--- origsrc/sqlite-autoconf-3100000/Makefile.am 2016-01-06 13:03:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/Makefile.am 2016-01-06 16:30:03.315880000 +0100 @@ -1,16 +1,75 @@ -AM_CFLAGS = @THREADSAFE_FLAGS@ @DYNAMIC_EXTENSION_FLAGS@ @FTS5_FLAGS@ @JSON1_FLAGS@ -DSQLITE_ENABLE_FTS3 -DSQLITE_ENABLE_RTREE +AM_CFLAGS = @THREADSAFE_FLAGS@ @DYNAMIC_EXTENSION_FLAGS@ @FTS5_FLAGS@ @JSON1_FLAGS@ -DUSE_SYSTEM_SQLITE=1 -DSQLITE_ENABLE_COLUMN_METADATA=1 -DSQLITE_ENABLE_DBSTAT_VTAB=1 -DSQLITE_ENABLE_FTS3_PARENTHESIS=1 -DSQLITE_ENABLE_FTS4=1 -DSQLITE_ENABLE_RTREE=1 -DSQLITE_ENABLE_UPDATE_DELETE_LIMIT=1 -DSQLITE_OMIT_LOOKASIDE=1 -DSQLITE_SECURE_DELETE=1 -lib_LTLIBRARIES = libsqlite3.la +lib_LTLIBRARIES = libsqlite3.la libsqlite3amatch.la libsqlite3closure.la \ + libsqlite3compress.la libsqlite3eval.la libsqlite3fileio.la \ + libsqlite3fuzzer.la libsqlite3icu.la \ + libsqlite3ieee754.la libsqlite3nextchar.la \ + libsqlite3percentile.la libsqlite3rbu.la libsqlite3regexp.la \ + libsqlite3rot13.la libsqlite3series.la libsqlite3showauth.la \ + libsqlite3spellfix.la libsqlite3totype.la libsqlite3vfslog.la \ + libsqlite3vtshim.la libsqlite3wholenumber.la libsqlite3_la_SOURCES = sqlite3.c -libsqlite3_la_LDFLAGS = -no-undefined -version-info 8:6:8 +libsqlite3_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3amatch_la_SOURCES = amatch.c +libsqlite3amatch_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3closure_la_SOURCES = closure.c +libsqlite3closure_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3compress_la_SOURCES = compress.c +libsqlite3compress_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc -lz +libsqlite3eval_la_SOURCES = eval.c +libsqlite3eval_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3fileio_la_SOURCES = fileio.c +libsqlite3fileio_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3fuzzer_la_SOURCES = fuzzer.c +libsqlite3fuzzer_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3icu_la_SOURCES = icu.c +libsqlite3icu_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc -licui18n -licuuc +libsqlite3ieee754_la_SOURCES = ieee754.c +libsqlite3ieee754_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3nextchar_la_SOURCES = nextchar.c +libsqlite3nextchar_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3percentile_la_SOURCES = percentile.c +libsqlite3percentile_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3rbu_la_SOURCES = sqlite3rbu.c +libsqlite3rbu_la_LIBADD = $(top_builddir)/libsqlite3.la +libsqlite3rbu_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3rbu_la_DEPENDENCIES = $(top_builddir)/libsqlite3.la +libsqlite3regexp_la_SOURCES = regexp.c +libsqlite3regexp_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3rot13_la_SOURCES = rot13.c +libsqlite3rot13_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3series_la_SOURCES = series.c +libsqlite3series_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3showauth_la_SOURCES = showauth.c +libsqlite3showauth_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3spellfix_la_SOURCES = spellfix.c +libsqlite3spellfix_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3totype_la_SOURCES = totype.c +libsqlite3totype_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3vfslog_la_SOURCES = vfslog.c +libsqlite3vfslog_la_LIBADD = $(top_builddir)/libsqlite3.la +libsqlite3vfslog_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3vfslog_la_DEPENDENCIES = $(top_builddir)/libsqlite3.la +libsqlite3vtshim_la_SOURCES = vtshim.c +libsqlite3vtshim_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc +libsqlite3wholenumber_la_SOURCES = wholenumber.c +libsqlite3wholenumber_la_LDFLAGS = -no-undefined -version-info 8:6:8 -static-libgcc -bin_PROGRAMS = sqlite3 +bin_PROGRAMS = sqlite3 sqldiff rbu sqlite3_SOURCES = shell.c sqlite3.h EXTRA_sqlite3_SOURCES = sqlite3.c -sqlite3_LDADD = @EXTRA_SHELL_OBJ@ @READLINE_LIBS@ -sqlite3_DEPENDENCIES = @EXTRA_SHELL_OBJ@ +sqlite3_LDADD = $(top_builddir)/libsqlite3vfslog.la @EXTRA_SHELL_OBJ@ @READLINE_LIBS@ +sqlite3_DEPENDENCIES = $(top_builddir)/libsqlite3vfslog.la @EXTRA_SHELL_OBJ@ sqlite3_CFLAGS = $(AM_CFLAGS) +sqldiff_SOURCES = sqldiff.c sqlite3.h +sqldiff_LDADD = @EXTRA_SHELL_OBJ@ +sqldiff_DEPENDENCIES = @EXTRA_SHELL_OBJ@ +sqldiff_CFLAGS = $(AM_CFLAGS) +rbu_SOURCES = rbu.c sqlite3.h +rbu_LDADD = $(top_builddir)/libsqlite3rbu.la @EXTRA_SHELL_OBJ@ +rbu_DEPENDENCIES = $(top_builddir)/libsqlite3rbu.la @EXTRA_SHELL_OBJ@ +rbu_CFLAGS = $(AM_CFLAGS) include_HEADERS = sqlite3.h sqlite3ext.h --- origsrc/sqlite-autoconf-3100000/amatch.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/amatch.c 2016-01-06 16:30:42.469119500 +0100 @@ -0,0 +1,1520 @@ +/* +** 2013-03-14 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** This file contains code for a demonstration virtual table that finds +** "approximate matches" - strings from a finite set that are nearly the +** same as a single input string. The virtual table is called "amatch". +** +** A amatch virtual table is created like this: +** +** CREATE VIRTUAL TABLE f USING approximate_match( +** vocabulary_table=, -- V +** vocabulary_word=, -- W +** vocabulary_language=, -- L +** edit_distances= +** ); +** +** When it is created, the new amatch table must be supplied with the +** the name of a table V and columns V.W and V.L such that +** +** SELECT W FROM V WHERE L=$language +** +** returns the allowed vocabulary for the match. If the "vocabulary_language" +** or L columnname is left unspecified or is an empty string, then no +** filtering of the vocabulary by language is performed. +** +** For efficiency, it is essential that the vocabulary table be indexed: +** +** CREATE vocab_index ON V(W) +** +** A separate edit-cost-table provides scoring information that defines +** what it means for one string to be "close" to another. +** +** The edit-cost-table must contain exactly four columns (more precisely, +** the statement "SELECT * FROM " must return records +** that consist of four columns). It does not matter what the columns are +** named. +** +** Each row in the edit-cost-table represents a single character +** transformation going from user input to the vocabulary. The leftmost +** column of the row (column 0) contains an integer identifier of the +** language to which the transformation rule belongs (see "MULTIPLE LANGUAGES" +** below). The second column of the row (column 1) contains the input +** character or characters - the characters of user input. The third +** column contains characters as they appear in the vocabulary table. +** And the fourth column contains the integer cost of making the +** transformation. For example: +** +** CREATE TABLE f_data(iLang, cFrom, cTo, Cost); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '', 'a', 100); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, 'b', '', 87); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, 'o', 'oe', 38); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, 'oe', 'o', 40); +** +** The first row inserted into the edit-cost-table by the SQL script +** above indicates that the cost of having an extra 'a' in the vocabulary +** table that is missing in the user input 100. (All costs are integers. +** Overall cost must not exceed 16777216.) The second INSERT statement +** creates a rule saying that the cost of having a single letter 'b' in +** user input which is missing in the vocabulary table is 87. The third +** INSERT statement mean that the cost of matching an 'o' in user input +** against an 'oe' in the vocabulary table is 38. And so forth. +** +** The following rules are special: +** +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '?', '', 97); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '', '?', 98); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '?', '?', 99); +** +** The '?' to '' rule is the cost of having any single character in the input +** that is not found in the vocabular. The '' to '?' rule is the cost of +** having a character in the vocabulary table that is missing from input. +** And the '?' to '?' rule is the cost of doing an arbitrary character +** substitution. These three generic rules apply across all languages. +** In other words, the iLang field is ignored for the generic substitution +** rules. If more than one cost is given for a generic substitution rule, +** then the lowest cost is used. +** +** Once it has been created, the amatch virtual table can be queried +** as follows: +** +** SELECT word, distance FROM f +** WHERE word MATCH 'abcdefg' +** AND distance<200; +** +** This query outputs the strings contained in the T(F) field that +** are close to "abcdefg" and in order of increasing distance. No string +** is output more than once. If there are multiple ways to transform the +** target string ("abcdefg") into a string in the vocabulary table then +** the lowest cost transform is the one that is returned. In this example, +** the search is limited to strings with a total distance of less than 200. +** +** For efficiency, it is important to put tight bounds on the distance. +** The time and memory space needed to perform this query is exponential +** in the maximum distance. A good rule of thumb is to limit the distance +** to no more than 1.5 or 2 times the maximum cost of any rule in the +** edit-cost-table. +** +** The amatch is a read-only table. Any attempt to DELETE, INSERT, or +** UPDATE on a amatch table will throw an error. +** +** It is important to put some kind of a limit on the amatch output. This +** can be either in the form of a LIMIT clause at the end of the query, +** or better, a "distance +#include +#include +#include +#include + +#ifndef SQLITE_OMIT_VIRTUALTABLE + +/* +** Forward declaration of objects used by this implementation +*/ +typedef struct amatch_vtab amatch_vtab; +typedef struct amatch_cursor amatch_cursor; +typedef struct amatch_rule amatch_rule; +typedef struct amatch_word amatch_word; +typedef struct amatch_avl amatch_avl; + + +/***************************************************************************** +** AVL Tree implementation +*/ +/* +** Objects that want to be members of the AVL tree should embedded an +** instance of this structure. +*/ +struct amatch_avl { + amatch_word *pWord; /* Points to the object being stored in the tree */ + char *zKey; /* Key. zero-terminated string. Must be unique */ + amatch_avl *pBefore; /* Other