUsage[iRowidTerm].argvIndex = 1;
2465+ pIdxInfo->aConstraintUsage[iRowidTerm].omit = 1;
2466+ pIdxInfo->estimatedCost = 5;
2467+ }else{
2468+ pIdxInfo->idxNum = 0;
2469+ pIdxInfo->estimatedCost = 1e50;
2470+ }
2471+ return SQLITE_OK;
2472+}
2473+
2474+/*
2475+** Open a new fuzzy-search cursor.
2476+*/
2477+static int spellfix1Open(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){
2478+ spellfix1_vtab *p = (spellfix1_vtab*)pVTab;
2479+ spellfix1_cursor *pCur;
2480+ pCur = sqlite3_malloc64( sizeof(*pCur) );
2481+ if( pCur==0 ) return SQLITE_NOMEM;
2482+ memset(pCur, 0, sizeof(*pCur));
2483+ pCur->pVTab = p;
2484+ *ppCursor = &pCur->base;
2485+ return SQLITE_OK;
2486+}
2487+
2488+/*
2489+** Adjust a distance measurement by the words rank in order to show
2490+** preference to common words.
2491+*/
2492+static int spellfix1Score(int iDistance, int iRank){
2493+ int iLog2;
2494+ for(iLog2=0; iRank>0; iLog2++, iRank>>=1){}
2495+ return iDistance + 32 - iLog2;
2496+}
2497+
2498+/*
2499+** Compare two spellfix1_row objects for sorting purposes in qsort() such
2500+** that they sort in order of increasing distance.
2501+*/
2502+static int SQLITE_CDECL spellfix1RowCompare(const void *A, const void *B){
2503+ const struct spellfix1_row *a = (const struct spellfix1_row*)A;
2504+ const struct spellfix1_row *b = (const struct spellfix1_row*)B;
2505+ return a->iScore - b->iScore;
2506+}
2507+
2508+/*
2509+** A structure used to pass information from spellfix1FilterForMatch()
2510+** into spellfix1RunQuery().
2511+*/
2512+typedef struct MatchQuery {
2513+ spellfix1_cursor *pCur; /* The cursor being queried */
2514+ sqlite3_stmt *pStmt; /* shadow table query statment */
2515+ char zHash[SPELLFIX_MX_HASH]; /* The current phonehash for zPattern */
2516+ const char *zPattern; /* Transliterated input string */
2517+ int nPattern; /* Length of zPattern */
2518+ EditDist3FromString *pMatchStr3; /* Original unicode string */
2519+ EditDist3Config *pConfig3; /* Edit-distance cost coefficients */
2520+ const EditDist3Lang *pLang; /* The selected language coefficients */
2521+ int iLang; /* The language id */
2522+ int iScope; /* Default scope */
2523+ int iMaxDist; /* Maximum allowed edit distance, or -1 */
2524+ int rc; /* Error code */
2525+ int nRun; /* Number of prior runs for the same zPattern */
2526+ char azPrior[SPELLFIX_MX_RUN][SPELLFIX_MX_HASH]; /* Prior hashes */
2527+} MatchQuery;
2528+
2529+/*
2530+** Run a query looking for the best matches against zPattern using
2531+** zHash as the character class seed hash.
2532+*/
2533+static void spellfix1RunQuery(MatchQuery *p, const char *zQuery, int nQuery){
2534+ const char *zK1;
2535+ const char *zWord;
2536+ int iDist;
2537+ int iRank;
2538+ int iScore;
2539+ int iWorst = 0;
2540+ int idx;
2541+ int idxWorst = -1;
2542+ int i;
2543+ int iScope = p->iScope;
2544+ spellfix1_cursor *pCur = p->pCur;
2545+ sqlite3_stmt *pStmt = p->pStmt;
2546+ char zHash1[SPELLFIX_MX_HASH];
2547+ char zHash2[SPELLFIX_MX_HASH];
2548+ char *zClass;
2549+ int nClass;
2550+ int rc;
2551+
2552+ if( pCur->a==0 || p->rc ) return; /* Prior memory allocation failure */
2553+ zClass = (char*)phoneticHash((unsigned char*)zQuery, nQuery);
2554+ if( zClass==0 ){
2555+ p->rc = SQLITE_NOMEM;
2556+ return;
2557+ }
2558+ nClass = (int)strlen(zClass);
2559+ if( nClass>SPELLFIX_MX_HASH-2 ){
2560+ nClass = SPELLFIX_MX_HASH-2;
2561+ zClass[nClass] = 0;
2562+ }
2563+ if( nClass<=iScope ){
