From 08119c361d1181b3e8f1abb429236e488a664753 Mon Sep 17 00:00:00 2001 From: Hans-Christoph Steiner Date: Tue, 13 Aug 2013 15:42:54 -0400 Subject: Imported Upstream version 2.2.1 --- ext/misc/spellfix.c | 2844 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 2844 insertions(+) create mode 100644 ext/misc/spellfix.c (limited to 'ext/misc/spellfix.c') diff --git a/ext/misc/spellfix.c b/ext/misc/spellfix.c new file mode 100644 index 0000000..eb5442e --- /dev/null +++ b/ext/misc/spellfix.c @@ -0,0 +1,2844 @@ +/* +** 2012 April 10 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** This module implements the spellfix1 VIRTUAL TABLE that can be used +** to search a large vocabulary for close matches. See separate +** documentation (http://www.sqlite.org/spellfix1.html) for details. +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 + +#ifndef SQLITE_AMALGAMATION +# include +# include +# include +# include +# define ALWAYS(X) 1 +# define NEVER(X) 0 + typedef unsigned char u8; + typedef unsigned short u16; +# include +#endif + +#ifndef SQLITE_OMIT_VIRTUALTABLE + +/* +** Character classes for ASCII characters: +** +** 0 '' Silent letters: H W +** 1 'A' Any vowel: A E I O U (Y) +** 2 'B' A bilabeal stop or fricative: B F P V W +** 3 'C' Other fricatives or back stops: C G J K Q S X Z +** 4 'D' Alveolar stops: D T +** 5 'H' Letter H at the beginning of a word +** 6 'L' Glide: L +** 7 'R' Semivowel: R +** 8 'M' Nasals: M N +** 9 'Y' Letter Y at the beginning of a word. +** 10 '9' Digits: 0 1 2 3 4 5 6 7 8 9 +** 11 ' ' White space +** 12 '?' Other. +*/ +#define CCLASS_SILENT 0 +#define CCLASS_VOWEL 1 +#define CCLASS_B 2 +#define CCLASS_C 3 +#define CCLASS_D 4 +#define CCLASS_H 5 +#define CCLASS_L 6 +#define CCLASS_R 7 +#define CCLASS_M 8 +#define CCLASS_Y 9 +#define CCLASS_DIGIT 10 +#define CCLASS_SPACE 11 +#define CCLASS_OTHER 12 + +/* +** The following table gives the character class for non-initial ASCII +** characters. +*/ +static const unsigned char midClass[] = { + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, + /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, + /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, + /* ' */ CCLASS_SILENT, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, + /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, + /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, + /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, + /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, + /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, + /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, + /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, + /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, + /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, + /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, + /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, + /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, + /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, + /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, + /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, + /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_VOWEL, + /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, + /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, + /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, + /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, + /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, + /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, + /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, + /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, + /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, + /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, + /* x */ CCLASS_C, /* y */ CCLASS_VOWEL, /* z */ CCLASS_C, + /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, + /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, +}; +/* +** This tables gives the character class for ASCII characters that form the +** initial character of a word. The only difference from midClass is with +** the letters H, W, and Y. +*/ +static const unsigned char initClass[] = { + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, + /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, + /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, + /* ' */ CCLASS_OTHER, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, + /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, + /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, + /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, + /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, + /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, + /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, + /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, + /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, + /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, + /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, + /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, + /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, + /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, + /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, + /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, + /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_Y, + /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, + /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, + /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, + /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, + /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, + /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, + /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, + /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, + /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, + /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, + /* x */ CCLASS_C, /* y */ CCLASS_Y, /* z */ CCLASS_C, + /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, + /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, +}; + +/* +** Mapping from the character class number (0-13) to a symbol for each +** character class. Note that initClass[] can be used to map the class +** symbol back into the class number. +*/ +static const unsigned char className[] = ".ABCDHLRMY9 ?"; + +/* +** Generate a "phonetic hash" from a string of ASCII characters +** in zIn[0..nIn-1]. +** +** * Map characters by character class as defined above. +** * Omit double-letters +** * Omit vowels beside R and L +** * Omit T when followed by CH +** * Omit W when followed by R +** * Omit D when followed by J or G +** * Omit K in KN or G in GN at the beginning of a word +** +** Space to hold the result is obtained from sqlite3_malloc() +** +** Return NULL if memory allocation fails. +*/ +static unsigned char *phoneticHash(const unsigned char *zIn, int nIn){ + unsigned char *zOut = sqlite3_malloc( nIn + 1 ); + int i; + int nOut = 0; + char cPrev = 0x77; + char cPrevX = 0x77; + const unsigned char *aClass = initClass; + + if( zOut==0 ) return 0; + if( nIn>2 ){ + switch( zIn[0] ){ + case 'g': + case 'k': { + if( zIn[1]=='n' ){ zIn++; nIn--; } + break; + } + } + } + for(i=0; i=0 ); + if( nOut==0 || c!