elements less than zKey */ + amatch_avl *pAfter; /* Other elements greater than zKey */ + amatch_avl *pUp; /* Parent element */ + short int height; /* Height of this node. Leaf==1 */ + short int imbalance; /* Height difference between pBefore and pAfter */ +}; + +/* Recompute the amatch_avl.height and amatch_avl.imbalance fields for p. +** Assume that the children of p have correct heights. +*/ +static void amatchAvlRecomputeHeight(amatch_avl *p){ + short int hBefore = p->pBefore ? p->pBefore->height : 0; + short int hAfter = p->pAfter ? p->pAfter->height : 0; + p->imbalance = hBefore - hAfter; /* -: pAfter higher. +: pBefore higher */ + p->height = (hBefore>hAfter ? hBefore : hAfter)+1; +} + +/* +** P B +** / \ / \ +** B Z ==> X P +** / \ / \ +** X Y Y Z +** +*/ +static amatch_avl *amatchAvlRotateBefore(amatch_avl *pP){ + amatch_avl *pB = pP->pBefore; + amatch_avl *pY = pB->pAfter; + pB->pUp = pP->pUp; + pB->pAfter = pP; + pP->pUp = pB; + pP->pBefore = pY; + if( pY ) pY->pUp = pP; + amatchAvlRecomputeHeight(pP); + amatchAvlRecomputeHeight(pB); + return pB; +} + +/* +** P A +** / \ / \ +** X A ==> P Z +** / \ / \ +** Y Z X Y +** +*/ +static amatch_avl *amatchAvlRotateAfter(amatch_avl *pP){ + amatch_avl *pA = pP->pAfter; + amatch_avl *pY = pA->pBefore; + pA->pUp = pP->pUp; + pA->pBefore = pP; + pP->pUp = pA; + pP->pAfter = pY; + if( pY ) pY->pUp = pP; + amatchAvlRecomputeHeight(pP); + amatchAvlRecomputeHeight(pA); + return pA; +} + +/* +** Return a pointer to the pBefore or pAfter pointer in the parent +** of p that points to p. Or if p is the root node, return pp. +*/ +static amatch_avl **amatchAvlFromPtr(amatch_avl *p, amatch_avl **pp){ + amatch_avl *pUp = p->pUp; + if( pUp==0 ) return pp; + if( pUp->pAfter==p ) return &pUp->pAfter; + return &pUp->pBefore; +} + +/* +** Rebalance all nodes starting with p and working up to the root. +** Return the new root. +*/ +static amatch_avl *amatchAvlBalance(amatch_avl *p){ + amatch_avl *pTop = p; + amatch_avl **pp; + while( p ){ + amatchAvlRecomputeHeight(p); + if( p->imbalance>=2 ){ + amatch_avl *pB = p->pBefore; + if( pB->imbalance<0 ) p->pBefore = amatchAvlRotateAfter(pB); + pp = amatchAvlFromPtr(p,&p); + p = *pp = amatchAvlRotateBefore(p); + }else if( p->imbalance<=(-2) ){ + amatch_avl *pA = p->pAfter; + if( pA->imbalance>0 ) p->pAfter = amatchAvlRotateBefore(pA); + pp = amatchAvlFromPtr(p,&p); + p = *pp = amatchAvlRotateAfter(p); + } + pTop = p; + p = p->pUp; + } + return pTop; +} + +/* Search the tree rooted at p for an entry with zKey. Return a pointer +** to the entry or return NULL. +*/ +static amatch_avl *amatchAvlSearch(amatch_avl *p, const char *zKey){ + int c; + while( p && (c = strcmp(zKey, p->zKey))!=0 ){ + p = (c<0) ? p->pBefore : p->pAfter; + } + return p; +} + +/* Find the first node (the one with the smallest key). +*/ +static amatch_avl *amatchAvlFirst(amatch_avl *p){ + if( p ) while( p->pBefore ) p = p->pBefore; + return p; +} + +#if 0 /* NOT USED */ +/* Return the node with the next larger key after p. +*/ +static amatch_avl *amatchAvlNext(amatch_avl *p){ + amatch_avl *pPrev = 0; + while( p && p->pAfter==pPrev ){ + pPrev = p; + p = p->pUp; + } + if( p && pPrev==0 ){ + p = amatchAvlFirst(p->pAfter); + } + return p; +} +#endif + +#if 0 /* NOT USED */ +/* Verify AVL tree integrity +*/ +static int amatchAvlIntegrity(amatch_avl *pHead){ + amatch_avl *p; + if( pHead==0 ) return 1; + if( (p = pHead->pBefore)!=0 ){ + assert( p->pUp==pHead ); + assert( amatchAvlIntegrity(p) ); + assert( strcmp(p->zKey, pHead->zKey)<0 ); + while( p->pAfter ) p = p->pAfter; + assert( strcmp(p->zKey, pHead->zKey)<0 ); + } + if( (p = pHead->pAfter)!=0 ){ + assert( p->pUp==pHead ); + assert( amatchAvlIntegrity(p) ); + assert( strcmp(p->zKey, pHead->zKey)>0 ); + p = amatchAvlFirst(p); + assert( strcmp(p->zKey, pHead->zKey)>0 ); + } + return 1; +} +static int amatchAvlIntegrity2(amatch_avl *pHead){ + amatch_avl *p, *pNext; + for(p=amatchAvlFirst(pHead); p; p=pNext){ + pNext = amatchAvlNext(p); + if( pNext==0 ) break; + assert( strcmp(p->zKey, pNext->zKey)<0 ); + } + return 1; +} +#endif + +/* Insert a new node pNew. Return NULL on success. If the key is not +** unique, then do not perform the insert but instead leave pNew unchanged +** and return a pointer to an existing node with the same key. +*/ +static amatch_avl *amatchAvlInsert(amatch_avl **ppHead, amatch_avl *pNew){ + int c; + amatch_avl *p = *ppHead; + if( p==0 ){ + p = pNew; + pNew->pUp = 0; + }else{ + while( p ){ + c = strcmp(pNew->zKey, p->zKey); + if( c<0 ){ + if( p->pBefore ){ + p = p->pBefore; + }else{ + p->pBefore = pNew; + pNew->pUp = p; + break; + } + }else if( c>0 ){ + if( p->pAfter ){ + p = p->pAfter; + }else{ + p->pAfter = pNew; + pNew->pUp = p; + break; + } + }else{ + return p; + } + } + } + pNew->pBefore = 0; + pNew->pAfter = 0; + pNew->height = 1; + pNew->imbalance = 0; + *ppHead = amatchAvlBalance(p); + /* assert( amatchAvlIntegrity(*ppHead) ); */ + /* assert( amatchAvlIntegrity2(*ppHead) ); */ + return 0; +} + +/* Remove node pOld from the tree. pOld must be an element of the tree or +** the AVL tree will become corrupt. +*/ +static void amatchAvlRemove(amatch_avl **ppHead, amatch_avl *pOld){ + amatch_avl **ppParent; + amatch_avl *pBalance = 0; + /* assert( amatchAvlSearch(*ppHead, pOld->zKey)==pOld ); */ + ppParent = amatchAvlFromPtr(pOld, ppHead); + if( pOld->pBefore==0 && pOld->pAfter==0 ){ + *ppParent = 0; + pBalance = pOld->pUp; + }else if( pOld->pBefore && pOld->pAfter ){ + amatch_avl *pX, *pY; + pX = amatchAvlFirst(pOld->pAfter); + *amatchAvlFromPtr(pX, 0) = pX->pAfter; + if( pX->pAfter ) pX->pAfter->pUp = pX->pUp; + pBalance = pX->pUp; + pX->pAfter = pOld->pAfter; + if( pX->pAfter ){ + pX->pAfter->pUp = pX; + }else{ + assert( pBalance==pOld ); + pBalance = pX; + } + pX->pBefore = pY = pOld->pBefore; + if( pY ) pY->pUp = pX; + pX->pUp = pOld->pUp; + *ppParent = pX; + }else if( pOld->pBefore==0 ){ + *ppParent = pBalance = pOld->pAfter; + pBalance->pUp = pOld->pUp; + }else if( pOld->pAfter==0 ){ + *ppParent = pBalance = pOld->pBefore; + pBalance->pUp = pOld->pUp; + } + *ppHead = amatchAvlBalance(pBalance); + pOld->pUp = 0; + pOld->pBefore = 0; + pOld->pAfter = 0; + /* assert( amatchAvlIntegrity(*ppHead) ); */ + /* assert( amatchAvlIntegrity2(*ppHead) ); */ +} +/* +** End of the AVL Tree implementation +******************************************************************************/ + + +/* +** Various types. +** +** amatch_cost is the "cost" of an edit operation. +** +** amatch_len is the length of a matching string. +** +** amatch_langid is an ruleset identifier. +*/ +typedef int amatch_cost; +typedef signed char amatch_len; +typedef int amatch_langid; + +/* +** Limits +*/ +#define AMATCH_MX_LENGTH 50 /* Maximum length of a rule string */ +#define AMATCH_MX_LANGID 2147483647 /* Maximum rule ID */ +#define AMATCH_MX_COST 1000 /* Maximum single-rule cost */ + +/* +** A match or partial match +*/ +struct amatch_word { + amatch_word *pNext; /* Next on a list of all amatch_words */ + amatch_avl sCost; /* Linkage of this node into the cost tree */ + amatch_avl sWord; /* Linkage of this node into the word tree */ + amatch_cost rCost; /* Cost of the match so far */ + int iSeq; /* Sequence number */ + char zCost[10]; /* Cost key (text rendering of rCost) */ + short int nMatch; /* Input characters matched */ + char zWord[4]; /* Text of the word. Extra space appended as needed */ +}; + +/* +** Each transformation rule is stored as an instance of this object. +** All rules are kept on a linked list sorted by rCost. +*/ +struct amatch_rule { + amatch_rule *pNext; /* Next rule in order of increasing rCost */ + char *zFrom; /* Transform from (a string from user input) */ + amatch_cost rCost; /* Cost of this transformation */ + amatch_langid iLang; /* The langauge to which this rule belongs */ + amatch_len nFrom, nTo; /* Length of the zFrom and zTo strings */ + char zTo[4]; /* Tranform to V.W value (extra space appended) */ +}; + +/* +** A amatch virtual-table object +*/ +struct amatch_vtab { + sqlite3_vtab base; /* Base class - must be first */ + char *zClassName; /* Name of this class. Default: "amatch" */ + char *zDb; /* Name of database. (ex: "main") */ + char *zSelf; /* Name of this virtual table */ + char *zCostTab; /* Name of edit-cost-table */ + char *zVocabTab; /* Name of vocabulary table */ + char *zVocabWord; /* Name of vocabulary table word column */ + char *zVocabLang; /* Name of vocabulary table language column */ + amatch_rule *pRule; /* All active rules in this amatch */ + amatch_cost rIns; /* Generic insertion cost '' -> ? */ + amatch_cost rDel; /* Generic deletion cost ? -> '' */ + amatch_cost rSub; /* Generic substitution cost ? -> ? */ + sqlite3 *db; /* The database connection */ + sqlite3_stmt *pVCheck; /* Query to check zVocabTab */ + int nCursor; /* Number of active cursors */ +}; + +/* A amatch cursor object */ +struct amatch_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + sqlite3_int64 iRowid; /* The rowid of the current word */ + amatch_langid iLang; /* Use this language ID */ + amatch_cost rLimit; /* Maximum cost of any term */ + int nBuf; /* Space allocated for zBuf */ + int oomErr; /* True following an OOM error */ + int nWord; /* Number of amatch_word objects */ + char *zBuf; /* Temp-use buffer space */ + char *zInput; /* Input word to match against */ + amatch_vtab *pVtab; /* The virtual table this cursor belongs to */ + amatch_word *pAllWords; /* List of all amatch_word objects */ + amatch_word *pCurrent; /* Most recent solution */ + amatch_avl *pCost; /* amatch_word objects keyed by iCost */ + amatch_avl *pWord; /* amatch_word objects keyed by zWord */ +}; + +/* +** The two input rule lists are both sorted in order of increasing +** cost. Merge them together into a single list, sorted by cost, and +** return a pointer to the head of that list. +*/ +static amatch_rule *amatchMergeRules(amatch_rule *pA, amatch_rule *pB){ + amatch_rule head; + amatch_rule *pTail; + + pTail = &head; + while( pA && pB ){ + if( pA->rCost<=pB->rCost ){ + pTail->pNext = pA; + pTail = pA; + pA = pA->pNext; + }else{ + pTail->pNext = pB; + pTail = pB; + pB = pB->pNext; + } + } + if( pA==0 ){ + pTail->pNext = pB; + }else{ + pTail->pNext = pA; + } + return head.pNext; +} + +/* +** Statement pStmt currently points to a row in the amatch data table. This +** function