2564+ if( nClass>2 ){
2565+ iScope = nClass-1;
2566+ }else{
2567+ iScope = nClass;
2568+ }
2569+ }
2570+ memcpy(zHash1, zClass, iScope);
2571+ sqlite3_free(zClass);
2572+ zHash1[iScope] = 0;
2573+ memcpy(zHash2, zHash1, iScope);
2574+ zHash2[iScope] = 'Z';
2575+ zHash2[iScope+1] = 0;
2576+#if SPELLFIX_MX_RUN>1
2577+ for(i=0; i<p->nRun; i++){
2578+ if( strcmp(p->azPrior[i], zHash1)==0 ) return;
2579+ }
2580+#endif
2581+ assert( p->nRun<SPELLFIX_MX_RUN );
2582+ memcpy(p->azPrior[p->nRun++], zHash1, iScope+1);
2583+ if( sqlite3_bind_text(pStmt, 1, zHash1, -1, SQLITE_STATIC)==SQLITE_NOMEM
2584+ || sqlite3_bind_text(pStmt, 2, zHash2, -1, SQLITE_STATIC)==SQLITE_NOMEM
2585+ ){
2586+ p->rc = SQLITE_NOMEM;
2587+ return;
2588+ }
2589+#if SPELLFIX_MX_RUN>1
2590+ for(i=0; i<pCur->nRow; i++){
2591+ if( pCur->a[i].iScore>iWorst ){
2592+ iWorst = pCur->a[i].iScore;
2593+ idxWorst = i;
2594+ }
2595+ }
2596+#endif
2597+ while( sqlite3_step(pStmt)==SQLITE_ROW ){
2598+ int iMatchlen = -1;
2599+ iRank = sqlite3_column_int(pStmt, 2);
2600+ if( p->pMatchStr3 ){
2601+ int nWord = sqlite3_column_bytes(pStmt, 1);
2602+ zWord = (const char*)sqlite3_column_text(pStmt, 1);
2603+ iDist = editDist3Core(p->pMatchStr3, zWord, nWord, p->pLang, &iMatchlen);
2604+ }else{
2605+ zK1 = (const char*)sqlite3_column_text(pStmt, 3);
2606+ if( zK1==0 ) continue;
2607+ iDist = editdist1(p->zPattern, zK1, 0);
2608+ }
2609+ if( iDist<0 ){
2610+ p->rc = SQLITE_NOMEM;
2611+ break;
2612+ }
2613+ pCur->nSearch++;
2614+
2615+ /* If there is a "distance < $dist" or "distance <= $dist" constraint,
2616+ ** check if this row meets it. If not, jump back up to the top of the
2617+ ** loop to process the next row. Otherwise, if the row does match the
2618+ ** distance constraint, check if the pCur->a[] array is already full.
2619+ ** If it is and no explicit "top = ?" constraint was present in the
2620+ ** query, grow the array to ensure there is room for the new entry. */
2621+ assert( (p->iMaxDist>=0)==((pCur->idxNum & SPELLFIX_IDXNUM_DIST) ? 1 : 0) );
2622+ if( p->iMaxDist>=0 ){
2623+ if( iDist>p->iMaxDist ) continue;
2624+ if( pCur->nRow>=pCur->nAlloc && (pCur->idxNum & SPELLFIX_IDXNUM_TOP)==0 ){
2625+ spellfix1ResizeCursor(pCur, pCur->nAlloc*2 + 10);
2626+ if( pCur->a==0 ) break;
2627+ }
2628+ }
2629+
2630+ iScore = spellfix1Score(iDist,iRank);
2631+ if( pCur->nRow<pCur->nAlloc ){
2632+ idx = pCur->nRow;
2633+ }else if( iScore<iWorst ){
2634+ idx = idxWorst;
2635+ sqlite3_free(pCur->a[idx].zWord);
2636+ }else{
2637+ continue;
2638+ }
2639+
2640+ pCur->a[idx].zWord = sqlite3_mprintf("%s", sqlite3_column_text(pStmt, 1));
2641+ if( pCur->a[idx].zWord==0 ){
2642+ p->rc = SQLITE_NOMEM;
2643+ break;
2644+ }
2645+ pCur->a[idx].iRowid = sqlite3_column_int64(pStmt, 0);
2646+ pCur->a[idx].iRank = iRank;
2647+ pCur->a[idx].iDistance = iDist;
2648+ pCur->a[idx].iScore = iScore;
2649+ pCur->a[idx].iMatchlen = iMatchlen;
2650+ memcpy(pCur->a[idx].zHash, zHash1, iScope+1);
2651+ if( pCur->nRow<pCur->nAlloc ) pCur->nRow++;
2652+ if( pCur->nRow==pCur->nAlloc ){
2653+ iWorst = pCur->a[0].iScore;
2654+ idxWorst = 0;
2655+ for(i=1; i<pCur->nRow; i++){
2656+ iScore = pCur->a[i].iScore;
2657+ if( iWorst<iScore ){
2658+ iWorst = iScore;
2659+ idxWorst = i;
2660+ }
2661+ }
2662+ }
2663+ }
2664+ rc = sqlite3_reset(pStmt);
2665+ if( rc ) p->rc = rc;
2666+}
2667+
2668+/*
2669+** This version of the xFilter method work if the MATCH term is present
2670+** and we are doing a scan.