=zOut[nOut-1] ) zOut[nOut++] = c; + } + zOut[nOut] = 0; + return zOut; +} + +/* +** This is an SQL function wrapper around phoneticHash(). See +** the description of phoneticHash() for additional information. +*/ +static void phoneticHashSqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn; + unsigned char *zOut; + + zIn = sqlite3_value_text(argv[0]); + if( zIn==0 ) return; + zOut = phoneticHash(zIn, sqlite3_value_bytes(argv[0])); + if( zOut==0 ){ + sqlite3_result_error_nomem(context); + }else{ + sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); + } +} + +/* +** Return the character class number for a character given its +** context. +*/ +static char characterClass(char cPrev, char c){ + return cPrev==0 ? initClass[c&0x7f] : midClass[c&0x7f]; +} + +/* +** Return the cost of inserting or deleting character c immediately +** following character cPrev. If cPrev==0, that means c is the first +** character of the word. +*/ +static int insertOrDeleteCost(char cPrev, char c, char cNext){ + char classC = characterClass(cPrev, c); + char classCprev; + + if( classC==CCLASS_SILENT ){ + /* Insert or delete "silent" characters such as H or W */ + return 1; + } + if( cPrev==c ){ + /* Repeated characters, or miss a repeat */ + return 10; + } + if( classC==CCLASS_VOWEL && (cPrev=='r' || cNext=='r') ){ + return 20; /* Insert a vowel before or after 'r' */ + } + classCprev = characterClass(cPrev, cPrev); + if( classC==classCprev ){ + if( classC==CCLASS_VOWEL ){ + /* Remove or add a new vowel to a vowel cluster */ + return 15; + }else{ + /* Remove or add a consonant not in the same class */ + return 50; + } + } + + /* any other character insertion or deletion */ + return 100; +} + +/* +** Divide the insertion cost by this factor when appending to the +** end of the word. +*/ +#define FINAL_INS_COST_DIV 4 + +/* +** Return the cost of substituting cTo in place of cFrom assuming +** the previous character is cPrev. If cPrev==0 then cTo is the first +** character of the word. +*/ +static int substituteCost(char cPrev, char cFrom, char cTo){ + char classFrom, classTo; + if( cFrom==cTo ){ + /* Exact match */ + return 0; + } + if( cFrom==(cTo^0x20) && ((cTo>='A' && cTo<='Z') || (cTo>='a' && cTo<='z')) ){ + /* differ only in case */ + return 0; + } + classFrom = characterClass(cPrev, cFrom); + classTo = characterClass(cPrev, cTo); + if( classFrom==classTo ){ + /* Same character class */ + return 40; + } + if( classFrom>=CCLASS_B && classFrom<=CCLASS_Y + && classTo>=CCLASS_B && classTo<=CCLASS_Y ){ + /* Convert from one consonant to another, but in a different class */ + return 75; + } + /* Any other subsitution */ + return 100; +} + +/* +** Given two strings zA and zB which are pure ASCII, return the cost +** of transforming zA into zB. If zA ends with '*' assume that it is +** a prefix of zB and give only minimal penalty for extra characters +** on the end of zB. +** +** Smaller numbers mean a closer match. +** +** Negative values indicate an error: +** -1 One of the inputs is NULL +** -2 Non-ASCII characters on input +** -3 Unable to allocate memory +** +** If pnMatch is not NULL, then *pnMatch is set to the number of bytes +** of zB that matched the pattern in zA. If zA does not end with a '*', +** then this value is always the number of bytes in zB (i.e. strlen(zB)). +** If zA does end in a '*', then it is the number of bytes in the prefix +** of zB that was deemed to match zA. +*/ +static int editdist1(const char *zA, const char *zB, int *pnMatch){ + int nA, nB; /* Number of characters in zA[] and zB[] */ + int xA, xB; /* Loop counters for zA[] and zB[] */ + char cA, cB; /* Current character of zA and zB */ + char cAprev, cBprev; /* Previous character of zA and zB */ + char cAnext, cBnext; /* Next character in zA and zB */ + int d; /* North-west cost value */ + int dc = 0; /* North-west character value */ + int res; /* Final result */ + int *m; /* The cost matrix */ + char *cx; /* Corresponding character values */ + int *toFree = 0; /* Malloced space */ + int mStack[60+15]; /* Stack space to use if not too much is needed */ + int nMatch = 0; + + /* Early out if either input is NULL */ + if( zA==0 || zB==0 ) return -1; + + /* Skip any common prefix */ + while( zA[0] && zA[0]==zB[0] ){ dc = zA[0]; zA++; zB++; nMatch++; } + if( pnMatch ) *pnMatch = nMatch; + if( zA[0]==0 && zB[0]==0 ) return 0; + +#if 0 + printf("A=\"%s\" B=\"%s\" dc=%c\n", zA, zB, dc?dc:' '); +#endif + + /* Verify input strings and measure their lengths */ + for(nA=0; zA[nA]; nA++){ + if( zA[nA]&0x80 ) return -2; + } + for(nB=0; zB[nB]; nB++){ + if( zB[nB]&0x80 ) return -2; + } + + /* Special processing if either string is empty */ + if( nA==0 ){ + cBprev = dc; + for(xB=res=0; (cB = zB[xB])!=0; xB++){ + res += insertOrDeleteCost(cBprev, cB, zB[xB+1])/FINAL_INS_COST_DIV; + cBprev = cB; + } + return res; + } + if( nB==0 ){ + cAprev = dc; + for(xA=res=0; (cA = zA[xA])!=0; xA++){ + res += insertOrDeleteCost(cAprev, cA, zA[xA+1]); + cAprev = cA; + } + return res; + } + + /* A is a prefix of B */ + if( zA[0]=='*' && zA[1]==0 ) return 0; + + /* Allocate and initialize the Wagner matrix */ + if( nB<(sizeof(mStack)*4)/(sizeof(mStack[0])*5) ){ + m = mStack; + }else{ + m = toFree = sqlite3_malloc( (nB+1)*5*sizeof(m[0])/4 ); + if( m==0 ) return -3; + } + cx = (char*)&m[nB+1]; + + /* Compute the Wagner edit distance */ + m[0] = 0; + cx[0] = dc; + cBprev = dc; + for(xB=1; xB<=nB; xB++){ + cBnext = zB[xB]; + cB = zB[xB-1]; + cx[xB] = cB; + m[xB] = m[xB-1] + insertOrDeleteCost(cBprev, cB, cBnext); + cBprev = cB; + } + cAprev = dc; + for(xA=1; xA<=nA; xA++){ + int lastA = (xA==nA); + cA = zA[xA-1]; + cAnext = zA[xA]; + if( cA=='*' && lastA ) break; + d = m[0]; + dc = cx[0]; + m[0] = d + insertOrDeleteCost(cAprev, cA, cAnext); + cBprev = 0; + for(xB=1; xB<=nB; xB++){ + int totalCost, insCost, delCost, subCost, ncx; + cB = zB[xB-1]; + cBnext = zB[xB]; + + /* Cost to insert cB */ + insCost = insertOrDeleteCost(cx[xB-1], cB, cBnext); + if( lastA ) insCost /= FINAL_INS_COST_DIV; + + /* Cost to delete cA */ + delCost = insertOrDeleteCost(cx[xB], cA, cBnext); + + /* Cost to substitute cA->cB */ + subCost = substituteCost(cx[xB-1], cA, cB); + + /* Best cost */ + totalCost = insCost + m[xB-1]; + ncx = cB; + if( (delCost + m[xB])nLang; i++){ + EditDist3Cost *pCost, *pNext; + pCost = p->a[i].pCost; + while( pCost ){ + pNext = pCost->pNext; + sqlite3_free(pCost); + pCost = pNext; + } + } + sqlite3_free(p->a); + memset(p, 0, sizeof(*p)); +} +static void editDist3ConfigDelete(void *pIn){ + EditDist3Config *p = (EditDist3Config*)pIn; + editDist3ConfigClear(p); + sqlite3_free(p); +} + +/* +** Load all edit-distance weights from a table. +*/ +static int editDist3ConfigLoad( + EditDist3Config *p, /* The edit distance configuration to load */ + sqlite3 *db, /* Load from this database */ + const char *zTable /* Name of the table from which to load */ +){ + sqlite3_stmt *pStmt; + int rc, rc2; + char *zSql; + int iLangPrev = -9999; + EditDist3Lang *pLang = 0; + + zSql = sqlite3_mprintf("SELECT iLang, cFrom, cTo, iCost" + " FROM \"%w\" WHERE iLang>=0 ORDER BY iLang", zTable); + if( zSql==0 ) return SQLITE_NOMEM; + rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + if( rc ) return rc; + editDist3ConfigClear(p); + while( sqlite3_step(pStmt)==SQLITE_ROW ){ + int iLang = sqlite3_column_int(pStmt, 0); + const char *zFrom = (const char*)sqlite3_column_text(pStmt, 1); + int nFrom = zFrom ? sqlite3_column_bytes(pStmt, 1) : 0; + const char *zTo = (const char*)sqlite3_column_text(pStmt, 2); + int nTo = zTo ? sqlite3_column_bytes(pStmt, 2) : 0; + int iCost = sqlite3_column_int(pStmt, 3); + + assert( zFrom!=0 || nFrom==0 ); + assert( zTo!=0 || nTo==0 ); + if( nFrom>100 || nTo>100 ) continue; + if( iCost<0 ) continue; + if( pLang==0 || iLang!=iLangPrev ){ + EditDist3Lang *pNew; + pNew = sqlite3_realloc(p->a, (p->nLang+1)*sizeof(p->a[0])); + if( pNew==0 ){ rc = SQLITE_NOMEM; break; } + p->a = pNew; + pLang = &p->a[p->nLang]; + p->nLang++; + pLang->iLang = iLang; + pLang->iInsCost = 100; + pLang->iDelCost = 100; + pLang->iSubCost = 150; + pLang->pCost = 0; + iLangPrev = iLang; + } + if( nFrom==1 && zFrom[0]=='?' && nTo==0 ){ + pLang->iDelCost = iCost; + }else if( nFrom==0 && nTo==1 && zTo[0]=='?' ){ + pLang->iInsCost = iCost; + }else if( nFrom==1 && nTo==1 && zFrom[0]=='?' && zTo[0]=='?' ){ + pLang->iSubCost = iCost; + }else{ + EditDist3Cost *pCost; + int nExtra = nFrom + nTo - 4; + if( nExtra<0 ) nExtra = 0; + pCost = sqlite3_malloc( sizeof(*pCost) + nExtra ); + if( pCost==0 ){ rc = SQLITE_NOMEM; break; } + pCost->nFrom = nFrom; + pCost->nTo = nTo; + pCost->iCost = iCost; + memcpy(pCost->a, zFrom, nFrom); + memcpy(pCost->a + nFrom, zTo, nTo); + pCost->pNext = pLang->pCost; + pLang->pCost = pCost; + } + } + rc2 = sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ) rc = rc2; + return rc; +} + +/* +** Return the length (in bytes) of a utf-8 character. Or return a maximum +** of N. +*/ +static int utf8Len(unsigned char c, int N){ + int len = 1; + if( c>0x7f ){ + if( (c&0xe0)==0xc0 ){ + len = 2; + }else if( (c&0xf0)==0xe0 ){ + len = 3; + }else{ + len = 4; + } + } + if( len>N ) len = N; + return len; +} + +/* +** Return TRUE (non-zero) if the To side of the given cost matches +** the given string. +*/ +static int matchTo(EditDist3Cost *p, const char *z, int n){ + if( p->nTo>n ) return 0; + if( strncmp(p->a+p->nFrom, z, p->nTo)!=0 ) return 0; + return 1; +} + +/* +** Return TRUE (non-zero) if the From side of the given cost matches +** the given string. +*/ +static int matchFrom(EditDist3Cost *p, const char *z, int n){ + assert( p->nFrom<=n ); + if( strncmp(p->a, z, p->nFrom)!=0 ) return 0; + return 1; +} + +/* +** Return TRUE (non-zero) of the next FROM character and the next TO +** character are the same. +*/ +static int matchFromTo( + EditDist3FromString *pStr, /* Left hand string */ + int n1, /* Index of comparison character on the left */ + const char *z2, /* Right-handl comparison character */ + int n2 /* Bytes remaining in z2[] */ +){ + int b1 = pStr->a[n1].nByte; + if( b1>n2 ) return 0; + if( memcmp(pStr->z+n1, z2, b1)!=0 ) return 0; + return 1; +} + +/* +** Delete an EditDist3FromString objecct +*/ +static void editDist3FromStringDelete(EditDist3FromString *p){ + int i; + if( p ){ + for(i=0; in; i++){ + sqlite3_free(p->a[i].apDel); + sqlite3_free(p->a[i].apSubst); + } + sqlite3_free(p); + } +} + +/* +** Create a EditDist3FromString object. +*/ +static EditDist3FromString *editDist3FromStringNew( + const EditDist3Lang *pLang, + const char *z, + int n +){ + EditDist3FromString *pStr; + EditDist3Cost *p; + int i; + + if( z==0 ) return 0; + if( n<0 ) n = (int)strlen(z); + pStr = sqlite3_malloc( sizeof(*pStr) + sizeof(pStr->a[0])*n + n + 1 ); + if( pStr==0 ) return 0; + pStr->a = (EditDist3From*)&pStr[1]; + memset(pStr->a, 0, sizeof(pStr->a[0])*n); + pStr->n = n; + pStr->z = (char*)&pStr->a[n]; + memcpy(pStr->z, z, n+1); + if( n && z[n-1]=='*' ){ + pStr->isPrefix = 1; + n--; + pStr->n--; + pStr->z[n] = 0; + }else{ + pStr->isPrefix = 0; + } + + for(i=0; ia[i]; + memset(pFrom, 0, sizeof(*pFrom)); + pFrom->nByte = utf8Len((unsigned char)z[i], n-i); + for(p=pLang->pCost; p; p=p->pNext){ + EditDist3Cost **apNew; + if( i+p->nFrom>n ) continue; + if( matchFrom(p, z+i, n-i)==0 ) continue; + if( p->nTo==0 ){ + apNew = sqlite3_realloc(pFrom->apDel, + sizeof(*apNew)*(pFrom->nDel+1)); + if( apNew==0 ) break; + pFrom->apDel = apNew; + apNew[pFrom->nDel++] = p; + }else{ + apNew = sqlite3_realloc(pFrom->apSubst, + sizeof(*apNew)*(pFrom->nSubst+1)); + if( apNew==0 ) break; + pFrom->apSubst = apNew; + apNew[pFrom->nSubst++] = p; + } + } + if( p ){ + editDist3FromStringDelete(pStr); + pStr = 0; + break; + } + } + return pStr; +} + +/* +** Update entry m[i] such that it is the minimum of its current value +** and m[j]+iCost. +** +** If the iCost is 1,000,000 or greater, then consider the cost to be +** infinite and skip the update. +*/ +static void updateCost( + unsigned int *m, + int i, + int j, + int iCost +){ + assert( iCost>=0 ); + if( iCost<10000 ){ + unsigned int b = m[j] + iCost; + if( bpCost; p; p=p->pNext){ + EditDist3Cost **apNew; + if( p->nFrom>0 ) continue; + if( i2+p->nTo>n2 ) continue; + if( matchTo(p, z2+i2, n2-i2)==0 ) continue; + a2[i2].nIns++; + apNew = sqlite3_realloc(a2[i2].apIns, sizeof(*apNew)*a2[i2].nIns); + if( apNew==0 ){ + res = -1; /* Out of memory */ + goto editDist3Abort; + } + a2[i2].apIns = apNew; + a2[i2].apIns[a2[i2].nIns-1] = p; + } + } + + /* Prepare to compute the minimum edit distance */ + szRow = f.n+1; + memset(m, 0x01, (n2+1)*szRow*sizeof(m[0])); + m[0] = 0; + + /* First fill in the top-row of the matrix with FROM deletion costs */ + for(i1=0; i1iDelCost); + for(k=0; knFrom, i1, p->iCost); + } + } + + /* Fill in all subsequent rows, top-to-bottom, left-to-right */ + for(i2=0; i2iInsCost); + for(k=0; knTo), rxp, p->iCost); + } + for(i1=0; i1iDelCost); + for(k=0; knFrom, cxp, p->iCost); + } + updateCost(m, cx, cxu, pLang->iInsCost); + if( matchFromTo(&f, i1, z2+i2, n2-i2) ){ + updateCost(m, cx, cxd, 0); + } + updateCost(m, cx, cxd, pLang->iSubCost); + for(k=0; knFrom+szRow*p->nTo, cxd, p->iCost); + } + } + } + } + +#if 0 /* Enable for debugging */ + printf(" ^"); + for(i1=0; i19999 ) printf(" ****"); + else printf(" %4d", v); + } + printf("\n"); + for(i2=0; i29999 ) printf(" ****"); + else printf(" %4d", v); + } + printf("\n"); + } +#endif + + /* Free memory allocations and return the result */ + res = (int)m[szRow*(n2+1)-1]; + n = n2; + if( f.isPrefix ){ + for(i2=1; i2<=n2; i2++){ + int b = m[szRow*i2-1]; + if( b<=res ){ + res = b; + n = i2 - 1; + } + } + } + if( pnMatch ){ + int nExtra = 0; + for(k=0; knLang; i++){ + if( pConfig->a[i].iLang==iLang ) return &pConfig->a[i]; + } + return &editDist3Lang; +} + +/* +** Function: editdist3(A,B,iLang) +** editdist3(tablename) +** +** Return the cost of transforming string A into string B using edit +** weights for iLang. +** +** The second form loads edit weights into memory from a table. +*/ +static void editDist3SqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + EditDist3Config *pConfig = (EditDist3Config*)sqlite3_user_data(context); + sqlite3 *db = sqlite3_context_db_handle(context); + int rc; + if( argc==1 ){ + const char *zTable = (const char*)sqlite3_value_text(argv[0]); + rc = editDist3ConfigLoad(pConfig, db, zTable); + if( rc ) sqlite3_result_error_code(context, rc); + }else{ + const char *zA = (const char*)sqlite3_value_text(argv[0]); + const char *zB = (const char*)sqlite3_value_text(argv[1]); + int nA = sqlite3_value_bytes(argv[0]); + int nB = sqlite3_value_bytes(argv[1]); + int iLang = argc==3 ? sqlite3_value_int(argv[2]) : 0; + const EditDist3Lang *pLang = editDist3FindLang(pConfig, iLang); + EditDist3FromString *pFrom; + int dist; + + pFrom = editDist3FromStringNew(pLang, zA, nA); + if( pFrom==0 ){ + sqlite3_result_error_nomem(context); + return; + } + dist = editDist3Core(pFrom, zB, nB, pLang, 0); + editDist3FromStringDelete(pFrom); + if( dist==(-1) ){ + sqlite3_result_error_nomem(context); + }else{ + sqlite3_result_int(context, dist); + } + } +} + +/* +** Register the editDist3 function with SQLite +*/ +static int editDist3Install(sqlite3 *db){ + int rc; + EditDist3Config *pConfig = sqlite3_malloc( sizeof(*pConfig) ); + if( pConfig==0 ) return SQLITE_NOMEM; + memset(pConfig, 0, sizeof(*pConfig)); + rc = sqlite3_create_function_v2(db, "editdist3", + 2, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function_v2(db, "editdist3", + 3, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function_v2(db, "editdist3", + 1, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, + editDist3ConfigDelete); + }else{ + sqlite3_free(pConfig); + } + return rc; +} +/* End configurable cost unicode edit distance routines +****************************************************************************** +****************************************************************************** +** Begin transliterate unicode-to-ascii implementation +*/ + +#if !SQLITE_AMALGAMATION +/* +** This lookup table is used to help decode the first byte of +** a multi-byte UTF8 character. +*/ +static const unsigned char sqlite3Utf8Trans1[] = { + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, + 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00, +}; +#endif + +/* +** Return the value of the first UTF-8 character in the string. +*/ +static int utf8Read(const unsigned char *z, int n, int *pSize){ + int c, i; + + /* All callers to this routine (in the current implementation) + ** always have n>0. */ + if( NEVER(n==0) ){ + c = i = 0; + }else{ + c = z[0]; + i = 1; + if( c>=0xc0 ){ + c = sqlite3Utf8Trans1[c-0xc0]; + while( i0 ){ + c = utf8Read(zIn, nIn, &sz); + zIn += sz; + nIn -= sz; + if( c<=127 ){ + zOut[nOut++] = c; + }else{ + int xTop, xBtm, x; + xTop = sizeof(translit)/sizeof(translit[0]) - 1; + xBtm = 0; + while( xTop>=xBtm ){ + x = (xTop + xBtm)/2; + if( translit[x].cFrom==c ){ + zOut[nOut++] = translit[x].cTo0; + if( translit[x].cTo1 ){ + zOut[nOut++] = translit[x].cTo1; + /* Add an extra "ch" after the "sh" for Щ and щ */ + if( c==0x0429 || c== 0x0449 ){ + zOut[nOut++] = 'c'; + zOut[nOut++] = 'h'; + } + } + c = 0; + break; + }else if( translit[x].cFrom>c ){ + xTop = x-1; + }else{ + xBtm = x+1; + } + } + if( c ) zOut[nOut++] = '?'; + } + } + zOut[nOut] = 0; + return zOut; +} + +/* +** Return the number of characters in the shortest prefix of the input +** string that transliterates to an ASCII string nTrans bytes or longer. +** Or, if the transliteration of the input string is less than nTrans +** bytes in size, return the number of characters in the input string. +*/ +static int translen_to_charlen(const char *zIn, int nIn, int nTrans){ + int i, c, sz, nOut; + int nChar; + + i = nOut = 0; + for(nChar=0; i=128 ){ + int xTop, xBtm, x; + xTop = sizeof(translit)/sizeof(translit[0]) - 1; + xBtm = 0; + while( xTop>=xBtm ){ + x = (xTop + xBtm)/2; + if( translit[x].cFrom==c ){ + if( translit[x].cTo1 ) nOut++; + if( c==0x0429 || c== 0x0449 ) nOut += 2; + break; + }else if( translit[x].cFrom>c ){ + xTop = x-1; + }else{ + xBtm = x+1; + } + } + } + } + + return nChar; +} + + +/* +** spellfix1_translit(X) +** +** Convert a string that contains non-ASCII Roman characters into +** pure ASCII. +*/ +static void transliterateSqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn = sqlite3_value_text(argv[0]); + int nIn = sqlite3_value_bytes(argv[0]); + unsigned char *zOut = transliterate(zIn, nIn); + if( zOut==0 ){ + sqlite3_result_error_nomem(context); + }else{ + sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); + } +} + +/* +** spellfix1_scriptcode(X) +** +** Try to determine the dominant script used by the word X and return +** its ISO 15924 numeric code. +** +** The current implementation only understands the following scripts: +** +** 215 (Latin) +** 220 (Cyrillic) +** 200 (Greek) +** +** This routine will return 998 if the input X contains characters from +** two or more of the above scripts or 999 if X contains no characters +** from any of the above scripts. +*/ +static void scriptCodeSqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn = sqlite3_value_text(argv[0]); + int nIn = sqlite3_value_bytes(argv[0]); + int c, sz; + int scriptMask = 0; + int res; +# define SCRIPT_LATIN 0x0001 +# define SCRIPT_CYRILLIC 0x0002 +# define SCRIPT_GREEK 0x0004 + + while( nIn>0 ){ + c = utf8Read(zIn, nIn, &sz); + zIn += sz; + nIn -= sz; + if( c<0x02af ){ + scriptMask |= SCRIPT_LATIN; + }else if( c>=0x0400 && c<=0x04ff ){ + scriptMask |= SCRIPT_CYRILLIC; + }else if( c>=0x0386 && c<=0x03ce ){ + scriptMask |= SCRIPT_GREEK; + } + } + switch( scriptMask ){ + case 0: res = 999; break; + case SCRIPT_LATIN: res = 215; break; + case SCRIPT_CYRILLIC: res = 220; break; + case SCRIPT_GREEK: res = 200; break; + default: res = 998; break; + } + sqlite3_result_int(context, res); +} + +/* End transliterate +****************************************************************************** +****************************************************************************** +** Begin spellfix1 virtual table. +*/ + +/* Maximum length of a phonehash used for querying the shadow table */ +#define SPELLFIX_MX_HASH 8 + +/* Maximum number of hash strings to examine per query */ +#define SPELLFIX_MX_RUN 1 + +typedef struct spellfix1_vtab spellfix1_vtab; +typedef struct spellfix1_cursor spellfix1_cursor; + +/* Fuzzy-search