allocates and populates a amatch_rule structure according to +** the content of the row. +** +** If successful, *ppRule is set to point to the new object and SQLITE_OK +** is returned. Otherwise, *ppRule is zeroed, *pzErr may be set to point +** to an error message and an SQLite error code returned. +*/ +static int amatchLoadOneRule( + amatch_vtab *p, /* Fuzzer virtual table handle */ + sqlite3_stmt *pStmt, /* Base rule on statements current row */ + amatch_rule **ppRule, /* OUT: New rule object */ + char **pzErr /* OUT: Error message */ +){ + sqlite3_int64 iLang = sqlite3_column_int64(pStmt, 0); + const char *zFrom = (const char *)sqlite3_column_text(pStmt, 1); + const char *zTo = (const char *)sqlite3_column_text(pStmt, 2); + amatch_cost rCost = sqlite3_column_int(pStmt, 3); + + int rc = SQLITE_OK; /* Return code */ + int nFrom; /* Size of string zFrom, in bytes */ + int nTo; /* Size of string zTo, in bytes */ + amatch_rule *pRule = 0; /* New rule object to return */ + + if( zFrom==0 ) zFrom = ""; + if( zTo==0 ) zTo = ""; + nFrom = (int)strlen(zFrom); + nTo = (int)strlen(zTo); + + /* Silently ignore null transformations */ + if( strcmp(zFrom, zTo)==0 ){ + if( zFrom[0]=='?' && zFrom[1]==0 ){ + if( p->rSub==0 || p->rSub>rCost ) p->rSub = rCost; + } + *ppRule = 0; + return SQLITE_OK; + } + + if( rCost<=0 || rCost>AMATCH_MX_COST ){ + *pzErr = sqlite3_mprintf("%s: cost must be between 1 and %d", + p->zClassName, AMATCH_MX_COST + ); + rc = SQLITE_ERROR; + }else + if( nFrom>AMATCH_MX_LENGTH || nTo>AMATCH_MX_LENGTH ){ + *pzErr = sqlite3_mprintf("%s: maximum string length is %d", + p->zClassName, AMATCH_MX_LENGTH + ); + rc = SQLITE_ERROR; + }else + if( iLang<0 || iLang>AMATCH_MX_LANGID ){ + *pzErr = sqlite3_mprintf("%s: iLang must be between 0 and %d", + p->zClassName, AMATCH_MX_LANGID + ); + rc = SQLITE_ERROR; + }else + if( strcmp(zFrom,"")==0 && strcmp(zTo,"?")==0 ){ + if( p->rIns==0 || p->rIns>rCost ) p->rIns = rCost; + }else + if( strcmp(zFrom,"?")==0 && strcmp(zTo,"")==0 ){ + if( p->rDel==0 || p->rDel>rCost ) p->rDel = rCost; + }else + { + pRule = sqlite3_malloc( sizeof(*pRule) + nFrom + nTo ); + if( pRule==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pRule, 0, sizeof(*pRule)); + pRule->zFrom = &pRule->zTo[nTo+1]; + pRule->nFrom = nFrom; + memcpy(pRule->zFrom, zFrom, nFrom+1); + memcpy(pRule->zTo, zTo, nTo+1); + pRule->nTo = nTo; + pRule->rCost = rCost; + pRule->iLang = (int)iLang; + } + } + + *ppRule = pRule; + return rc; +} + +/* +** Free all the content in the edit-cost-table +*/ +static void amatchFreeRules(amatch_vtab *p){ + while( p->pRule ){ + amatch_rule *pRule = p->pRule; + p->pRule = pRule->pNext; + sqlite3_free(pRule); + } + p->pRule = 0; +} + +/* +** Load the content of the amatch data table into memory. +*/ +static int amatchLoadRules( + sqlite3 *db, /* Database handle */ + amatch_vtab *p, /* Virtual amatch table to configure */ + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; /* Return code */ + char *zSql; /* SELECT used to read from rules table */ + amatch_rule *pHead = 0; + + zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", p->zDb, p->zCostTab); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + int rc2; /* finalize() return code */ + sqlite3_stmt *pStmt = 0; + rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); + if( rc!=SQLITE_OK ){ + *pzErr = sqlite3_mprintf("%s: %s", p->zClassName, sqlite3_errmsg(db)); + }else if( sqlite3_column_count(pStmt)!=4 ){ + *pzErr = sqlite3_mprintf("%s: %s has %d columns, expected 4", + p->zClassName, p->zCostTab, sqlite3_column_count(pStmt) + ); + rc = SQLITE_ERROR; + }else{ + while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ + amatch_rule *pRule = 0; + rc = amatchLoadOneRule(p, pStmt, &pRule, pzErr); + if( pRule ){ + pRule->pNext = pHead; + pHead = pRule; + } + } + } + rc2 = sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ) rc = rc2; + } + sqlite3_free(zSql); + + /* All rules are now in a singly linked list starting at pHead. This + ** block sorts them by cost and then sets amatch_vtab.pRule to point to + ** point to the head of the sorted list. + */ + if( rc==SQLITE_OK ){ + unsigned int i; + amatch_rule *pX; + amatch_rule *a[15]; + for(i=0; ipNext; + pX->pNext = 0; + for(i=0; a[i] && ipRule = amatchMergeRules(p->pRule, pX); + }else{ + /* An error has occurred. Setting p->pRule to point to the head of the + ** allocated list ensures that the list will be cleaned up in this case. + */ + assert( p->pRule==0 ); + p->pRule = pHead; + } + + return rc; +} + +/* +** This function converts an SQL quoted string into an unquoted string +** and returns a pointer to a buffer allocated using sqlite3_malloc() +** containing the result. The caller should eventually free this buffer +** using sqlite3_free. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static char *amatchDequote(const char *zIn){ + int nIn; /* Size of input string, in bytes */ + char *zOut; /* Output (dequoted) string */ + + nIn = (int)strlen(zIn); + zOut = sqlite3_malloc(nIn+1); + if( zOut ){ + char q = zIn[0]; /* Quote character (if any ) */ + + if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ + memcpy(zOut, zIn, nIn+1); + }else{ + int iOut = 0; /* Index of next byte to write to output */ + int iIn; /* Index of next byte to read from input */ + + if( q=='[' ) q = ']'; + for(iIn=1; iInpVCheck ){ + sqlite3_finalize(p->pVCheck); + p->pVCheck = 0; + } +} + +/* +** Deallocate an amatch_vtab object +*/ +static void amatchFree(amatch_vtab *p){ + if( p ){ + amatchFreeRules(p); + amatchVCheckClear(p); + sqlite3_free(p->zClassName); + sqlite3_free(p->zDb); + sqlite3_free(p->zCostTab); + sqlite3_free(p->zVocabTab); + sqlite3_free(p->zVocabWord); + sqlite3_free(p->zVocabLang); + sqlite3_free(p->zSelf); + memset(p, 0, sizeof(*p)); + sqlite3_free(p); + } +} + +/* +** xDisconnect/xDestroy method for the amatch module. +*/ +static int amatchDisconnect(sqlite3_vtab *pVtab){ + amatch_vtab *p = (amatch_vtab*)pVtab; + assert( p->nCursor==0 ); + amatchFree(p); + return SQLITE_OK; +} + +/* +** Check to see if the argument is of the form: +** +** KEY = VALUE +** +** If it is, return a pointer to the first character of VALUE. +** If not, return NULL. Spaces around the = are ignored. +*/ +static const char *amatchValueOfKey(const char *zKey, const char *zStr){ + int nKey = (int)strlen(zKey); + int nStr = (int)strlen(zStr); + int i; + if( nStr module name ("approximate_match") +** argv[1] -> database name +** argv[2] -> table name +** argv[3...] -> arguments +*/ +static int amatchConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + int rc = SQLITE_OK; /* Return code */ + amatch_vtab *pNew = 0; /* New virtual table */ + const char *zModule = argv[0]; + const char *zDb = argv[1]; + const char *zVal; + int i; + + (void)pAux; + *ppVtab = 0; + pNew = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + rc = SQLITE_NOMEM; + memset(pNew, 0, sizeof(*pNew)); + pNew->db = db; + pNew->zClassName = sqlite3_mprintf("%s", zModule); + if( pNew->zClassName==0 ) goto amatchConnectError; + pNew->zDb = sqlite3_mprintf("%s", zDb); + if( pNew->zDb==0 ) goto amatchConnectError; + pNew->zSelf = sqlite3_mprintf("%s", argv[2]); + if( pNew->zSelf==0 ) goto amatchConnectError; + for(i=3; izVocabTab); + pNew->zVocabTab = amatchDequote(zVal); + if( pNew->zVocabTab==0 ) goto amatchConnectError; + continue; + } + zVal = amatchValueOfKey("vocabulary_word", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zVocabWord); + pNew->zVocabWord = amatchDequote(zVal); + if( pNew->zVocabWord==0 ) goto amatchConnectError; + continue; + } + zVal = amatchValueOfKey("vocabulary_language", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zVocabLang); + pNew->zVocabLang = amatchDequote(zVal); + if( pNew->zVocabLang==0 ) goto amatchConnectError; + continue; + } + zVal = amatchValueOfKey("edit_distances", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zCostTab); + pNew->zCostTab = amatchDequote(zVal); + if( pNew->zCostTab==0 ) goto amatchConnectError; + continue; + } + *pzErr = sqlite3_mprintf("unrecognized argument: [%s]\n", argv[i]); + amatchFree(pNew); + *ppVtab = 0; + return SQLITE_ERROR; + } + rc = SQLITE_OK; + if( pNew->zCostTab==0 ){ + *pzErr = sqlite3_mprintf("no edit_distances table specified"); + rc = SQLITE_ERROR; + }else{ + rc = amatchLoadRules(db, pNew, pzErr); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_declare_vtab(db, + "CREATE TABLE x(word,distance,language," + "command HIDDEN,nword HIDDEN)" + ); +#define AMATCH_COL_WORD 0 +#define AMATCH_COL_DISTANCE 1 +#define AMATCH_COL_LANGUAGE 2 +#define AMATCH_COL_COMMAND 3 +#define AMATCH_COL_NWORD 4 + } + if( rc!=SQLITE_OK ){ + amatchFree(pNew); + } + *ppVtab = &pNew->base; + return rc; + +amatchConnectError: + amatchFree(pNew); + return rc; +} + +/* +** Open a new amatch cursor. +*/ +static int amatchOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + amatch_vtab *p = (amatch_vtab*)pVTab; + amatch_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVtab = p; + *ppCursor = &pCur->base; + p->nCursor++; + return SQLITE_OK; +} + +/* +** Free up all the memory allocated by a cursor. Set it rLimit to 0 +** to indicate that it is at EOF. +*/ +static void amatchClearCursor(amatch_cursor *pCur){ + amatch_word *pWord, *pNextWord; + for(pWord=pCur->pAllWords; pWord; pWord=pNextWord){ + pNextWord = pWord->pNext; + sqlite3_free(pWord); + } + pCur->pAllWords = 0; + sqlite3_free(pCur->zInput); + pCur->zInput = 0; + sqlite3_free(pCur->zBuf); + pCur->zBuf = 0; + pCur->nBuf = 0; + pCur->pCost = 0; + pCur->pWord = 0; + pCur->pCurrent = 0; + pCur->rLimit = 1000000; + pCur->iLang = 0; + pCur->nWord = 0; +} + +/* +** Close a amatch cursor. +*/ +static int amatchClose(sqlite3_vtab_cursor *cur){ + amatch_cursor *pCur = (amatch_cursor *)cur; + amatchClearCursor(pCur); + pCur->pVtab->nCursor--; + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Render a 24-bit unsigned integer as a 4-byte base-64 number. +*/ +static void amatchEncodeInt(int x, char *z){ + static const char a[] = + "0123456789" + "ABCDEFGHIJ" + "KLMNOPQRST" + "UVWXYZ^abc" + "defghijklm" + "nopqrstuvw" + "xyz~"; + z[0] = a[(x>>18)&0x3f]; + z[1] = a[(x>>12)&0x3f]; + z[2] = a[(x>>6)&0x3f]; + z[3] = a[x&0x3f]; +} + +/* +** Write the zCost[] field for a amatch_word object +*/ +static void amatchWriteCost(amatch_word *pWord){ + amatchEncodeInt(pWord->rCost, pWord->zCost); + amatchEncodeInt(pWord->iSeq, pWord->zCost+4); + pWord->zCost[8] = 0; +} + +/* Circumvent compiler warnings about the use of strcpy() by supplying +** our own implementation. +*/ +#if defined(__OpenBSD__) +static void amatchStrcpy(char *dest, const char *src){ + while( (*(dest++) = *(src++))!