2671+*/
2672+static int spellfix1FilterForMatch(
2673+ spellfix1_cursor *pCur,
2674+ int argc,
2675+ sqlite3_value **argv
2676+){
2677+ int idxNum = pCur->idxNum;
2678+ const unsigned char *zMatchThis; /* RHS of the MATCH operator */
2679+ EditDist3FromString *pMatchStr3 = 0; /* zMatchThis as an editdist string */
2680+ char *zPattern; /* Transliteration of zMatchThis */
2681+ int nPattern; /* Length of zPattern */
2682+ int iLimit = 20; /* Max number of rows of output */
2683+ int iScope = 3; /* Use this many characters of zClass */
2684+ int iLang = 0; /* Language code */
2685+ char *zSql; /* SQL of shadow table query */
2686+ sqlite3_stmt *pStmt = 0; /* Shadow table query */
2687+ int rc; /* Result code */
2688+ int idx = 1; /* Next available filter parameter */
2689+ spellfix1_vtab *p = pCur->pVTab; /* The virtual table that owns pCur */
2690+ MatchQuery x; /* For passing info to RunQuery() */
2691+
2692+ /* Load the cost table if we have not already done so */
2693+ if( p->zCostTable!=0 && p->pConfig3==0 ){
2694+ p->pConfig3 = sqlite3_malloc64( sizeof(p->pConfig3[0]) );
2695+ if( p->pConfig3==0 ) return SQLITE_NOMEM;
2696+ memset(p->pConfig3, 0, sizeof(p->pConfig3[0]));
2697+ rc = editDist3ConfigLoad(p->pConfig3, p->db, p->zCostTable);
2698+ if( rc ) return rc;
2699+ }
2700+ memset(&x, 0, sizeof(x));
2701+ x.iScope = 3; /* Default scope if none specified by "WHERE scope=N" */
2702+ x.iMaxDist = -1; /* Maximum allowed edit distance */
2703+
2704+ if( idxNum&2 ){
2705+ iLang = sqlite3_value_int(argv[idx++]);
2706+ }
2707+ if( idxNum&4 ){
2708+ iLimit = sqlite3_value_int(argv[idx++]);
2709+ if( iLimit<1 ) iLimit = 1;
2710+ }
2711+ if( idxNum&8 ){
2712+ x.iScope = sqlite3_value_int(argv[idx++]);
2713+ if( x.iScope<1 ) x.iScope = 1;
2714+ if( x.iScope>SPELLFIX_MX_HASH-2 ) x.iScope = SPELLFIX_MX_HASH-2;
2715+ }
2716+ if( idxNum&(16|32) ){
2717+ x.iMaxDist = sqlite3_value_int(argv[idx++]);
2718+ if( idxNum&16 ) x.iMaxDist--;
2719+ if( x.iMaxDist<0 ) x.iMaxDist = 0;
2720+ }
2721+ spellfix1ResetCursor(pCur);
2722+ spellfix1ResizeCursor(pCur, iLimit);
2723+ zMatchThis = sqlite3_value_text(argv[0]);
2724+ if( zMatchThis==0 ) return SQLITE_OK;
2725+ if( p->pConfig3 ){
2726+ x.pLang = editDist3FindLang(p->pConfig3, iLang);
2727+ pMatchStr3 = editDist3FromStringNew(x.pLang, (const char*)zMatchThis, -1);
2728+ if( pMatchStr3==0 ){
2729+ x.rc = SQLITE_NOMEM;
2730+ goto filter_exit;
2731+ }
2732+ }else{
2733+ x.pLang = 0;
2734+ }
2735+ zPattern = (char*)transliterate(zMatchThis, sqlite3_value_bytes(argv[0]));
2736+ sqlite3_free(pCur->zPattern);
2737+ pCur->zPattern = zPattern;
2738+ if( zPattern==0 ){
2739+ x.rc = SQLITE_NOMEM;
2740+ goto filter_exit;
2741+ }
2742+ nPattern = (int)strlen(zPattern);
2743+ if( zPattern[nPattern-1]=='*' ) nPattern--;
2744+ zSql = sqlite3_mprintf(
2745+ "SELECT id, word, rank, coalesce(k1,word)"
2746+ " FROM \"%w\".\"%w_vocab\""
2747+ " WHERE langid=%d AND k2>=?1 AND k2<?2",
2748+ p->zDbName, p->zTableName, iLang
2749+ );
2750+ if( zSql==0 ){
2751+ x.rc = SQLITE_NOMEM;
2752+ pStmt = 0;
2753+ goto filter_exit;
2754+ }
2755+ rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0);
2756+ sqlite3_free(zSql);
2757+ pCur->iLang = iLang;
2758+ x.pCur = pCur;