virtual table object */ +struct spellfix1_vtab { + sqlite3_vtab base; /* Base class - must be first */ + sqlite3 *db; /* Database connection */ + char *zDbName; /* Name of database holding this table */ + char *zTableName; /* Name of the virtual table */ + char *zCostTable; /* Table holding edit-distance cost numbers */ + EditDist3Config *pConfig3; /* Parsed edit distance costs */ +}; + +/* Fuzzy-search cursor object */ +struct spellfix1_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + spellfix1_vtab *pVTab; /* The table to which this cursor belongs */ + char *zPattern; /* rhs of MATCH clause */ + int nRow; /* Number of rows of content */ + int nAlloc; /* Number of allocated rows */ + int iRow; /* Current row of content */ + int iLang; /* Value of the langid= constraint */ + int iTop; /* Value of the top= constraint */ + int iScope; /* Value of the scope= constraint */ + int nSearch; /* Number of vocabulary items checked */ + sqlite3_stmt *pFullScan; /* Shadow query for a full table scan */ + struct spellfix1_row { /* For each row of content */ + sqlite3_int64 iRowid; /* Rowid for this row */ + char *zWord; /* Text for this row */ + int iRank; /* Rank for this row */ + int iDistance; /* Distance from pattern for this row */ + int iScore; /* Score for sorting */ + int iMatchlen; /* Value of matchlen column (or -1) */ + char zHash[SPELLFIX_MX_HASH]; /* the phonehash used for this match */ + } *a; +}; + +/* +** Construct one or more SQL statements from the format string given +** and then evaluate those statements. The success code is written +** into *pRc. +** +** If *pRc is initially non-zero then this routine is a no-op. +*/ +static void spellfix1DbExec( + int *pRc, /* Success code */ + sqlite3 *db, /* Database in which to run SQL */ + const char *zFormat, /* Format string for SQL */ + ... /* Arguments to the format string */ +){ + va_list ap; + char *zSql; + if( *pRc ) return; + va_start(ap, zFormat); + zSql = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + if( zSql==0 ){ + *pRc = SQLITE_NOMEM; + }else{ + *pRc = sqlite3_exec(db, zSql, 0, 0, 0); + sqlite3_free(zSql); + } +} + +/* +** xDisconnect/xDestroy method for the fuzzy-search module. +*/ +static int spellfix1Uninit(int isDestroy, sqlite3_vtab *pVTab){ + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + int rc = SQLITE_OK; + if( isDestroy ){ + sqlite3 *db = p->db; + spellfix1DbExec(&rc, db, "DROP TABLE IF EXISTS \"%w\".\"%w_vocab\"", + p->zDbName, p->zTableName); + } + if( rc==SQLITE_OK ){ + sqlite3_free(p->zTableName); + editDist3ConfigDelete(p->pConfig3); + sqlite3_free(p->zCostTable); + sqlite3_free(p); + } + return rc; +} +static int spellfix1Disconnect(sqlite3_vtab *pVTab){ + return spellfix1Uninit(0, pVTab); +} +static int spellfix1Destroy(sqlite3_vtab *pVTab){ + return spellfix1Uninit(1, pVTab); +} + +/* +** Make a copy of a string. Remove leading and trailing whitespace +** and dequote it. +*/ +static char *spellfix1Dequote(const char *zIn){ + char *zOut; + int i, j; + char c; + while( isspace(zIn[0]) ) zIn++; + zOut = sqlite3_mprintf("%s", zIn); + if( zOut==0 ) return 0; + i = (int)strlen(zOut); +#if 0 /* The parser will never leave spaces at the end */ + while( i>0 && isspace(zOut[i-1]) ){ i--; } +#endif + zOut[i] = 0; + c = zOut[0]; + if( c=='\'' || c=='"' ){ + for(i=1, j=0; ALWAYS(zOut[i]); i++){ + zOut[j++] = zOut[i]; + if( zOut[i]==c ){ + if( zOut[i+1]==c ){ + i++; + }else{ + zOut[j-1] = 0; + break; + } + } + } + } + return zOut; +} + + +/* +** xConnect/xCreate method for the spellfix1 module. Arguments are: +** +** argv[0] -> module name ("spellfix1") +** argv[1] -> database name +** argv[2] -> table name +** argv[3].. -> optional arguments (i.e. "edit_cost_table" parameter) +*/ +static int spellfix1Init( + int isCreate, + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVTab, + char **pzErr +){ + spellfix1_vtab *pNew = 0; + const char *zModule = argv[0]; + const char *zDbName = argv[1]; + const char *zTableName = argv[2]; + int nDbName; + int rc = SQLITE_OK; + int i; + + nDbName = (int)strlen(zDbName); + pNew = sqlite3_malloc( sizeof(*pNew) + nDbName + 1); + if( pNew==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pNew, 0, sizeof(*pNew)); + pNew->zDbName = (char*)&pNew[1]; + memcpy(pNew->zDbName, zDbName, nDbName+1); + pNew->zTableName = sqlite3_mprintf("%s", zTableName); + pNew->db = db; + if( pNew->zTableName==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = sqlite3_declare_vtab(db, + "CREATE TABLE x(word,rank,distance,langid, " + "score, matchlen, phonehash HIDDEN, " + "top HIDDEN, scope HIDDEN, srchcnt HIDDEN, " + "soundslike HIDDEN, command HIDDEN)" + ); +#define SPELLFIX_COL_WORD 0 +#define SPELLFIX_COL_RANK 1 +#define SPELLFIX_COL_DISTANCE 2 +#define SPELLFIX_COL_LANGID 3 +#define SPELLFIX_COL_SCORE 4 +#define SPELLFIX_COL_MATCHLEN 5 +#define SPELLFIX_COL_PHONEHASH 6 +#define SPELLFIX_COL_TOP 7 +#define SPELLFIX_COL_SCOPE 8 +#define SPELLFIX_COL_SRCHCNT 9 +#define SPELLFIX_COL_SOUNDSLIKE 10 +#define SPELLFIX_COL_COMMAND 11 + } + if( rc==SQLITE_OK && isCreate ){ + sqlite3_uint64 r; + spellfix1DbExec(&rc, db, + "CREATE TABLE IF NOT EXISTS \"%w\".\"%w_vocab\"(\n" + " id INTEGER PRIMARY KEY,\n" + " rank INT,\n" + " langid INT,\n" + " word TEXT,\n" + " k1 TEXT,\n" + " k2 TEXT\n" + ");\n", + zDbName, zTableName + ); + sqlite3_randomness(sizeof(r), &r); + spellfix1DbExec(&rc, db, + "CREATE INDEX IF NOT EXISTS \"%w\".\"%w_index_%llx\" " + "ON \"%w_vocab\"(langid,k2);", + zDbName, zModule, r, zTableName + ); + } + for(i=3; rc==SQLITE_OK && ibase); + }else{ + *ppVTab = (sqlite3_vtab *)pNew; + } + return rc; +} + +/* +** The xConnect and xCreate methods +*/ +static int spellfix1Connect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVTab, + char **pzErr +){ + return spellfix1Init(0, db, pAux, argc, argv, ppVTab, pzErr); +} +static int spellfix1Create( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVTab, + char **pzErr +){ + return spellfix1Init(1, db, pAux, argc, argv, ppVTab, pzErr); +} + +/* +** Clear all of the content from a cursor. +*/ +static void spellfix1ResetCursor(spellfix1_cursor *pCur){ + int i; + for(i=0; inRow; i++){ + sqlite3_free(pCur->a[i].zWord); + } + pCur->nRow = 0; + pCur->iRow = 0; + pCur->nSearch = 0; + if( pCur->pFullScan ){ + sqlite3_finalize(pCur->pFullScan); + pCur->pFullScan = 0; + } +} + +/* +** Resize the cursor to hold up to N rows of content +*/ +static void spellfix1ResizeCursor(spellfix1_cursor *pCur, int N){ + struct spellfix1_row *aNew; + assert( N>=pCur->nRow ); + aNew = sqlite3_realloc(pCur->a, sizeof(pCur->a[0])*N); + if( aNew==0 && N>0 ){ + spellfix1ResetCursor(pCur); + sqlite3_free(pCur->a); + pCur->nAlloc = 0; + pCur->a = 0; + }else{ + pCur->nAlloc = N; + pCur->a = aNew; + } +} + + +/* +** Close a fuzzy-search cursor. +*/ +static int spellfix1Close(sqlite3_vtab_cursor *cur){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + spellfix1ResetCursor(pCur); + spellfix1ResizeCursor(pCur, 0); + sqlite3_free(pCur->zPattern); + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Search for terms of these forms: +** +** (A) word MATCH $str +** (B) langid == $langid +** (C) top = $top +** (D) scope = $scope +** (E) distance < $distance +** (F) distance <= $distance +** +** The plan number is a bit mask formed with these bits: +** +** 0x01 (A) is found +** 0x02 (B) is found +** 0x04 (C) is found +** 0x08 (D) is found +** 0x10 (E) is found +** 0x20 (F) is found +** +** filter.argv[*] values contains $str, $langid, $top, and $scope, +** if specified and in that order. +*/ +static int spellfix1BestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ + int iPlan = 0; + int iLangTerm = -1; + int iTopTerm = -1; + int iScopeTerm = -1; + int iDistTerm = -1; + int i; + const struct sqlite3_index_constraint *pConstraint; + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + + /* Terms of the form: word MATCH $str */ + if( (iPlan & 1)==0 + && pConstraint->iColumn==SPELLFIX_COL_WORD + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + + /* Terms of the form: langid = $langid */ + if( (iPlan & 2)==0 + && pConstraint->iColumn==SPELLFIX_COL_LANGID + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 2; + iLangTerm = i; + } + + /* Terms of the form: top = $top */ + if( (iPlan & 4)==0 + && pConstraint->iColumn==SPELLFIX_COL_TOP + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 4; + iTopTerm = i; + } + + /* Terms of the form: scope = $scope */ + if( (iPlan & 8)==0 + && pConstraint->iColumn==SPELLFIX_COL_SCOPE + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 8; + iScopeTerm = i; + } + + /* Terms of the form: distance < $dist or distance <= $dist */ + if( (iPlan & (16|32))==0 + && pConstraint->iColumn==SPELLFIX_COL_DISTANCE + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) + ){ + iPlan |= pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ? 16 : 32; + iDistTerm = i; + } + } + if( iPlan&1 ){ + int idx = 2; + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==SPELLFIX_COL_SCORE + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; /* Default order by iScore */ + } + if( iPlan&2 ){ + pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iLangTerm].omit = 1; + } + if( iPlan&4 ){ + pIdxInfo->aConstraintUsage[iTopTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iTopTerm].omit = 1; + } + if( iPlan&8 ){ + pIdxInfo->aConstraintUsage[iScopeTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iScopeTerm].omit = 1; + } + if( iPlan&(16|32) ){ + pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iDistTerm].omit = 1; + } + pIdxInfo->estimatedCost = (double)10000; + }else{ + pIdxInfo->idxNum = 0; + pIdxInfo->estimatedCost = (double)10000000; + } + return SQLITE_OK; +} + +/* +** Open a new fuzzy-search cursor. +*/ +static int spellfix1Open(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + spellfix1_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVTab = p; + *ppCursor = &pCur->base; + return SQLITE_OK; +} + +/* +** Adjust a distance measurement by the words rank in order to show +** preference to common words. +*/ +static int spellfix1Score(int iDistance, int iRank){ + int iLog2; + for(iLog2=0; iRank>0; iLog2++, iRank>>=1){} + return iDistance + 32 - iLog2; +} + +/* +** Compare two spellfix1_row objects for sorting purposes in qsort() such +** that they sort in order of increasing distance. +*/ +static int spellfix1RowCompare(const void *A, const void *B){ + const struct spellfix1_row *a = (const struct spellfix1_row*)A; + const struct spellfix1_row *b = (const struct spellfix1_row*)B; + return a->iScore - b->iScore; +} + +/* +** A structure used to pass information from spellfix1FilterForMatch() +** into spellfix1RunQuery(). +*/ +typedef struct MatchQuery { + spellfix1_cursor *pCur; /* The cursor being queried */ + sqlite3_stmt *pStmt; /* shadow table query statment */ + char zHash[SPELLFIX_MX_HASH]; /* The current phonehash for zPattern */ + const char *zPattern; /* Transliterated input string */ + int nPattern; /* Length of zPattern */ + EditDist3FromString *pMatchStr3; /* Original unicode string */ + EditDist3Config *pConfig3; /* Edit-distance cost coefficients */ + const EditDist3Lang *pLang; /* The selected language coefficients */ + int iLang; /* The language id */ + int iScope; /* Default scope */ + int iMaxDist; /* Maximum allowed edit distance, or -1 */ + int rc; /* Error code */ + int nRun; /* Number of prior runs for the same zPattern */ + char azPrior[SPELLFIX_MX_RUN][SPELLFIX_MX_HASH]; /* Prior hashes */ +} MatchQuery; + +/* +** Run a query looking for the best matches against zPattern using +** zHash as the character class seed hash. +*/ +static void spellfix1RunQuery(MatchQuery *p, const char *zQuery, int nQuery){ + const char *zK1; + const char *zWord; + int iDist; + int iRank; + int iScore; + int iWorst = 0; + int idx; + int idxWorst = -1; + int i; + int iScope = p->iScope; + spellfix1_cursor *pCur = p->pCur; + sqlite3_stmt *pStmt = p->pStmt; + char zHash1[SPELLFIX_MX_HASH]; + char zHash2[SPELLFIX_MX_HASH]; + char *zClass; + int nClass; + int rc; + + if( pCur->a==0 || p->rc ) return; /* Prior memory allocation failure */ + zClass = (char*)phoneticHash((unsigned char*)zQuery, nQuery); + if( zClass==0 ){ + p->rc = SQLITE_NOMEM; + return; + } + nClass = (int)strlen(zClass); + if( nClass>SPELLFIX_MX_HASH-2 ){ + nClass = SPELLFIX_MX_HASH-2; + zClass[nClass] = 0; + } + if( nClass<=iScope ){ + if( nClass>2 ){ + iScope = nClass-1; + }else{ + iScope = nClass; + } + } + memcpy(zHash1, zClass, iScope); + sqlite3_free(zClass); + zHash1[iScope] = 0; + memcpy(zHash2, zHash1, iScope); + zHash2[iScope] = 'Z'; + zHash2[iScope+1] = 0; +#if SPELLFIX_MX_RUN>1 + for(i=0; inRun; i++){ + if( strcmp(p->azPrior[i], zHash1)==0 ) return; + } +#endif + assert( p->nRunazPrior[p->nRun++], zHash1, iScope+1); + if( sqlite3_bind_text(pStmt, 1, zHash1, -1, SQLITE_STATIC)==SQLITE_NOMEM + || sqlite3_bind_text(pStmt, 2, zHash2, -1, SQLITE_STATIC)==SQLITE_NOMEM + ){ + p->rc = SQLITE_NOMEM; + return; + } +#if SPELLFIX_MX_RUN>1 + for(i=0; inRow; i++){ + if( pCur->a[i].iScore>iWorst ){ + iWorst = pCur->a[i].iScore; + idxWorst = i; + } + } +#endif + while( sqlite3_step(pStmt)==SQLITE_ROW ){ + int iMatchlen = -1; + iRank = sqlite3_column_int(pStmt, 2); + if( p->pMatchStr3 ){ + int nWord = sqlite3_column_bytes(pStmt, 1); + zWord = (const char*)sqlite3_column_text(pStmt, 1); + iDist = editDist3Core(p->pMatchStr3, zWord, nWord, p->pLang, &iMatchlen); + }else{ + zK1 = (const char*)sqlite3_column_text(pStmt, 3); + if( zK1==0 ) continue; + iDist = editdist1(p->zPattern, zK1, 0); + } + if( iDist<0 ){ + p->rc = SQLITE_NOMEM; + break; + } + pCur->nSearch++; + iScore = spellfix1Score(iDist,iRank); + if( p->iMaxDist>=0 ){ + if( iDist>p->iMaxDist ) continue; + if( pCur->nRow>=pCur->nAlloc-1 ){ + spellfix1ResizeCursor(pCur, pCur->nAlloc*2 + 10); + if( pCur->a==0 ) break; + } + idx = pCur->nRow; + }else if( pCur->nRownAlloc ){ + idx = pCur->nRow; + }else if( iScorea[idx].zWord); + }else{ + continue; + } + pCur->a[idx].zWord = sqlite3_mprintf("%s", sqlite3_column_text(pStmt, 1)); + if( pCur->a[idx].zWord==0 ){ + p->rc = SQLITE_NOMEM; + break; + } + pCur->a[idx].iRowid = sqlite3_column_int64(pStmt, 0); + pCur->a[idx].iRank = iRank; + pCur->a[idx].iDistance = iDist; + pCur->a[idx].iScore = iScore; + pCur->a[idx].iMatchlen = iMatchlen; + memcpy(pCur->a[idx].zHash, zHash1, iScope+1); + if( pCur->nRownAlloc ) pCur->nRow++; + if( pCur->nRow==pCur->nAlloc ){ + iWorst = pCur->a[0].iScore; + idxWorst = 0; + for(i=1; inRow; i++){ + iScore = pCur->a[i].iScore; + if( iWorstrc = rc; +} + +/* +** This version of the xFilter method work if the MATCH term is present +** and we are doing a scan. +*/ +static int spellfix1FilterForMatch( + spellfix1_cursor *pCur, + int idxNum, + int argc, + sqlite3_value **argv +){ + const unsigned char *zMatchThis; /* RHS of the MATCH operator */ + EditDist3FromString *pMatchStr3 = 0; /* zMatchThis as an editdist string */ + char *zPattern; /* Transliteration of zMatchThis */ + int nPattern; /* Length of zPattern */ + int iLimit = 20; /* Max number of rows of output */ + int iScope = 3; /* Use this many characters of zClass */ + int iLang = 0; /* Language code */ + char *zSql; /* SQL of shadow table query */ + sqlite3_stmt *pStmt = 0; /* Shadow table query */ + int rc; /* Result code */ + int idx = 1; /* Next available filter parameter */ + spellfix1_vtab *p = pCur->pVTab; /* The virtual table that owns pCur */ + MatchQuery x; /* For passing info to RunQuery() */ + + /* Load the cost table if we have not already done so */ + if( p->zCostTable!=0 && p->pConfig3==0 ){ + p->pConfig3 = sqlite3_malloc( sizeof(p->pConfig3[0]) ); + if( p->pConfig3==0 ) return SQLITE_NOMEM; + memset(p->pConfig3, 0, sizeof(p->pConfig3[0])); + rc = editDist3ConfigLoad(p->pConfig3, p->db, p->zCostTable); + if( rc ) return rc; + } + memset(&x, 0, sizeof(x)); + x.iScope = 3; /* Default scope if none specified by "WHERE scope=N" */ + x.iMaxDist = -1; /* Maximum allowed edit distance */ + + if( idxNum&2 ){ + iLang = sqlite3_value_int(argv[idx++]); + } + if( idxNum&4 ){ + iLimit = sqlite3_value_int(argv[idx++]); + if( iLimit<1 ) iLimit = 1; + } + if( idxNum&8 ){ + x.iScope = sqlite3_value_int(argv[idx++]); + if( x.iScope<1 ) x.iScope = 1; + if( x.iScope>SPELLFIX_MX_HASH-2 ) x.iScope = SPELLFIX_MX_HASH-2; + } + if( idxNum&(16|32) ){ + x.iMaxDist = sqlite3_value_int(argv[idx++]); + if( idxNum&16 ) x.iMaxDist--; + if( x.iMaxDist<0 ) x.iMaxDist = 0; + } + spellfix1ResetCursor(pCur); + spellfix1ResizeCursor(pCur, iLimit); + zMatchThis = sqlite3_value_text(argv[0]); + if( zMatchThis==0 ) return SQLITE_OK; + if( p->pConfig3 ){ + x.pLang = editDist3FindLang(p->pConfig3, iLang); + pMatchStr3 = editDist3FromStringNew(x.pLang, (const char*)zMatchThis, -1); + if( pMatchStr3==0 ){ + x.rc = SQLITE_NOMEM; + goto filter_exit; + } + }else{ + x.pLang = 0; + } + zPattern = (char*)transliterate(zMatchThis, sqlite3_value_bytes(argv[0])); + sqlite3_free(pCur->zPattern); + pCur->zPattern = zPattern; + if( zPattern==0 ){ + x.rc = SQLITE_NOMEM; + goto filter_exit; + } + nPattern = (int)strlen(zPattern); + if( zPattern[nPattern-1]=='*' ) nPattern--; + zSql = sqlite3_mprintf( + "SELECT id, word, rank, k1" + " FROM \"%w\".\"%w_vocab\"" + " WHERE langid=%d AND k2>=?1 AND k2zDbName, p->zTableName, iLang + ); + if( zSql==0 ){ + x.rc = SQLITE_NOMEM; + pStmt = 0; + goto filter_exit; + } + rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + pCur->iLang = iLang; + x.pCur = pCur; + x.pStmt = pStmt; + x.zPattern = zPattern; + x.nPattern = nPattern; + x.pMatchStr3 = pMatchStr3; + x.iLang = iLang; + x.rc = rc; + x.pConfig3 = p->pConfig3; + if( x.rc==SQLITE_OK ){ + spellfix1RunQuery(&x, zPattern, nPattern); + } + + if( pCur->a ){ + qsort(pCur->a, pCur->nRow, sizeof(pCur->a[0]), spellfix1RowCompare); + pCur->iTop = iLimit; + pCur->iScope = iScope; + }else{ + x.rc = SQLITE_NOMEM; + } + +filter_exit: + sqlite3_finalize(pStmt); + editDist3FromStringDelete(pMatchStr3); + return x.rc; +} + +/* +** This version of xFilter handles a full-table scan case +*/ +static int spellfix1FilterForFullScan( + spellfix1_cursor *pCur, + int idxNum, + int argc, + sqlite3_value **argv +){ + int rc; + char *zSql; + spellfix1_vtab *pVTab = pCur->pVTab; + spellfix1ResetCursor(pCur); + zSql = sqlite3_mprintf( + "SELECT word, rank, NULL, langid, id FROM \"%w\".\"%w_vocab\"", + pVTab->zDbName, pVTab->zTableName); + if( zSql==0 ) return SQLITE_NOMEM; + rc = sqlite3_prepare_v2(pVTab->db, zSql, -1, &pCur->pFullScan, 0); + sqlite3_free(zSql); + pCur->nRow = pCur->iRow = 0; + if( rc==SQLITE_OK ){ + rc = sqlite3_step(pCur->pFullScan); + if( rc==SQLITE_ROW ){ pCur->iRow = -1; rc = SQLITE_OK; } + if( rc==SQLITE_DONE ){ rc = SQLITE_OK; } + }else{ + pCur->iRow = 0; + } + return rc; +} + + +/* +** Called to "rewind" a cursor back to the beginning so that +** it starts its output over again. Always called at least once +** prior to any spellfix1Column, spellfix1Rowid, or spellfix1Eof call. +*/ +static int spellfix1Filter( + sqlite3_vtab_cursor *cur, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + int rc; + if( idxNum & 1 ){ + rc = spellfix1FilterForMatch(pCur, idxNum, argc, argv); + }else{ + rc = spellfix1FilterForFullScan(pCur, idxNum, argc, argv); + } + return rc; +} + + +/* +** Advance a cursor to its next row of output +*/ +static int spellfix1Next(sqlite3_vtab_cursor *cur){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + int rc = SQLITE_OK; + if( pCur->iRow < pCur->nRow ){ + if( pCur->pFullScan ){ + rc = sqlite3_step(pCur->pFullScan); + if( rc!=SQLITE_ROW ) pCur->iRow = pCur->nRow; + if( rc==SQLITE_ROW || rc==SQLITE_DONE ) rc = SQLITE_OK; + }else{ + pCur->iRow++; + } + } + return rc; +} + +/* +** Return TRUE if we are at the end-of-file +*/ +static int spellfix1Eof(sqlite3_vtab_cursor *cur){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + return pCur->iRow>=pCur->nRow; +} + +/* +** Return columns from the current row. +*/ +static int spellfix1Column( + sqlite3_vtab_cursor *cur, + sqlite3_context *ctx, + int i +){ + spellfix1_cursor *pCur = (spellfix1_cursor*)cur; + if( pCur->pFullScan ){ + if( i<=SPELLFIX_COL_LANGID ){ + sqlite3_result_value(ctx, sqlite3_column_value(pCur->pFullScan, i)); + }else{ + sqlite3_result_null(ctx); + } + return SQLITE_OK; + } + switch( i ){ + case SPELLFIX_COL_WORD: { + sqlite3_result_text(ctx, pCur->a[pCur->iRow].zWord, -1, SQLITE_STATIC); + break; + } + case SPELLFIX_COL_RANK: { + sqlite3_result_int(ctx, pCur->a[pCur->iRow].iRank); + break; + } + case SPELLFIX_COL_DISTANCE: { + sqlite3_result_int(ctx, pCur->a[pCur->iRow].iDistance); + break; + } + case SPELLFIX_COL_LANGID: { + sqlite3_result_int(ctx, pCur->iLang); + break; + } + case SPELLFIX_COL_SCORE: { + sqlite3_result_int(ctx, pCur->a[pCur->iRow].iScore); + break; + } + case