=0 ){} +} +static void amatchStrcat(char *dest, const char *src){ + while( *dest ) dest++; + amatchStrcpy(dest, src); +} +#else +# define amatchStrcpy strcpy +# define amatchStrcat strcat +#endif + + +/* +** Add a new amatch_word object to the queue. +** +** If a prior amatch_word object with the same zWord, and nMatch +** already exists, update its rCost (if the new rCost is less) but +** otherwise leave it unchanged. Do not add a duplicate. +** +** Do nothing if the cost exceeds threshold. +*/ +static void amatchAddWord( + amatch_cursor *pCur, + amatch_cost rCost, + int nMatch, + const char *zWordBase, + const char *zWordTail +){ + amatch_word *pWord; + amatch_avl *pNode; + amatch_avl *pOther; + int nBase, nTail; + char zBuf[4]; + + if( rCost>pCur->rLimit ){ + return; + } + nBase = (int)strlen(zWordBase); + nTail = (int)strlen(zWordTail); + if( nBase+nTail+3>pCur->nBuf ){ + pCur->nBuf = nBase+nTail+100; + pCur->zBuf = sqlite3_realloc(pCur->zBuf, pCur->nBuf); + if( pCur->zBuf==0 ){ + pCur->nBuf = 0; + return; + } + } + amatchEncodeInt(nMatch, zBuf); + memcpy(pCur->zBuf, zBuf+2, 2); + memcpy(pCur->zBuf+2, zWordBase, nBase); + memcpy(pCur->zBuf+2+nBase, zWordTail, nTail+1); + pNode = amatchAvlSearch(pCur->pWord, pCur->zBuf); + if( pNode ){ + pWord = pNode->pWord; + if( pWord->rCost>rCost ){ +#ifdef AMATCH_TRACE_1 + printf("UPDATE [%s][%.*s^%s] %d (\"%s\" \"%s\")\n", + pWord->zWord+2, pWord->nMatch, pCur->zInput, pCur->zInput, + pWord->rCost, pWord->zWord, pWord->zCost); +#endif + amatchAvlRemove(&pCur->pCost, &pWord->sCost); + pWord->rCost = rCost; + amatchWriteCost(pWord); +#ifdef AMATCH_TRACE_1 + printf(" ---> %d (\"%s\" \"%s\")\n", + pWord->rCost, pWord->zWord, pWord->zCost); +#endif + pOther = amatchAvlInsert(&pCur->pCost, &pWord->sCost); + assert( pOther==0 ); (void)pOther; + } + return; + } + pWord = sqlite3_malloc( sizeof(*pWord) + nBase + nTail - 1 ); + if( pWord==0 ) return; + memset(pWord, 0, sizeof(*pWord)); + pWord->rCost = rCost; + pWord->iSeq = pCur->nWord++; + amatchWriteCost(pWord); + pWord->nMatch = nMatch; + pWord->pNext = pCur->pAllWords; + pCur->pAllWords = pWord; + pWord->sCost.zKey = pWord->zCost; + pWord->sCost.pWord = pWord; + pOther = amatchAvlInsert(&pCur->pCost, &pWord->sCost); + assert( pOther==0 ); (void)pOther; + pWord->sWord.zKey = pWord->zWord; + pWord->sWord.pWord = pWord; + amatchStrcpy(pWord->zWord, pCur->zBuf); + pOther = amatchAvlInsert(&pCur->pWord, &pWord->sWord); + assert( pOther==0 ); (void)pOther; +#ifdef AMATCH_TRACE_1 + printf("INSERT [%s][%.*s^%s] %d (\"%s\" \"%s\")\n", pWord->zWord+2, + pWord->nMatch, pCur->zInput, pCur->zInput+pWord->nMatch, rCost, + pWord->zWord, pWord->zCost); +#endif +} + + +/* +** Advance a cursor to its next row of output +*/ +static int amatchNext(sqlite3_vtab_cursor *cur){ + amatch_cursor *pCur = (amatch_cursor*)cur; + amatch_word *pWord = 0; + amatch_avl *pNode; + int isMatch = 0; + amatch_vtab *p = pCur->pVtab; + int nWord; + int rc; + int i; + const char *zW; + amatch_rule *pRule; + char *zBuf = 0; + char nBuf = 0; + char zNext[8]; + char zNextIn[8]; + int nNextIn; + + if( p->pVCheck==0 ){ + char *zSql; + if( p->zVocabLang && p->zVocabLang[0] ){ + zSql = sqlite3_mprintf( + "SELECT \"%w\" FROM \"%w\"", + " WHERE \"%w\">=?1 AND \"%w\"=?2" + " ORDER BY 1", + p->zVocabWord, p->zVocabTab, + p->zVocabWord, p->zVocabLang + ); + }else{ + zSql = sqlite3_mprintf( + "SELECT \"%w\" FROM \"%w\"" + " WHERE \"%w\">=?1" + " ORDER BY 1", + p->zVocabWord, p->zVocabTab, + p->zVocabWord + ); + } + rc = sqlite3_prepare_v2(p->db, zSql, -1, &p->pVCheck, 0); + sqlite3_free(zSql); + if( rc ) return rc; + } + sqlite3_bind_int(p->pVCheck, 2, pCur->iLang); + + do{ + pNode = amatchAvlFirst(pCur->pCost); + if( pNode==0 ){ + pWord = 0; + break; + } + pWord = pNode->pWord; + amatchAvlRemove(&pCur->pCost, &pWord->sCost); + +#ifdef AMATCH_TRACE_1 + printf("PROCESS [%s][%.*s^%s] %d (\"%s\" \"%s\")\n", + pWord->zWord+2, pWord->nMatch, pCur->zInput, pCur->zInput+pWord->nMatch, + pWord->rCost, pWord->zWord, pWord->zCost); +#endif + nWord = (int)strlen(pWord->zWord+2); + if( nWord+20>nBuf ){ + nBuf = nWord+100; + zBuf = sqlite3_realloc(zBuf, nBuf); + if( zBuf==0 ) return SQLITE_NOMEM; + } + amatchStrcpy(zBuf, pWord->zWord+2); + zNext[0] = 0; + zNextIn[0] = pCur->zInput[pWord->nMatch]; + if( zNextIn[0] ){ + for(i=1; i<=4 && (pCur->zInput[pWord->nMatch+i]&0xc0)==0x80; i++){ + zNextIn[i] = pCur->zInput[pWord->nMatch+i]; + } + zNextIn[i] = 0; + nNextIn = i; + }else{ + nNextIn = 0; + } + + if( zNextIn[0] && zNextIn[0]!='*' ){ + sqlite3_reset(p->pVCheck); + amatchStrcat(zBuf, zNextIn); + sqlite3_bind_text(p->pVCheck, 1, zBuf, nWord+nNextIn, SQLITE_STATIC); + rc = sqlite3_step(p->pVCheck); + if( rc==SQLITE_ROW ){ + zW = (const char*)sqlite3_column_text(p->pVCheck, 0); + if( strncmp(zBuf, zW, nWord+nNextIn)==0 ){ + amatchAddWord(pCur, pWord->rCost, pWord->nMatch+nNextIn, zBuf, ""); + } + } + zBuf[nWord] = 0; + } + + while( 1 ){ + amatchStrcpy(zBuf+nWord, zNext); + sqlite3_reset(p->pVCheck); + sqlite3_bind_text(p->pVCheck, 1, zBuf, -1, SQLITE_TRANSIENT); + rc = sqlite3_step(p->pVCheck); + if( rc!=SQLITE_ROW ) break; + zW = (const char*)sqlite3_column_text(p->pVCheck, 0); + amatchStrcpy(zBuf+nWord, zNext); + if( strncmp(zW, zBuf, nWord)!=0 ) break; + if( (zNextIn[0]=='*' && zNextIn[1]==0) + || (zNextIn[0]==0 && zW[nWord]==0) + ){ + isMatch = 1; + zNextIn[0] = 0; + nNextIn = 0; + break; + } + zNext[0] = zW[nWord]; + for(i=1; i<=4 && (zW[nWord+i]&0xc0)==0x80; i++){ + zNext[i] = zW[nWord+i]; + } + zNext[i] = 0; + zBuf[nWord] = 0; + if( p->rIns>0 ){ + amatchAddWord(pCur, pWord->rCost+p->rIns, pWord->nMatch, + zBuf, zNext); + } + if( p->rSub>0 ){ + amatchAddWord(pCur, pWord->rCost+p->rSub, pWord->nMatch+nNextIn, + zBuf, zNext); + } + if( p->rIns<0 && p->rSub<0 ) break; + zNext[i-1]++; /* FIX ME */ + } + sqlite3_reset(p->pVCheck); + + if( p->rDel>0 ){ + zBuf[nWord] = 0; + amatchAddWord(pCur, pWord->rCost+p->rDel, pWord->nMatch+nNextIn, + zBuf, ""); + } + + for(pRule=p->pRule; pRule; pRule=pRule->pNext){ + if( pRule->iLang!=pCur->iLang ) continue; + if( strncmp(pRule->zFrom, pCur->zInput+pWord->nMatch, pRule->nFrom)==0 ){ + amatchAddWord(pCur, pWord->rCost+pRule->rCost, + pWord->nMatch+pRule->nFrom, pWord->zWord+2, pRule->zTo); + } + } + }while( !isMatch ); + pCur->pCurrent = pWord; + sqlite3_free(zBuf); + return SQLITE_OK; +} + +/* +** 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 amatchColumn, amatchRowid, or amatchEof call. +*/ +static int amatchFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + amatch_cursor *pCur = (amatch_cursor *)pVtabCursor; + const char *zWord = "*"; + int idx; + + amatchClearCursor(pCur); + idx = 0; + if( idxNum & 1 ){ + zWord = (const char*)sqlite3_value_text(argv[0]); + idx++; + } + if( idxNum & 2 ){ + pCur->rLimit = (amatch_cost)sqlite3_value_int(argv[idx]); + idx++; + } + if( idxNum & 4 ){ + pCur->iLang = (amatch_cost)sqlite3_value_int(argv[idx]); + idx++; + } + pCur->zInput = sqlite3_mprintf("%s", zWord); + if( pCur->zInput==0 ) return SQLITE_NOMEM; + amatchAddWord(pCur, 0, 0, "", ""); + amatchNext(pVtabCursor); + + return SQLITE_OK; +} + +/* +** Only the word and distance columns have values. All other columns +** return NULL +*/ +static int amatchColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ + amatch_cursor *pCur = (amatch_cursor*)cur; + switch( i ){ + case AMATCH_COL_WORD: { + sqlite3_result_text(ctx, pCur->pCurrent->zWord+2, -1, SQLITE_STATIC); + break; + } + case AMATCH_COL_DISTANCE: { + sqlite3_result_int(ctx, pCur->pCurrent->rCost); + break; + } + case AMATCH_COL_LANGUAGE: { + sqlite3_result_int(ctx, pCur->iLang); + break; + } + case AMATCH_COL_NWORD: { + sqlite3_result_int(ctx, pCur->nWord); + break; + } + default: { + sqlite3_result_null(ctx); + break; + } + } + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int amatchRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + amatch_cursor *pCur = (amatch_cursor*)cur; + *pRowid = pCur->iRowid; + return SQLITE_OK; +} + +/* +** EOF indicator +*/ +static int amatchEof(sqlite3_vtab_cursor *cur){ + amatch_cursor *pCur = (amatch_cursor*)cur; + return pCur->pCurrent==0; +} + +/* +** Search for terms of these forms: +** +** (A) word MATCH $str +** (B1) distance < $value +** (B2) distance <= $value +** (C) language == $language +** +** The distance< and distance<= are both treated as distance<=. +** The query plan number is a bit vector: +** +** bit 1: Term of the form (A) found +** bit 2: Term like (B1) or (B2) found +** bit 3: Term like (C) found +** +** If bit-1 is set, $str is always in filter.argv[0]. If bit-2 is set +** then $value is in filter.argv[0] if bit-1 is clear and is in +** filter.argv[1] if bit-1 is set. If bit-3 is set, then $ruleid is +** in filter.argv[0] if bit-1 and bit-2 are both zero, is in +** filter.argv[1] if exactly one of bit-1 and bit-2 are set, and is in +** filter.argv[2] if both bit-1 and bit-2 are set. +*/ +static int amatchBestIndex( + sqlite3_vtab *tab, + sqlite3_index_info *pIdxInfo +){ + int iPlan = 0; + int iDistTerm = -1; + int iLangTerm = -1; + int i; + const struct sqlite3_index_constraint *pConstraint; + + (void)tab; + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + if( (iPlan & 1)==0 + && pConstraint->iColumn==0 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 2)==0 + && pConstraint->iColumn==1 + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) + ){ + iPlan |= 2; + iDistTerm = i; + } + if( (iPlan & 4)==0 + && pConstraint->iColumn==2 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 4; + pIdxInfo->aConstraintUsage[i].omit = 1; + iLangTerm = i; + } + } + if( iPlan & 2 ){ + pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = 1+((iPlan&1)!