2759+ x.pStmt = pStmt;
2760+ x.zPattern = zPattern;
2761+ x.nPattern = nPattern;
2762+ x.pMatchStr3 = pMatchStr3;
2763+ x.iLang = iLang;
2764+ x.rc = rc;
2765+ x.pConfig3 = p->pConfig3;
2766+ if( x.rc==SQLITE_OK ){
2767+ spellfix1RunQuery(&x, zPattern, nPattern);
2768+ }
2769+
2770+ if( pCur->a ){
2771+ qsort(pCur->a, pCur->nRow, sizeof(pCur->a[0]), spellfix1RowCompare);
2772+ pCur->iTop = iLimit;
2773+ pCur->iScope = iScope;
2774+ }else{
2775+ x.rc = SQLITE_NOMEM;
2776+ }
2777+
2778+filter_exit:
2779+ sqlite3_finalize(pStmt);
2780+ editDist3FromStringDelete(pMatchStr3);
2781+ return x.rc;
2782+}
2783+
2784+/*
2785+** This version of xFilter handles a full-table scan case
2786+*/
2787+static int spellfix1FilterForFullScan(
2788+ spellfix1_cursor *pCur,
2789+ int argc,
2790+ sqlite3_value **argv
2791+){
2792+ int rc = SQLITE_OK;
2793+ int idxNum = pCur->idxNum;
2794+ char *zSql;
2795+ spellfix1_vtab *pVTab = pCur->pVTab;
2796+ spellfix1ResetCursor(pCur);
2797+ assert( idxNum==0 || idxNum==64 );
2798+ zSql = sqlite3_mprintf(
2799+ "SELECT word, rank, NULL, langid, id FROM \"%w\".\"%w_vocab\"%s",
2800+ pVTab->zDbName, pVTab->zTableName,
2801+ ((idxNum & 64) ? " WHERE rowid=?" : "")
2802+ );
2803+ if( zSql==0 ) return SQLITE_NOMEM;
2804+ rc = sqlite3_prepare_v2(pVTab->db, zSql, -1, &pCur->pFullScan, 0);
2805+ sqlite3_free(zSql);
2806+ if( rc==SQLITE_OK && (idxNum & 64) ){
2807+ assert( argc==1 );
2808+ rc = sqlite3_bind_value(pCur->pFullScan, 1, argv[0]);
2809+ }
2810+ pCur->nRow = pCur->iRow = 0;
2811+ if( rc==SQLITE_OK ){
2812+ rc = sqlite3_step(pCur->pFullScan);
2813+ if( rc==SQLITE_ROW ){ pCur->iRow = -1; rc = SQLITE_OK; }
2814+ if( rc==SQLITE_DONE ){ rc = SQLITE_OK; }
2815+ }else{
2816+ pCur->iRow = 0;
2817+ }
2818+ return rc;
2819+}
2820+
2821+
2822+/*
2823+** Called to "rewind" a cursor back to the beginning so that
2824+** it starts its output over again. Always called at least once
2825+** prior to any spellfix1Column, spellfix1Rowid, or spellfix1Eof call.
2826+*/
2827+static int spellfix1Filter(
2828+ sqlite3_vtab_cursor *cur,
2829+ int idxNum, const char *idxStr,
2830+ int argc, sqlite3_value **argv
2831+){
2832+ spellfix1_cursor *pCur = (spellfix1_cursor *)cur;
2833+ int rc;
2834+ pCur->idxNum = idxNum;
2835+ if( idxNum & 1 ){
2836+ rc = spellfix1FilterForMatch(pCur, argc, argv);
2837+ }else{
2838+ rc = spellfix1FilterForFullScan(pCur, argc, argv);
2839+ }
2840+ return rc;
2841+}
2842+
2843+
2844+/*
2845+** Advance a cursor to its next row of output
2846+*/
2847+static int spellfix1Next(sqlite3_vtab_cursor *cur){
2848+ spellfix1_cursor *pCur = (spellfix1_cursor *)cur;
2849+ int rc = SQLITE_OK;
2850+ if( pCur->iRow < pCur->nRow ){
2851+ if( pCur->pFullScan ){
2852+ rc = sqlite3_step(pCur->pFullScan);
2853+ if( rc!=SQLITE_ROW ) pCur->iRow = pCur->nRow;
2854+ if( rc==SQLITE_ROW || rc==SQLITE_DONE ) rc = SQLITE_OK;
2855+ }else{
2856+ pCur->iRow++;
2857+ }
2858+ }
2859+ return rc;
2860+}
2861+
2862+/*
2863+** Return TRUE if we are at the end-of-file
2864+*/
2865+static int spellfix1Eof(sqlite3_vtab_cursor *cur){
2866+ spellfix1_cursor *pCur = (spellfix1_cursor *)cur;
2867+ return pCur->iRow>=pCur->nRow;
2868+}
2869+
2870+/*
2871+** Return columns from the current row.