SPELLFIX_COL_MATCHLEN: { + int iMatchlen = pCur->a[pCur->iRow].iMatchlen; + if( iMatchlen<0 ){ + int nPattern = (int)strlen(pCur->zPattern); + char *zWord = pCur->a[pCur->iRow].zWord; + int nWord = (int)strlen(zWord); + + if( nPattern>0 && pCur->zPattern[nPattern-1]=='*' ){ + char *zTranslit; + int res; + zTranslit = (char *)transliterate((unsigned char *)zWord, nWord); + if( !zTranslit ) return SQLITE_NOMEM; + res = editdist1(pCur->zPattern, zTranslit, &iMatchlen); + sqlite3_free(zTranslit); + if( res<0 ) return SQLITE_NOMEM; + iMatchlen = translen_to_charlen(zWord, nWord, iMatchlen); + }else{ + iMatchlen = utf8Charlen(zWord, nWord); + } + } + + sqlite3_result_int(ctx, iMatchlen); + break; + } + case SPELLFIX_COL_PHONEHASH: { + sqlite3_result_text(ctx, pCur->a[pCur->iRow].zHash, -1, SQLITE_STATIC); + break; + } + case SPELLFIX_COL_TOP: { + sqlite3_result_int(ctx, pCur->iTop); + break; + } + case SPELLFIX_COL_SCOPE: { + sqlite3_result_int(ctx, pCur->iScope); + break; + } + case SPELLFIX_COL_SRCHCNT: { + sqlite3_result_int(ctx, pCur->nSearch); + break; + } + default: { + sqlite3_result_null(ctx); + break; + } + } + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int spellfix1Rowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + spellfix1_cursor *pCur = (spellfix1_cursor*)cur; + if( pCur->pFullScan ){ + *pRowid = sqlite3_column_int64(pCur->pFullScan, 4); + }else{ + *pRowid = pCur->a[pCur->iRow].iRowid; + } + return SQLITE_OK; +} + +/* +** The xUpdate() method. +*/ +static int spellfix1Update( + sqlite3_vtab *pVTab, + int argc, + sqlite3_value **argv, + sqlite_int64 *pRowid +){ + int rc = SQLITE_OK; + sqlite3_int64 rowid, newRowid; + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + sqlite3 *db = p->db; + + if( argc==1 ){ + /* A delete operation on the rowid given by argv[0] */ + rowid = *pRowid = sqlite3_value_int64(argv[0]); + spellfix1DbExec(&rc, db, "DELETE FROM \"%w\".\"%w_vocab\" " + " WHERE id=%lld", + p->zDbName, p->zTableName, rowid); + }else{ + const unsigned char *zWord = sqlite3_value_text(argv[SPELLFIX_COL_WORD+2]); + int nWord = sqlite3_value_bytes(argv[SPELLFIX_COL_WORD+2]); + int iLang = sqlite3_value_int(argv[SPELLFIX_COL_LANGID+2]); + int iRank = sqlite3_value_int(argv[SPELLFIX_COL_RANK+2]); + const unsigned char *zSoundslike = + sqlite3_value_text(argv[SPELLFIX_COL_SOUNDSLIKE+2]); + int nSoundslike = sqlite3_value_bytes(argv[SPELLFIX_COL_SOUNDSLIKE+2]); + char *zK1, *zK2; + int i; + char c; + + if( zWord==0 ){ + /* Inserts of the form: INSERT INTO table(command) VALUES('xyzzy'); + ** cause zWord to be NULL, so we look at the "command" column to see + ** what special actions to take */ + const char *zCmd = + (const char*)sqlite3_value_text(argv[SPELLFIX_COL_COMMAND+2]); + if( zCmd==0 ){ + pVTab->zErrMsg = sqlite3_mprintf("%s.word may not be NULL", + p->zTableName); + return SQLITE_CONSTRAINT_NOTNULL; + } + if( strcmp(zCmd,"reset")==0 ){ + /* Reset the edit cost table (if there is one). */ + editDist3ConfigDelete(p->pConfig3); + p->pConfig3 = 0; + return SQLITE_OK; + } + if( strncmp(zCmd,"edit_cost_table=",16)==0 ){ + editDist3ConfigDelete(p->pConfig3); + p->pConfig3 = 0; + sqlite3_free(p->zCostTable); + p->zCostTable = spellfix1Dequote(zCmd+16); + if( p->zCostTable==0 ) return SQLITE_NOMEM; + if( p->zCostTable[0]==0 || sqlite3_stricmp(p->zCostTable,"null")==0 ){ + sqlite3_free(p->zCostTable); + p->zCostTable = 0; + } + return SQLITE_OK; + } + pVTab->zErrMsg = sqlite3_mprintf("unknown value for %s.command: \"%w\"", + p->zTableName, zCmd); + return SQLITE_ERROR; + } + if( iRank<1 ) iRank = 1; + if( zSoundslike ){ + zK1 = (char*)transliterate(zSoundslike, nSoundslike); + }else{ + zK1 = (char*)transliterate(zWord, nWord); + } + if( zK1==0 ) return SQLITE_NOMEM; + for(i=0; (c = zK1[i])!=0; i++){ + if( c>='A' && c<='Z' ) zK1[i] += 'a' - 'A'; + } + zK2 = (char*)phoneticHash((const unsigned char*)zK1, i); + if( zK2==0 ){ + sqlite3_free(zK1); + return SQLITE_NOMEM; + } + if( sqlite3_value_type(argv[0])==SQLITE_NULL ){ + spellfix1DbExec(&rc, db, + "INSERT INTO \"%w\".\"%w_vocab\"(rank,langid,word,k1,k2) " + "VALUES(%d,%d,%Q,%Q,%Q)", + p->zDbName, p->zTableName, + iRank, iLang, zWord, zK1, zK2 + ); + *pRowid = sqlite3_last_insert_rowid(db); + }else{ + rowid = sqlite3_value_int64(argv[0]); + newRowid = *pRowid = sqlite3_value_int64(argv[1]); + spellfix1DbExec(&rc, db, + "UPDATE \"%w\".\"%w_vocab\" SET id=%lld, rank=%d, langid=%d," + " word=%Q, k1=%Q, k2=%Q WHERE id=%lld", + p->zDbName, p->zTableName, newRowid, iRank, iLang, + zWord, zK1, zK2, rowid + ); + } + sqlite3_free(zK1); + sqlite3_free(zK2); + } + return rc; +} + +/* +** Rename the spellfix1 table. +*/ +static int spellfix1Rename(sqlite3_vtab *pVTab, const char *zNew){ + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + sqlite3 *db = p->db; + int rc = SQLITE_OK; + char *zNewName = sqlite3_mprintf("%s", zNew); + if( zNewName==0 ){ + return SQLITE_NOMEM; + } + spellfix1DbExec(&rc, db, + "ALTER TABLE \"%w\".\"%w_vocab\" RENAME TO \"%w_vocab\"", + p->zDbName, p->zTableName, zNewName + ); + if( rc==SQLITE_OK ){ + sqlite3_free(p->zTableName); + p->zTableName = zNewName; + }else{ + sqlite3_free(zNewName); + } + return rc; +} + + +/* +** A virtual table module that provides fuzzy search. +*/ +static sqlite3_module spellfix1Module = { + 0, /* iVersion */ + spellfix1Create, /* xCreate - handle CREATE VIRTUAL TABLE */ + spellfix1Connect, /* xConnect - reconnected to an existing table */ + spellfix1BestIndex, /* xBestIndex - figure out how to do a query */ + spellfix1Disconnect, /* xDisconnect - close a connection */ + spellfix1Destroy, /* xDestroy - handle DROP TABLE */ + spellfix1Open, /* xOpen - open a cursor */ + spellfix1Close, /* xClose - close a cursor */ + spellfix1Filter, /* xFilter - configure scan constraints */ + spellfix1Next, /* xNext - advance a cursor */ + spellfix1Eof, /* xEof - check for end of scan */ + spellfix1Column, /* xColumn - read data */ + spellfix1Rowid, /* xRowid - read data */ + spellfix1Update, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + spellfix1Rename, /* xRename */ +}; + +/* +** Register the various functions and the virtual table. +*/ +static int spellfix1Register(sqlite3 *db){ + int rc = SQLITE_OK; + int i; + rc = sqlite3_create_function(db, "spellfix1_translit", 1, SQLITE_UTF8, 0, + transliterateSqlFunc, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "spellfix1_editdist", 2, SQLITE_UTF8, 0, + editdistSqlFunc, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "spellfix1_phonehash", 1, SQLITE_UTF8, 0, + phoneticHashSqlFunc, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "spellfix1_scriptcode", 1, SQLITE_UTF8, 0, + scriptCodeSqlFunc, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_module(db, "spellfix1", &spellfix1Module, 0); + } + if( rc==SQLITE_OK ){ + rc = editDist3Install(db); + } + + /* Verify sanity of the translit[] table */ + for(i=0; i