=0); + } + if( iPlan & 4 ){ + int idx = 1; + if( iPlan & 1 ) idx++; + if( iPlan & 2 ) idx++; + pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx; + } + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==1 + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + pIdxInfo->estimatedCost = (double)10000; + + return SQLITE_OK; +} + +/* +** The xUpdate() method. +** +** This implementation disallows DELETE and UPDATE. The only thing +** allowed is INSERT into the "command" column. +*/ +static int amatchUpdate( + sqlite3_vtab *pVTab, + int argc, + sqlite3_value **argv, + sqlite_int64 *pRowid +){ + amatch_vtab *p = (amatch_vtab*)pVTab; + const unsigned char *zCmd; + (void)pRowid; + if( argc==1 ){ + pVTab->zErrMsg = sqlite3_mprintf("DELETE from %s is not allowed", + p->zSelf); + return SQLITE_ERROR; + } + if( sqlite3_value_type(argv[0])!=SQLITE_NULL ){ + pVTab->zErrMsg = sqlite3_mprintf("UPDATE of %s is not allowed", + p->zSelf); + return SQLITE_ERROR; + } + if( sqlite3_value_type(argv[2+AMATCH_COL_WORD])!=SQLITE_NULL + || sqlite3_value_type(argv[2+AMATCH_COL_DISTANCE])!=SQLITE_NULL + || sqlite3_value_type(argv[2+AMATCH_COL_LANGUAGE])!=SQLITE_NULL + ){ + pVTab->zErrMsg = sqlite3_mprintf( + "INSERT INTO %s allowed for column [command] only", p->zSelf); + return SQLITE_ERROR; + } + zCmd = sqlite3_value_text(argv[2+AMATCH_COL_COMMAND]); + if( zCmd==0 ) return SQLITE_OK; + + return SQLITE_OK; +} + +/* +** A virtual table module that implements the "approximate_match". +*/ +static const sqlite3_module amatchModule = { + 0, /* iVersion */ + amatchConnect, /* xCreate */ + amatchConnect, /* xConnect */ + amatchBestIndex, /* xBestIndex */ + amatchDisconnect, /* xDisconnect */ + amatchDisconnect, /* xDestroy */ + amatchOpen, /* xOpen - open a cursor */ + amatchClose, /* xClose - close a cursor */ + amatchFilter, /* xFilter - configure scan constraints */ + amatchNext, /* xNext - advance a cursor */ + amatchEof, /* xEof - check for end of scan */ + amatchColumn, /* xColumn - read data */ + amatchRowid, /* xRowid - read data */ + amatchUpdate, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ + 0, /* xSavepoint */ + 0, /* xRelease */ + 0 /* xRollbackTo */ +}; + +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + +/* +** Register the amatch virtual table +*/ +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_amatch_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Not used */ +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "approximate_match", &amatchModule, 0); +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Not used */ +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "approximate_match", &amatchModule, 0); +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/closure.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/closure.c 2016-01-06 16:30:42.472119600 +0100 @@ -0,0 +1,973 @@ +/* +** 2013-04-16 +** +** 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 file contains code for a virtual table that finds the transitive +** closure of a parent/child relationship in a real table. The virtual +** table is called "transitive_closure". +** +** A transitive_closure virtual table is created like this: +** +** CREATE VIRTUAL TABLE x USING transitive_closure( +** tablename=, -- T +** idcolumn=, -- X +** parentcolumn= -- P +** ); +** +** When it is created, the new transitive_closure table may be supplied +** with default values for the name of a table T and columns T.X and T.P. +** The T.X and T.P columns must contain integers. The ideal case is for +** T.X to be the INTEGER PRIMARY KEY. The T.P column should reference +** the T.X column. The row referenced by T.P is the parent of the current row. +** +** The tablename, idcolumn, and parentcolumn supplied by the CREATE VIRTUAL +** TABLE statement may be overridden in individual queries by including +** terms like tablename='newtable', idcolumn='id2', or +** parentcolumn='parent3' in the WHERE clause of the query. +** +** For efficiency, it is essential that there be an index on the P column: +** +** CREATE Tidx1 ON T(P) +** +** Suppose a specific instance of the closure table is as follows: +** +** CREATE VIRTUAL TABLE ct1 USING transitive_closure( +** tablename='group', +** idcolumn='groupId', +** parentcolumn='parentId' +** ); +** +** Such an instance of the transitive_closure virtual table would be +** appropriate for walking a tree defined using a table like this, for example: +** +** CREATE TABLE group( +** groupId INTEGER PRIMARY KEY, +** parentId INTEGER REFERENCES group +** ); +** CREATE INDEX group_idx1 ON group(parentId); +** +** The group table above would presumably have other application-specific +** fields. The key point here is that rows of the group table form a +** tree. The purpose of the ct1 virtual table is to easily extract +** branches of that tree. +** +** Once it has been created, the ct1 virtual table can be queried +** as follows: +** +** SELECT * FROM element +** WHERE element.groupId IN (SELECT id FROM ct1 WHERE root=?1); +** +** The above query will return all elements that are part of group ?1 +** or children of group ?1 or grand-children of ?1 and so forth for all +** descendents of group ?1. The same query can be formulated as a join: +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupid=ct1.id +** AND ct1.root=?1; +** +** The depth of the transitive_closure (the number of generations of +** parent/child relations to follow) can be limited by setting "depth" +** column in the WHERE clause. So, for example, the following query +** finds only children and grandchildren but no further descendents: +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupid=ct1.id +** AND ct1.root=?1 +** AND ct1.depth<=2; +** +** The "ct1.depth<=2" term could be a strict equality "ct1.depth=2" in +** order to find only the grandchildren of ?1, not ?1 itself or the +** children of ?1. +** +** The root=?1 term must be supplied in WHERE clause or else the query +** of the ct1 virtual table will return an empty set. The tablename, +** idcolumn, and parentcolumn attributes can be overridden in the WHERE +** clause if desired. So, for example, the ct1 table could be repurposed +** to find ancestors rather than descendents by inverting the roles of +** the idcolumn and parentcolumn: +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupid=ct1.id +** AND ct1.root=?1 +** AND ct1.idcolumn='parentId' +** AND ct1.parentcolumn='groupId'; +** +** Multiple calls to ct1 could be combined. For example, the following +** query finds all elements that "cousins" of groupId ?1. That is to say +** elements where the groupId is a grandchild of the grandparent of ?1. +** (This definition of "cousins" also includes siblings and self.) +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupId=ct1.id +** AND ct1.depth=2 +** AND ct1.root IN (SELECT id FROM ct1 +** WHERE root=?1 +** AND depth=2 +** AND idcolumn='parentId' +** AND parentcolumn='groupId'); +** +** In our example, the group.groupId column is unique and thus the +** subquery will return exactly one row. For that reason, the IN +** operator could be replaced by "=" to get the same result. But +** in the general case where the idcolumn is not unique, an IN operator +** would be required for this kind of query. +** +** Note that because the tablename, idcolumn, and parentcolumn can +** all be specified in the query, it is possible for an application +** to define a single transitive_closure virtual table for use on lots +** of different hierarchy tables. One might say: +** +** CREATE VIRTUAL TABLE temp.closure USING transitive_closure; +** +** As each database connection is being opened. Then the application +** would always have a "closure" virtual table handy to use for querying. +** +** SELECT element.* FROM element, closure +** WHERE element.groupid=ct1.id +** AND closure.root=?1 +** AND closure.tablename='group' +** AND closure.idname='groupId' +** AND closure.parentname='parentId'; +** +** See the documentation at http://www.sqlite.org/loadext.html for information +** on how to compile and use loadable extensions such as this one. +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include +#include +#include +#include +#include + +#ifndef SQLITE_OMIT_VIRTUALTABLE + +/* +** Forward declaration of objects used by this implementation +*/ +typedef struct closure_vtab closure_vtab; +typedef struct closure_cursor closure_cursor; +typedef struct closure_queue closure_queue; +typedef struct closure_avl closure_avl; + +/***************************************************************************** +** AVL Tree implementation +*/ +/* +** Objects that want to be members of the AVL tree should embedded an +** instance of this structure. +*/ +struct closure_avl { + sqlite3_int64 id; /* Id of this entry in the table */ + int iGeneration; /* Which generation is this entry part of */ + closure_avl *pList; /* A linked list of nodes */ + closure_avl *pBefore; /* Other elements less than id */ + closure_avl *pAfter; /* Other elements greater than id */ + closure_avl *pUp; /* Parent element */ + short int height; /* Height of this node. Leaf==1 */ + short int imbalance; /* Height difference between pBefore and pAfter */ +}; + +/* Recompute the closure_avl.height and closure_avl.imbalance fields for p. +** Assume that the children of p have correct heights. +*/ +static void closureAvlRecomputeHeight(closure_avl *p){ + short int hBefore = p->pBefore ? p->pBefore->height : 0; + short int hAfter = p->pAfter ? p->pAfter->height : 0; + p->imbalance = hBefore - hAfter; /* -: pAfter higher. +: pBefore higher */ + p->height = (hBefore>hAfter ? hBefore : hAfter)+1; +} + +/* +** P B +** / \ / \ +** B Z ==> X P +** / \ / \ +** X Y Y Z +** +*/ +static closure_avl *closureAvlRotateBefore(closure_avl *pP){ + closure_avl *pB = pP->pBefore; + closure_avl *pY = pB->pAfter; + pB->pUp = pP->pUp; + pB->pAfter = pP; + pP->pUp = pB; + pP->pBefore = pY; + if( pY ) pY->pUp = pP; + closureAvlRecomputeHeight(pP); + closureAvlRecomputeHeight(pB); + return pB; +} + +/* +** P A +** / \ / \ +** X A ==> P Z +** / \ / \ +** Y Z X Y +** +*/ +static closure_avl *closureAvlRotateAfter(closure_avl *pP){ + closure_avl *pA = pP->pAfter; + closure_avl *pY = pA->pBefore; + pA->pUp = pP->pUp; + pA->pBefore = pP; + pP->pUp = pA; + pP->pAfter = pY; + if( pY ) pY->pUp = pP; + closureAvlRecomputeHeight(pP); + closureAvlRecomputeHeight(pA); + return pA; +} + +/* +** Return a pointer to the pBefore or pAfter pointer in the parent +** of p that points to p. Or if p is the root node, return pp. +*/ +static closure_avl **closureAvlFromPtr(closure_avl *p, closure_avl **pp){ + closure_avl *pUp = p->pUp; + if( pUp==0 ) return pp; + if( pUp->pAfter==p ) return &pUp->pAfter; + return &pUp->pBefore; +} + +/* +** Rebalance all nodes starting with p and working up to the root. +** Return the new root. +*/ +static closure_avl *closureAvlBalance(closure_avl *p){ + closure_avl *pTop = p; + closure_avl **pp; + while( p ){ + closureAvlRecomputeHeight(p); + if( p->imbalance>=2 ){ + closure_avl *pB = p->pBefore; + if( pB->imbalance<0 ) p->pBefore = closureAvlRotateAfter(pB); + pp = closureAvlFromPtr(p,&p); + p = *pp = closureAvlRotateBefore(p); + }else if( p->imbalance<=(-2) ){ + closure_avl *pA = p->pAfter; + if( pA->imbalance>0 ) p->pAfter = closureAvlRotateBefore(pA); + pp = closureAvlFromPtr(p,&p); + p = *pp = closureAvlRotateAfter(p); + } + pTop = p; + p = p->pUp; + } + return pTop; +} + +/* Search the tree rooted at p for an entry with id. Return a pointer +** to the entry or return NULL. +*/ +static closure_avl *closureAvlSearch(closure_avl *p, sqlite3_int64 id){ + while( p && id!