2872+*/
2873+static int spellfix1Column(
2874+ sqlite3_vtab_cursor *cur,
2875+ sqlite3_context *ctx,
2876+ int i
2877+){
2878+ spellfix1_cursor *pCur = (spellfix1_cursor*)cur;
2879+ if( pCur->pFullScan ){
2880+ if( i<=SPELLFIX_COL_LANGID ){
2881+ sqlite3_result_value(ctx, sqlite3_column_value(pCur->pFullScan, i));
2882+ }else{
2883+ sqlite3_result_null(ctx);
2884+ }
2885+ return SQLITE_OK;
2886+ }
2887+ switch( i ){
2888+ case SPELLFIX_COL_WORD: {
2889+ sqlite3_result_text(ctx, pCur->a[pCur->iRow].zWord, -1, SQLITE_STATIC);
2890+ break;
2891+ }
2892+ case SPELLFIX_COL_RANK: {
2893+ sqlite3_result_int(ctx, pCur->a[pCur->iRow].iRank);
2894+ break;
2895+ }
2896+ case SPELLFIX_COL_DISTANCE: {
2897+ sqlite3_result_int(ctx, pCur->a[pCur->iRow].iDistance);
2898+ break;
2899+ }
2900+ case SPELLFIX_COL_LANGID: {
2901+ sqlite3_result_int(ctx, pCur->iLang);
2902+ break;
2903+ }
2904+ case SPELLFIX_COL_SCORE: {
2905+ sqlite3_result_int(ctx, pCur->a[pCur->iRow].iScore);
2906+ break;
2907+ }
2908+ case SPELLFIX_COL_MATCHLEN: {
2909+ int iMatchlen = pCur->a[pCur->iRow].iMatchlen;
2910+ if( iMatchlen<0 ){
2911+ int nPattern = (int)strlen(pCur->zPattern);
2912+ char *zWord = pCur->a[pCur->iRow].zWord;
2913+ int nWord = (int)strlen(zWord);
2914+
2915+ if( nPattern>0 && pCur->zPattern[nPattern-1]=='*' ){
2916+ char *zTranslit;
2917+ int res;
2918+ zTranslit = (char *)transliterate((unsigned char *)zWord, nWord);
2919+ if( !zTranslit ) return SQLITE_NOMEM;
2920+ res = editdist1(pCur->zPattern, zTranslit, &iMatchlen);
2921+ sqlite3_free(zTranslit);
2922+ if( res<0 ) return SQLITE_NOMEM;
2923+ iMatchlen = translen_to_charlen(zWord, nWord, iMatchlen);
2924+ }else{
2925+ iMatchlen = utf8Charlen(zWord, nWord);
2926+ }
2927+ }
2928+
2929+ sqlite3_result_int(ctx, iMatchlen);
2930+ break;
2931+ }
2932+ case SPELLFIX_COL_PHONEHASH: {
2933+ sqlite3_result_text(ctx, pCur->a[pCur->iRow].zHash, -1, SQLITE_STATIC);
2934+ break;
2935+ }
2936+ case SPELLFIX_COL_TOP: {
2937+ sqlite3_result_int(ctx, pCur->iTop);
2938+ break;
2939+ }
2940+ case SPELLFIX_COL_SCOPE: {
2941+ sqlite3_result_int(ctx, pCur->iScope);
2942+ break;
2943+ }
2944+ case SPELLFIX_COL_SRCHCNT: {
2945+ sqlite3_result_int(ctx, pCur->nSearch);
2946+ break;
2947+ }
2948+ default: {
2949+ sqlite3_result_null(ctx);
2950+ break;
2951+ }
2952+ }
2953+ return SQLITE_OK;
2954+}
2955+
2956+/*
2957+** The rowid.
2958+*/
2959+static int spellfix1Rowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){
2960+ spellfix1_cursor *pCur = (spellfix1_cursor*)cur;
2961+ if( pCur->pFullScan ){
2962+ *pRowid = sqlite3_column_int64(pCur->pFullScan, 4);
2963+ }else{
2964+ *pRowid = pCur->a[pCur->iRow].iRowid;
2965+ }
2966+ return SQLITE_OK;
2967+}
2968+
2969+/*
2970+** This function is called by the xUpdate() method. It returns a string
2971+** containing the conflict mode that xUpdate() should use for the current
2972+** operation. One of: "ROLLBACK", "IGNORE", "ABORT" or "REPLACE".