=p->id ){ + p = (idid) ? p->pBefore : p->pAfter; + } + return p; +} + +/* Find the first node (the one with the smallest key). +*/ +static closure_avl *closureAvlFirst(closure_avl *p){ + if( p ) while( p->pBefore ) p = p->pBefore; + return p; +} + +/* Return the node with the next larger key after p. +*/ +closure_avl *closureAvlNext(closure_avl *p){ + closure_avl *pPrev = 0; + while( p && p->pAfter==pPrev ){ + pPrev = p; + p = p->pUp; + } + if( p && pPrev==0 ){ + p = closureAvlFirst(p->pAfter); + } + return p; +} + +/* Insert a new node pNew. Return NULL on success. If the key is not +** unique, then do not perform the insert but instead leave pNew unchanged +** and return a pointer to an existing node with the same key. +*/ +static closure_avl *closureAvlInsert( + closure_avl **ppHead, /* Head of the tree */ + closure_avl *pNew /* New node to be inserted */ +){ + closure_avl *p = *ppHead; + if( p==0 ){ + p = pNew; + pNew->pUp = 0; + }else{ + while( p ){ + if( pNew->idid ){ + if( p->pBefore ){ + p = p->pBefore; + }else{ + p->pBefore = pNew; + pNew->pUp = p; + break; + } + }else if( pNew->id>p->id ){ + if( p->pAfter ){ + p = p->pAfter; + }else{ + p->pAfter = pNew; + pNew->pUp = p; + break; + } + }else{ + return p; + } + } + } + pNew->pBefore = 0; + pNew->pAfter = 0; + pNew->height = 1; + pNew->imbalance = 0; + *ppHead = closureAvlBalance(p); + return 0; +} + +/* Walk the tree can call xDestroy on each node +*/ +static void closureAvlDestroy(closure_avl *p, void (*xDestroy)(closure_avl*)){ + if( p ){ + closureAvlDestroy(p->pBefore, xDestroy); + closureAvlDestroy(p->pAfter, xDestroy); + xDestroy(p); + } +} +/* +** End of the AVL Tree implementation +******************************************************************************/ + +/* +** A closure virtual-table object +*/ +struct closure_vtab { + sqlite3_vtab base; /* Base class - must be first */ + char *zDb; /* Name of database. (ex: "main") */ + char *zSelf; /* Name of this virtual table */ + char *zTableName; /* Name of table holding parent/child relation */ + char *zIdColumn; /* Name of ID column of zTableName */ + char *zParentColumn; /* Name of PARENT column in zTableName */ + sqlite3 *db; /* The database connection */ + int nCursor; /* Number of pending cursors */ +}; + +/* A closure cursor object */ +struct closure_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + closure_vtab *pVtab; /* The virtual table this cursor belongs to */ + char *zTableName; /* Name of table holding parent/child relation */ + char *zIdColumn; /* Name of ID column of zTableName */ + char *zParentColumn; /* Name of PARENT column in zTableName */ + closure_avl *pCurrent; /* Current element of output */ + closure_avl *pClosure; /* The complete closure tree */ +}; + +/* A queue of AVL nodes */ +struct closure_queue { + closure_avl *pFirst; /* Oldest node on the queue */ + closure_avl *pLast; /* Youngest node on the queue */ +}; + +/* +** Add a node to the end of the queue +*/ +static void queuePush(closure_queue *pQueue, closure_avl *pNode){ + pNode->pList = 0; + if( pQueue->pLast ){ + pQueue->pLast->pList = pNode; + }else{ + pQueue->pFirst = pNode; + } + pQueue->pLast = pNode; +} + +/* +** Extract the oldest element (the front element) from the queue. +*/ +static closure_avl *queuePull(closure_queue *pQueue){ + closure_avl *p = pQueue->pFirst; + if( p ){ + pQueue->pFirst = p->pList; + if( pQueue->pFirst==0 ) pQueue->pLast = 0; + } + return p; +} + +/* +** This function converts an SQL quoted string into an unquoted string +** and returns a pointer to a buffer allocated using sqlite3_malloc() +** containing the result. The caller should eventually free this buffer +** using sqlite3_free. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static char *closureDequote(const char *zIn){ + int nIn; /* Size of input string, in bytes */ + char *zOut; /* Output (dequoted) string */ + + nIn = (int)strlen(zIn); + zOut = sqlite3_malloc(nIn+1); + if( zOut ){ + char q = zIn[0]; /* Quote character (if any ) */ + + if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ + memcpy(zOut, zIn, nIn+1); + }else{ + int iOut = 0; /* Index of next byte to write to output */ + int iIn; /* Index of next byte to read from input */ + + if( q=='[' ) q = ']'; + for(iIn=1; iInzDb); + sqlite3_free(p->zSelf); + sqlite3_free(p->zTableName); + sqlite3_free(p->zIdColumn); + sqlite3_free(p->zParentColumn); + memset(p, 0, sizeof(*p)); + sqlite3_free(p); + } +} + +/* +** xDisconnect/xDestroy method for the closure module. +*/ +static int closureDisconnect(sqlite3_vtab *pVtab){ + closure_vtab *p = (closure_vtab*)pVtab; + assert( p->nCursor==0 ); + closureFree(p); + return SQLITE_OK; +} + +/* +** Check to see if the argument is of the form: +** +** KEY = VALUE +** +** If it is, return a pointer to the first character of VALUE. +** If not, return NULL. Spaces around the = are ignored. +*/ +static const char *closureValueOfKey(const char *zKey, const char *zStr){ + int nKey = (int)strlen(zKey); + int nStr = (int)strlen(zStr); + int i; + if( nStr module name ("transitive_closure") +** argv[1] -> database name +** argv[2] -> table name +** argv[3...] -> arguments +*/ +static int closureConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + int rc = SQLITE_OK; /* Return code */ + closure_vtab *pNew = 0; /* New virtual table */ + const char *zDb = argv[1]; + const char *zVal; + int i; + + (void)pAux; + *ppVtab = 0; + pNew = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + rc = SQLITE_NOMEM; + memset(pNew, 0, sizeof(*pNew)); + pNew->db = db; + pNew->zDb = sqlite3_mprintf("%s", zDb); + if( pNew->zDb==0 ) goto closureConnectError; + pNew->zSelf = sqlite3_mprintf("%s", argv[2]); + if( pNew->zSelf==0 ) goto closureConnectError; + for(i=3; izTableName); + pNew->zTableName = closureDequote(zVal); + if( pNew->zTableName==0 ) goto closureConnectError; + continue; + } + zVal = closureValueOfKey("idcolumn", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zIdColumn); + pNew->zIdColumn = closureDequote(zVal); + if( pNew->zIdColumn==0 ) goto closureConnectError; + continue; + } + zVal = closureValueOfKey("parentcolumn", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zParentColumn); + pNew->zParentColumn = closureDequote(zVal); + if( pNew->zParentColumn==0 ) goto closureConnectError; + continue; + } + *pzErr = sqlite3_mprintf("unrecognized argument: [%s]\n", argv[i]); + closureFree(pNew); + *ppVtab = 0; + return SQLITE_ERROR; + } + rc = sqlite3_declare_vtab(db, + "CREATE TABLE x(id,depth,root HIDDEN,tablename HIDDEN," + "idcolumn HIDDEN,parentcolumn HIDDEN)" + ); +#define CLOSURE_COL_ID 0 +#define CLOSURE_COL_DEPTH 1 +#define CLOSURE_COL_ROOT 2 +#define CLOSURE_COL_TABLENAME 3 +#define CLOSURE_COL_IDCOLUMN 4 +#define CLOSURE_COL_PARENTCOLUMN 5 + if( rc!=SQLITE_OK ){ + closureFree(pNew); + } + *ppVtab = &pNew->base; + return rc; + +closureConnectError: + closureFree(pNew); + return rc; +} + +/* +** Open a new closure cursor. +*/ +static int closureOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + closure_vtab *p = (closure_vtab*)pVTab; + closure_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVtab = p; + *ppCursor = &pCur->base; + p->nCursor++; + return SQLITE_OK; +} + +/* +** Free up all the memory allocated by a cursor. Set it rLimit to 0 +** to indicate that it is at EOF. +*/ +static void closureClearCursor(closure_cursor *pCur){ + closureAvlDestroy(pCur->pClosure, (void(*)(closure_avl*))sqlite3_free); + sqlite3_free(pCur->zTableName); + sqlite3_free(pCur->zIdColumn); + sqlite3_free(pCur->zParentColumn); + pCur->zTableName = 0; + pCur->zIdColumn = 0; + pCur->zParentColumn = 0; + pCur->pCurrent = 0; + pCur->pClosure = 0; +} + +/* +** Close a closure cursor. +*/ +static int closureClose(sqlite3_vtab_cursor *cur){ + closure_cursor *pCur = (closure_cursor *)cur; + closureClearCursor(pCur); + pCur->pVtab->nCursor--; + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Advance a cursor to its next row of output +*/ +static int closureNext(sqlite3_vtab_cursor *cur){ + closure_cursor *pCur = (closure_cursor*)cur; + pCur->pCurrent = closureAvlNext(pCur->pCurrent); + return SQLITE_OK; +} + +/* +** Allocate and insert a node +*/ +static int closureInsertNode( + closure_queue *pQueue, /* Add new node to this queue */ + closure_cursor *pCur, /* The cursor into which to add the node */ + sqlite3_int64 id, /* The node ID */ + int iGeneration /* The generation number for this node */ +){ + closure_avl *pNew = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + memset(pNew, 0, sizeof(*pNew)); + pNew->id = id; + pNew->iGeneration = iGeneration; + closureAvlInsert(&pCur->pClosure, pNew); + queuePush(pQueue, pNew); + return SQLITE_OK; +} + +/* +** 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 closureColumn, closureRowid, or closureEof call. +** +** This routine actually computes the closure. +** +** See the comment at the beginning of closureBestIndex() for a +** description of the meaning of idxNum. The idxStr parameter is +** not used. +*/ +static int closureFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + closure_cursor *pCur = (closure_cursor *)pVtabCursor; + closure_vtab *pVtab = pCur->pVtab; + sqlite3_int64 iRoot; + int mxGen = 999999999; + char *zSql; + sqlite3_stmt *pStmt; + closure_avl *pAvl; + int rc = SQLITE_OK; + const char *zTableName = pVtab->zTableName; + const char *zIdColumn = pVtab->zIdColumn; + const char *zParentColumn = pVtab->zParentColumn; + closure_queue sQueue; + + (void)idxStr; /* Unused parameter */ + (void)argc; /* Unused parameter */ + closureClearCursor(pCur); + memset(&sQueue, 0, sizeof(sQueue)); + if( (idxNum & 1)==0 ){ + /* No root=$root in the WHERE clause. Return an empty set */ + return SQLITE_OK; + } + iRoot = sqlite3_value_int64(argv[0]); + if( (idxNum & 0x000f0)!