2973+*/
2974+static const char *spellfix1GetConflict(sqlite3 *db){
2975+ static const char *azConflict[] = {
2976+ /* Note: Instead of "FAIL" - "ABORT". */
2977+ "ROLLBACK", "IGNORE", "ABORT", "ABORT", "REPLACE"
2978+ };
2979+ int eConflict = sqlite3_vtab_on_conflict(db);
2980+
2981+ assert( eConflict==SQLITE_ROLLBACK || eConflict==SQLITE_IGNORE
2982+ || eConflict==SQLITE_FAIL || eConflict==SQLITE_ABORT
2983+ || eConflict==SQLITE_REPLACE
2984+ );
2985+ assert( SQLITE_ROLLBACK==1 );
2986+ assert( SQLITE_IGNORE==2 );
2987+ assert( SQLITE_FAIL==3 );
2988+ assert( SQLITE_ABORT==4 );
2989+ assert( SQLITE_REPLACE==5 );
2990+
2991+ return azConflict[eConflict-1];
2992+}
2993+
2994+/*
2995+** The xUpdate() method.
2996+*/
2997+static int spellfix1Update(
2998+ sqlite3_vtab *pVTab,
2999+ int argc,
3000+ sqlite3_value **argv,
3001+ sqlite_int64 *pRowid
3002+){
3003+ int rc = SQLITE_OK;
3004+ sqlite3_int64 rowid, newRowid;
3005+ spellfix1_vtab *p = (spellfix1_vtab*)pVTab;
3006+ sqlite3 *db = p->db;
3007+
3008+ if( argc==1 ){
3009+ /* A delete operation on the rowid given by argv[0] */
3010+ rowid = *pRowid = sqlite3_value_int64(argv[0]);
3011+ spellfix1DbExec(&rc, db, "DELETE FROM \"%w\".\"%w_vocab\" "
3012+ " WHERE id=%lld",
3013+ p->zDbName, p->zTableName, rowid);
3014+ }else{
3015+ const unsigned char *zWord = sqlite3_value_text(argv[SPELLFIX_COL_WORD+2]);
3016+ int nWord = sqlite3_value_bytes(argv[SPELLFIX_COL_WORD+2]);
3017+ int iLang = sqlite3_value_int(argv[SPELLFIX_COL_LANGID+2]);
3018+ int iRank = sqlite3_value_int(argv[SPELLFIX_COL_RANK+2]);
3019+ const unsigned char *zSoundslike =
3020+ sqlite3_value_text(argv[SPELLFIX_COL_SOUNDSLIKE+2]);
3021+ int nSoundslike = sqlite3_value_bytes(argv[SPELLFIX_COL_SOUNDSLIKE+2]);
3022+ char *zK1, *zK2;
3023+ int i;
3024+ char c;
3025+ const char *zConflict = spellfix1GetConflict(db);
3026+
3027+ if( zWord==0 ){
3028+ /* Inserts of the form: INSERT INTO table(command) VALUES('xyzzy');
3029+ ** cause zWord to be NULL, so we look at the "command" column to see
3030+ ** what special actions to take */
3031+ const char *zCmd =
3032+ (const char*)sqlite3_value_text(argv[SPELLFIX_COL_COMMAND+2]);
3033+ if( zCmd==0 ){
3034+ pVTab->zErrMsg = sqlite3_mprintf("NOT NULL constraint failed: %s.word",
3035+ p->zTableName);
3036+ return SQLITE_CONSTRAINT_NOTNULL;
3037+ }
3038+ if( strcmp(zCmd,"reset")==0 ){
3039+ /* Reset the edit cost table (if there is one). */
3040+ editDist3ConfigDelete(p->pConfig3);
3041+ p->pConfig3 = 0;
3042+ return SQLITE_OK;
3043+ }
3044+ if( strncmp(zCmd,"edit_cost_table=",16)==0 ){
3045+ editDist3ConfigDelete(p->pConfig3);
3046+ p->pConfig3 = 0;
3047+ sqlite3_free(p->zCostTable);
3048+ p->zCostTable = spellfix1Dequote(zCmd+16);
3049+ if( p->zCostTable==0 ) return SQLITE_NOMEM;
3050+ if( p->zCostTable[0]==0 || sqlite3_stricmp(p->zCostTable,"null")==0 ){
3051+ sqlite3_free(p->zCostTable);
3052+ p->zCostTable = 0;
3053+ }
3054+ return SQLITE_OK;
3055+ }
3056+ pVTab->zErrMsg = sqlite3_mprintf("unknown value for %s.command: \"%w\"",
3057+ p->zTableName, zCmd);
3058+ return SQLITE_ERROR;
3059+ }
3060+ if( iRank<1 ) iRank = 1;
3061+ if( zSoundslike ){
3062+ zK1 = (char*)transliterate(zSoundslike, nSoundslike);
3063+ }else{
3064+ zK1 = (char*)transliterate(zWord, nWord);
3065+ }
3066+ if( zK1==0 ) return SQLITE_NOMEM;
3067+ for(i=0; (c = zK1[i])!=0; i++){
3068+ if( c>='A' && c<='Z' ) zK1[i] += 'a' - 'A';
3069+ }
3070+ zK2 = (char*)phoneticHash((const unsigned char*)zK1, i);
3071+ if( zK2==0 ){
3072+ sqlite3_free(zK1);
3073+ return SQLITE_NOMEM;
3074+ }
3075+ if( sqlite3_value_type(argv[0])==SQLITE_NULL ){
3076+ if( sqlite3_value_type(argv[1])==SQLITE_NULL ){
3077+ spellfix1DbExec(&rc, db,
3078+ "INSERT INTO \"%w\".\"%w_vocab\"(rank,langid,word,k1,k2) "
3079+ "VALUES(%d,%d,%Q,nullif(%Q,%Q),%Q)",
3080+ p->zDbName, p->zTableName,
3081+ iRank, iLang, zWord, zK1, zWord, zK2
3082+ );
3083+ }else{
3084+ newRowid = sqlite3_value_int64(argv[1]);
3085+ spellfix1DbExec(&rc, db,
3086+ "INSERT OR %s INTO \"%w\".\"%w_vocab\"(id,rank,langid,word,k1,k2) "
3087+ "VALUES(%lld,%d,%d,%Q,nullif(%Q,%Q),%Q)",
3088+ zConflict, p->zDbName, p->zTableName,
3089+ newRowid, iRank, iLang, zWord, zK1, zWord, zK2
3090+ );
3091+ }
3092+ *pRowid = sqlite3_last_insert_rowid(db);
3093+ }else{
3094+ rowid = sqlite3_value_int64(argv[0]);
3095+ newRowid = *pRowid = sqlite3_value_int64(argv[1]);
3096+ spellfix1DbExec(&rc, db,
3097+ "UPDATE OR %s \"%w\".\"%w_vocab\" SET id=%lld, rank=%d, langid=%d,"
3098+ " word=%Q, k1=nullif(%Q,%Q), k2=%Q WHERE id=%lld",
3099+ zConflict, p->zDbName, p->zTableName, newRowid, iRank, iLang,
3100+ zWord, zK1, zWord, zK2, rowid
3101+ );
3102+ }
3103+ sqlite3_free(zK1);
3104+ sqlite3_free(zK2);
3105+ }
3106+ return rc;
3107+}
3108+
3109+/*
3110+** Rename the spellfix1 table.
3111+*/
3112+static int spellfix1Rename(sqlite3_vtab *pVTab, const char *zNew){
3113+ spellfix1_vtab *p = (spellfix1_vtab*)pVTab;
3114+ sqlite3 *db = p->db;
3115+ int rc = SQLITE_OK;
3116+ char *zNewName = sqlite3_mprintf("%s", zNew);
3117+ if( zNewName==0 ){
3118+ return SQLITE_NOMEM;
3119+ }
3120+ spellfix1DbExec(&rc, db,
3121+ "ALTER TABLE \"%w\".\"%w_vocab\" RENAME TO \"%w_vocab\"",
3122+ p->zDbName, p->zTableName, zNewName
3123+ );
3124+ if( rc==SQLITE_OK ){
3125+ sqlite3_free(p->zTableName);
3126+ p->zTableName = zNewName;
3127+ }else{
3128+ sqlite3_free(zNewName);
3129+ }
3130+ return rc;
3131+}
3132+
3133+
3134+/*
3135+** A virtual table module that provides fuzzy search.
3136+*/
3137+static sqlite3_module spellfix1Module = {
3138+ 0, /* iVersion */
3139+ spellfix1Create, /* xCreate - handle CREATE VIRTUAL TABLE */
3140+ spellfix1Connect, /* xConnect - reconnected to an existing table */
3141+ spellfix1BestIndex, /* xBestIndex - figure out how to do a query */
3142+ spellfix1Disconnect, /* xDisconnect - close a connection */
3143+ spellfix1Destroy, /* xDestroy - handle DROP TABLE */
3144+ spellfix1Open, /* xOpen - open a cursor */
3145+ spellfix1Close, /* xClose - close a cursor */
3146+ spellfix1Filter, /* xFilter - configure scan constraints */
3147+ spellfix1Next, /* xNext - advance a cursor */
3148+ spellfix1Eof, /* xEof - check for end of scan */
3149+ spellfix1Column, /* xColumn - read data */
3150+ spellfix1Rowid, /* xRowid - read data */
3151+ spellfix1Update, /* xUpdate */
3152+ 0, /* xBegin */
3153+ 0, /* xSync */
3154+ 0, /* xCommit */
3155+ 0, /* xRollback */
3156+ 0, /* xFindMethod */
3157+ spellfix1Rename, /* xRename */
3158+ 0, /* xSavepoint */
3159+ 0, /* xRelease */
3160+ 0, /* xRollbackTo */
3161+ 0, /* xShadowName */
3162+ 0 /* xIntegrity */
3163+};
3164+
3165+/*
3166+** Register the various functions and the virtual table.