=0 ){ + mxGen = sqlite3_value_int(argv[(idxNum>>4)&0x0f]); + if( (idxNum & 0x00002)!=0 ) mxGen--; + } + if( (idxNum & 0x00f00)!=0 ){ + zTableName = (const char*)sqlite3_value_text(argv[(idxNum>>8)&0x0f]); + pCur->zTableName = sqlite3_mprintf("%s", zTableName); + } + if( (idxNum & 0x0f000)!=0 ){ + zIdColumn = (const char*)sqlite3_value_text(argv[(idxNum>>12)&0x0f]); + pCur->zIdColumn = sqlite3_mprintf("%s", zIdColumn); + } + if( (idxNum & 0x0f0000)!=0 ){ + zParentColumn = (const char*)sqlite3_value_text(argv[(idxNum>>16)&0x0f]); + pCur->zParentColumn = sqlite3_mprintf("%s", zParentColumn); + } + + zSql = sqlite3_mprintf( + "SELECT \"%w\".\"%w\" FROM \"%w\" WHERE \"%w\".\"%w\"=?1", + zTableName, zIdColumn, zTableName, zTableName, zParentColumn); + if( zSql==0 ){ + return SQLITE_NOMEM; + }else{ + rc = sqlite3_prepare_v2(pVtab->db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + if( rc ){ + sqlite3_free(pVtab->base.zErrMsg); + pVtab->base.zErrMsg = sqlite3_mprintf("%s", sqlite3_errmsg(pVtab->db)); + return rc; + } + } + if( rc==SQLITE_OK ){ + rc = closureInsertNode(&sQueue, pCur, iRoot, 0); + } + while( (pAvl = queuePull(&sQueue))!=0 ){ + if( pAvl->iGeneration>=mxGen ) continue; + sqlite3_bind_int64(pStmt, 1, pAvl->id); + while( rc==SQLITE_OK && sqlite3_step(pStmt)==SQLITE_ROW ){ + if( sqlite3_column_type(pStmt,0)==SQLITE_INTEGER ){ + sqlite3_int64 iNew = sqlite3_column_int64(pStmt, 0); + if( closureAvlSearch(pCur->pClosure, iNew)==0 ){ + rc = closureInsertNode(&sQueue, pCur, iNew, pAvl->iGeneration+1); + } + } + } + sqlite3_reset(pStmt); + } + sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ){ + pCur->pCurrent = closureAvlFirst(pCur->pClosure); + } + + return rc; +} + +/* +** Only the word and distance columns have values. All other columns +** return NULL +*/ +static int closureColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ + closure_cursor *pCur = (closure_cursor*)cur; + switch( i ){ + case CLOSURE_COL_ID: { + sqlite3_result_int64(ctx, pCur->pCurrent->id); + break; + } + case CLOSURE_COL_DEPTH: { + sqlite3_result_int(ctx, pCur->pCurrent->iGeneration); + break; + } + case CLOSURE_COL_ROOT: { + sqlite3_result_null(ctx); + break; + } + case CLOSURE_COL_TABLENAME: { + sqlite3_result_text(ctx, + pCur->zTableName ? pCur->zTableName : pCur->pVtab->zTableName, + -1, SQLITE_TRANSIENT); + break; + } + case CLOSURE_COL_IDCOLUMN: { + sqlite3_result_text(ctx, + pCur->zIdColumn ? pCur->zIdColumn : pCur->pVtab->zIdColumn, + -1, SQLITE_TRANSIENT); + break; + } + case CLOSURE_COL_PARENTCOLUMN: { + sqlite3_result_text(ctx, + pCur->zParentColumn ? pCur->zParentColumn : pCur->pVtab->zParentColumn, + -1, SQLITE_TRANSIENT); + break; + } + } + return SQLITE_OK; +} + +/* +** The rowid. For the closure table, this is the same as the "id" column. +*/ +static int closureRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + closure_cursor *pCur = (closure_cursor*)cur; + *pRowid = pCur->pCurrent->id; + return SQLITE_OK; +} + +/* +** EOF indicator +*/ +static int closureEof(sqlite3_vtab_cursor *cur){ + closure_cursor *pCur = (closure_cursor*)cur; + return pCur->pCurrent==0; +} + +/* +** Search for terms of these forms: +** +** (A) root = $root +** (B1) depth < $depth +** (B2) depth <= $depth +** (B3) depth = $depth +** (C) tablename = $tablename +** (D) idcolumn = $idcolumn +** (E) parentcolumn = $parentcolumn +** +** +** +** idxNum meaning +** ---------- ------------------------------------------------------ +** 0x00000001 Term of the form (A) found +** 0x00000002 The term of bit-2 is like (B1) +** 0x000000f0 Index in filter.argv[] of $depth. 0 if not used. +** 0x00000f00 Index in filter.argv[] of $tablename. 0 if not used. +** 0x0000f000 Index in filter.argv[] of $idcolumn. 0 if not used +** 0x000f0000 Index in filter.argv[] of $parentcolumn. 0 if not used. +** +** There must be a term of type (A). If there is not, then the index type +** is 0 and the query will return an empty set. +*/ +static int closureBestIndex( + sqlite3_vtab *pTab, /* The virtual table */ + sqlite3_index_info *pIdxInfo /* Information about the query */ +){ + int iPlan = 0; + int i; + int idx = 1; + int seenMatch = 0; + const struct sqlite3_index_constraint *pConstraint; + closure_vtab *pVtab = (closure_vtab*)pTab; + double rCost = 10000000.0; + + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->iColumn==CLOSURE_COL_ROOT + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ ){ + seenMatch = 1; + } + if( pConstraint->usable==0 ) continue; + if( (iPlan & 1)==0 + && pConstraint->iColumn==CLOSURE_COL_ROOT + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + rCost /= 100.0; + } + if( (iPlan & 0x0000f0)==0 + && pConstraint->iColumn==CLOSURE_COL_DEPTH + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ) + ){ + iPlan |= idx<<4; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + if( pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ) iPlan |= 0x000002; + rCost /= 5.0; + } + if( (iPlan & 0x000f00)==0 + && pConstraint->iColumn==CLOSURE_COL_TABLENAME + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= idx<<8; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + pIdxInfo->aConstraintUsage[i].omit = 1; + rCost /= 5.0; + } + if( (iPlan & 0x00f000)==0 + && pConstraint->iColumn==CLOSURE_COL_IDCOLUMN + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= idx<<12; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 0x0f0000)==0 + && pConstraint->iColumn==CLOSURE_COL_PARENTCOLUMN + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= idx<<16; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + } + if( (pVtab->zTableName==0 && (iPlan & 0x000f00)==0) + || (pVtab->zIdColumn==0 && (iPlan & 0x00f000)==0) + || (pVtab->zParentColumn==0 && (iPlan & 0x0f0000)==0) + ){ + /* All of tablename, idcolumn, and parentcolumn must be specified + ** in either the CREATE VIRTUAL TABLE or in the WHERE clause constraints + ** or else the result is an empty set. */ + iPlan = 0; + } + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==CLOSURE_COL_ID + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + if( seenMatch && (iPlan&1)==0 ) rCost *= 1e30; + pIdxInfo->estimatedCost = rCost; + + return SQLITE_OK; +} + +/* +** A virtual table module that implements the "transitive_closure". +*/ +static const sqlite3_module closureModule = { + 0, /* iVersion */ + closureConnect, /* xCreate */ + closureConnect, /* xConnect */ + closureBestIndex, /* xBestIndex */ + closureDisconnect, /* xDisconnect */ + closureDisconnect, /* xDestroy */ + closureOpen, /* xOpen - open a cursor */ + closureClose, /* xClose - close a cursor */ + closureFilter, /* xFilter - configure scan constraints */ + closureNext, /* xNext - advance a cursor */ + closureEof, /* xEof - check for end of scan */ + closureColumn, /* xColumn - read data */ + closureRowid, /* xRowid - read data */ + 0, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ + 0, /* xSavepoint */ + 0, /* xRelease */ + 0 /* xRollbackTo */ +}; + +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + +/* +** Register the closure virtual table +*/ +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_closure_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "transitive_closure", &closureModule, 0); +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "transitive_closure", &closureModule, 0); +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/compress.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/compress.c 2016-01-06 16:30:42.475119800 +0100 @@ -0,0 +1,178 @@ +/* +** 2014-06-13 +** +** 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 SQLite extension implements SQL compression functions +** compress() and uncompress() using ZLIB. +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include + +/* +** Implementation of the "compress(X)" SQL function. The input X is +** compressed using zLib and the output is returned. +** +** The output is a BLOB that begins with an integer that forming the +** input size in bytes (the size of X before compression). The variable- +** length integer is implemented as 1 to 5 bytes. If the first byte is 0, +** 4 bytes are used, the most significant bits first. Otherwise there are +** seven bits per integer stored in the lower seven bits of each byte. +** More significant bits occur first. The most significant bit (0x80) +** is a flag to indicate the end of the integer. +*/ +static void compressFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *pIn; + unsigned char *pOut; + size_t nIn; + unsigned long int nOut; + int j, rc; + + pIn = sqlite3_value_blob(argv[0]); + nIn = sqlite3_value_bytes(argv[0]); + nOut = 13 + nIn + (nIn+999)/1000; + pOut = sqlite3_malloc( nOut+5 ); + if( (nIn<=0xffffff) || (!(nIn&0x80)&&!(nIn&0x8000)&&!