3167+*/
3168+static int spellfix1Register(sqlite3 *db){
3169+ int rc = SQLITE_OK;
3170+ int i;
3171+ rc = sqlite3_create_function(db, "spellfix1_translit", 1,
3172+ SQLITE_UTF8|SQLITE_DETERMINISTIC, 0,
3173+ transliterateSqlFunc, 0, 0);
3174+ if( rc==SQLITE_OK ){
3175+ rc = sqlite3_create_function(db, "spellfix1_editdist", 2,
3176+ SQLITE_UTF8|SQLITE_DETERMINISTIC, 0,
3177+ editdistSqlFunc, 0, 0);
3178+ }
3179+ if( rc==SQLITE_OK ){
3180+ rc = sqlite3_create_function(db, "spellfix1_phonehash", 1,
3181+ SQLITE_UTF8|SQLITE_DETERMINISTIC, 0,
3182+ phoneticHashSqlFunc, 0, 0);
3183+ }
3184+ if( rc==SQLITE_OK ){
3185+ rc = sqlite3_create_function(db, "spellfix1_scriptcode", 1,
3186+ SQLITE_UTF8|SQLITE_DETERMINISTIC, 0,
3187+ scriptCodeSqlFunc, 0, 0);
3188+ }
3189+ if( rc==SQLITE_OK ){
3190+ rc = sqlite3_create_module(db, "spellfix1", &spellfix1Module, 0);
3191+ }
3192+ if( rc==SQLITE_OK ){
3193+ rc = editDist3Install(db);
3194+ }
3195+
3196+ /* Verify sanity of the translit[] table */
3197+ for(i=0; i<sizeof(translit)/sizeof(translit[0])-1; i++){
3198+ assert( translit[i].cFrom<translit[i+1].cFrom );
3199+ }
3200+
3201+ return rc;
3202+}
3203+
3204+#endif /* SQLITE_OMIT_VIRTUALTABLE */
3205+
3206+/*
3207+** Extension load function.
3208+*/
3209+#ifdef _WIN32
3210+__declspec(dllexport)
3211+#endif
3212+int sqlite3_spellfix_init(
3213+ sqlite3 *db,
3214+ char **pzErrMsg,
3215+ const sqlite3_api_routines *pApi
3216+){
3217+ SQLITE_EXTENSION_INIT2(pApi);
3218+#ifndef SQLITE_OMIT_VIRTUALTABLE
3219+ return spellfix1Register(db);
3220+#endif
3221+ return SQLITE_OK;
3222+}
3223diff --git a/main.c b/main.c
3224index 69a68b8de4df79435dead3ef0535736d7c6422e0..f95a4659d4627d23ca5804896b624110400ada9c 100644
3225--- a/main.c
3226+++ b/main.c
3227@@ -11,13 +11,14 @@ unsigned int count_lines(FILE* file);
3228 int load_or_save_db(sqlite3 *pInMemory, const char *zFilename, int isSave);
3229
3230 int main() {
3231+
3232+ Data *data = new_data(":memory:");
3233+ bootstrap(data);
3234+
3235 setlocale(LC_ALL, "");
3236 initscr();
3237 noecho();
3238 cbreak();
3239-
3240- Data *data = new_data(":memory:");
3241- bootstrap(data);
3242
3243 FILE *f = fopen("dict.txt", "r");
3244 unsigned int lines = count_lines(f);
3245diff --git a/ui.c b/ui.c
3246index 8837bab19ca394e22205abe5312b37350ddad5e5..2e779e3a4c4dc055987446d7be311e27ba0b7a7e 100644
3247--- a/ui.c
3248+++ b/ui.c
3249@@ -5,30 +5,11 @@ #include "ui.h"
3250
3251 const char *uload = "█";
3252
3253-
3254 PROGRESS_BAR* new_progress_bar(WINDOW* scr, float total) {
3255 PROGRESS_BAR *bar = (PROGRESS_BAR*)malloc(sizeof(PROGRESS_BAR));
3256 bar->scr = scr;
3257 bar->total = total;
3258 bar->current = 0;
3259-
3260- int x, y;
3261- int hx, hy;
3262-
3263- getmaxyx(scr, y, x);
3264-
3265- hx = x/2;
3266- hy = y/2;
3267-
3268- wmove(scr, hy-1, 0);
3269- wprintw(scr, uload);
3270-
3271- wmove(scr, hy, hx-4);
3272- wprintw(scr, "000%%");
3273-
3274- wmove(scr, hy+1, hx);
3275- wprintw(scr, "%.0f/%.0f", 0.0, total);
3276-
3277 return bar;
3278 }
3279