(nIn&0x800000))) { + pOut[0] = nIn>>24 & 0xff; + pOut[1] = nIn>>16 & 0xff; + pOut[2] = nIn>>8 & 0xff; + pOut[3] = nIn & 0xff; + j = 4; + }else{ + int i; + unsigned char x[8]; + for(i=4; i>=0; i--){ + x[i] = (nIn >> (7*(4-i)))&0x7f; + } + for(i=0; i<4 && x[i]==0; i++){} + for(j=0; i<=4; i++, j++) pOut[j] = x[i]; + pOut[j-1] |= 0x80; + } + rc = compress(&pOut[j], &nOut, pIn, nIn); + if( rc==Z_OK ){ + sqlite3_result_blob(context, pOut, nOut+j, sqlite3_free); + }else{ + sqlite3_free(pOut); + sqlite3_result_error(context, "input cannot be zlib compressed", -1); + } +} + +/* +** Implementation of the "uncompress(X)" SQL function. The argument X +** is a blob which was obtained from compress(Y). The output will be +** the value Y. +*/ +static void uncompressFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *pIn; + unsigned char *pOut; + unsigned int nIn; + unsigned long nOut; + int rc; + + pIn = sqlite3_value_blob(argv[0]); + nIn = sqlite3_value_bytes(argv[0]); + nOut = (pIn[0]<<24) + (pIn[1]<<16) + (pIn[2]<<8) + pIn[3]; + if( pIn[0] ){ + unsigned long nOut2 = 0; + int i; + + /* + ** If the high-byte of the blob length > 0, there are actually + ** two possibilities: + ** 1) The blob is > 16MByte, possible but unlikely as most blobs + ** are not that big. + ** 2) The content was compressed with SQLite's ext/misc/compress.c + ** + ** Just try both possibilities (smallest first), if the decompression + ** fails (either by Z_BUF_ERROR or Z_DATA_ERROR) we will find out quick + ** enough which one was correct. + */ + for(i=0; i + +/* +** Structure used to accumulate the output +*/ +struct EvalResult { + char *z; /* Accumulated output */ + const char *zSep; /* Separator */ + int szSep; /* Size of the separator string */ + sqlite3_int64 nAlloc; /* Number of bytes allocated for z[] */ + sqlite3_int64 nUsed; /* Number of bytes of z[] actually used */ +}; + +/* +** Callback from sqlite_exec() for the eval() function. +*/ +static int callback(void *pCtx, int argc, char **argv, char **colnames){ + struct EvalResult *p = (struct EvalResult*)pCtx; + int i; + for(i=0; inUsed+p->szSep+1 > p->nAlloc ){ + char *zNew; + p->nAlloc = p->nAlloc*2 + sz + p->szSep + 1; + /* Using sqlite3_realloc64() would be better, but it is a recent + ** addition and will cause a segfault if loaded by an older version + ** of SQLite. */ + zNew = p->nAlloc<=0x7fffffff ? sqlite3_realloc(p->z, (int)p->nAlloc) : 0; + if( zNew==0 ){ + sqlite3_free(p->z); + memset(p, 0, sizeof(*p)); + return 1; + } + p->z = zNew; + } + if( p->nUsed>0 ){ + memcpy(&p->z[p->nUsed], p->zSep, p->szSep); + p->nUsed += p->szSep; + } + memcpy(&p->z[p->nUsed], z, sz); + p->nUsed += sz; + } + return 0; +} + +/* +** Implementation of the eval(X) and eval(X,Y) SQL functions. +** +** Evaluate the SQL text in X. Return the results, using string +** Y as the separator. If Y is omitted, use a single space character. +*/ +static void sqlEvalFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const char *zSql; + sqlite3 *db; + char *zErr = 0; + int rc; + struct EvalResult x; + + memset(&x, 0, sizeof(x)); + x.zSep = " "; + zSql = (const char*)sqlite3_value_text(argv[0]); + if( zSql==0 ) return; + if( argc>1 ){ + x.zSep = (const char*)sqlite3_value_text(argv[1]); + if( x.zSep==0 ) return; + } + x.szSep = (int)strlen(x.zSep); + db = sqlite3_context_db_handle(context); + rc = sqlite3_exec(db, zSql, callback, &x, &zErr); + if( rc!=SQLITE_OK ){ + sqlite3_result_error(context, zErr, -1); + sqlite3_free(zErr); + }else if( x.zSep==0 ){ + sqlite3_result_error_nomem(context); + sqlite3_free(x.z); + }else{ + sqlite3_result_text(context, x.z, (int)x.nUsed, sqlite3_free); + } +} + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_eval_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "eval", 1, SQLITE_UTF8, 0, + sqlEvalFunc, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "eval", 2, SQLITE_UTF8, 0, + sqlEvalFunc, 0, 0); + } + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "eval", 1, SQLITE_UTF8, 0, + sqlEvalFunc, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "eval", 2, SQLITE_UTF8, 0, + sqlEvalFunc, 0, 0); + } + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/fileio.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/fileio.c 2016-01-06 16:30:42.501121300 +0100 @@ -0,0 +1,118 @@ +/* +** 2014-06-13 +** +** 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 SQLite extension implements SQL functions readfile() and +** writefile(). +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include + +/* +** Implementation of the "readfile(X)" SQL function. The entire content +** of the file named X is read and returned as a BLOB. NULL is returned +** if the file does not exist or is unreadable. +*/ +static void readfileFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const char *zName; + FILE *in; + long nIn; + void *pBuf; + + zName = (const char*)sqlite3_value_text(argv[0]); + if( zName==0 ) return; + in = fopen(zName, "rb"); + if( in==0 ) return; + fseek(in, 0, SEEK_END); + nIn = ftell(in); + rewind(in); + pBuf = sqlite3_malloc( nIn ); + if( pBuf && 1==fread(pBuf, nIn, 1, in) ){ + sqlite3_result_blob(context, pBuf, nIn, sqlite3_free); + }else{ + sqlite3_free(pBuf); + } + fclose(in); +} + +/* +** Implementation of the "writefile(X,Y)" SQL function. The argument Y +** is written into file X. The number of bytes written is returned. Or +** NULL is returned if something goes wrong, such as being unable to open +** file X for writing. +*/ +static void writefileFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + FILE *out; + const char *z; + sqlite3_int64 rc; + const char *zFile; + + zFile = (const char*)sqlite3_value_text(argv[0]); + if( zFile==0 ) return; + out = fopen(zFile, "wb"); + if( out==0 ) return; + z = (const char*)sqlite3_value_blob(argv[1]); + if( z==0 ){ + rc = 0; + }else{ + rc = fwrite(z, 1, sqlite3_value_bytes(argv[1]), out); + } + fclose(out); + sqlite3_result_int64(context, rc); +} + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_fileio_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "readfile", 1, SQLITE_UTF8, 0, + readfileFunc, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "writefile", 2, SQLITE_UTF8, 0, + writefileFunc, 0, 0); + } + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "readfile", 1, SQLITE_UTF8, 0, + readfileFunc, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "writefile", 2, SQLITE_UTF8, 0, + writefileFunc, 0, 0); + } + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/fuzzer.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/fuzzer.c 2016-01-06 16:30:42.503121400 +0100 @@ -0,0 +1,1199 @@ +/* +** 2011 March 24 +** +** 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. +** +************************************************************************* +** +** Code for a demonstration virtual table that generates variations +** on an input word at increasing edit distances from the original. +** +** A fuzzer virtual table is created like this: +** +** CREATE VIRTUAL TABLE f USING fuzzer(); +** +** When it is created, the new fuzzer table must be supplied with the +** name of a "fuzzer data table", which must reside in the same database +** file as the new fuzzer table. The fuzzer data table contains the various +** transformations and their costs that the fuzzer logic uses to generate +** variations. +** +** The fuzzer data table must contain exactly four columns (more precisely, +** the statement "SELECT * FROM " must return records +** that consist of four columns). It does not matter what the columns are +** named. +** +** Each row in the fuzzer data table represents a single character +** transformation. The left most column of the row (column 0) contains an +** integer value - the identifier of the ruleset to which the transformation +** rule belongs (see "MULTIPLE RULE SETS" below). The second column of the +** row (column 0) contains the input character or characters. The third +** column contains the output character or characters. And the fourth column +** contains the integer cost of making the transformation. For example: +** +** CREATE TABLE f_data(ruleset, cFrom, cTo, Cost); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, '', 'a', 100); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'b', '', 87); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'o', 'oe', 38); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'oe', 'o', 40); +** +** The first row inserted into the fuzzer data table by the SQL script +** above indicates that the cost of inserting a letter 'a' is 100. (All +** costs are integers. We recommend that costs be scaled so that the +** average cost is around 100.) The second INSERT statement creates a rule +** saying that the cost of deleting a single letter 'b' is 87. The third +** and fourth INSERT statements mean that the cost of transforming a +** single letter "o" into the two-letter sequence "oe" is 38 and that the +** cost of transforming "oe" back into "o" is 40. +** +** The contents of the fuzzer data table are loaded into main memory when +** a fuzzer table is first created, and may be internally reloaded by the +** system at any subsequent time. Therefore, the fuzzer data table should be +** populated before the fuzzer table is created and not modified thereafter. +** If you do need to modify the contents of the fuzzer data table, it is +** recommended that the associated fuzzer table be dropped, the fuzzer data +** table edited, and the fuzzer table recreated within a single transaction. +** Alternatively, the fuzzer data table can be edited then the database +** connection can be closed and reopened. +** +** Once it has been created, the fuzzer table can be queried as follows: +** +** SELECT word, distance FROM f +** WHERE word MATCH 'abcdefg' +** AND distance<200; +** +** This first query outputs the string "abcdefg" and all strings that +** can be derived from that string by appling the specified transformations. +** The strings are output together with their total transformation cost +** (called "distance") and appear in order of increasing cost. No string +** is output more than once. If there are multiple ways to transform the +** target string into the output string then the lowest cost transform is +** the one that is returned. In the example, the search is limited to +** strings with a total distance of less than 200. +** +** The fuzzer is a read-only table. Any attempt to DELETE, INSERT, or +** UPDATE on a fuzzer table will throw an error. +** +** It is important to put some kind of a limit on the fuzzer output. This +** can be either in the form of a LIMIT clause at the end of the query, +** or better, a "distance +#include +#include +#include + +#ifndef SQLITE_OMIT_VIRTUALTABLE + +/* +** Forward declaration of objects used by this implementation +*/ +typedef struct fuzzer_vtab fuzzer_vtab; +typedef struct fuzzer_cursor fuzzer_cursor; +typedef struct fuzzer_rule fuzzer_rule; +typedef struct fuzzer_seen fuzzer_seen; +typedef struct fuzzer_stem fuzzer_stem; + +/* +** Various types. +** +** fuzzer_cost is the "cost" of an edit operation. +** +** fuzzer_len is the length of a matching string. +** +** fuzzer_ruleid is an ruleset identifier. +*/ +typedef int fuzzer_cost; +typedef signed char fuzzer_len; +typedef int fuzzer_ruleid; + +/* +** Limits +*/ +#define FUZZER_MX_LENGTH 50 /* Maximum length of a rule string */ +#define FUZZER_MX_RULEID 2147483647 /* Maximum rule ID */ +#define FUZZER_MX_COST 1000 /* Maximum single-rule cost */ +#define FUZZER_MX_OUTPUT_LENGTH 100 /* Maximum length of an output string */ + + +/* +** Each transformation rule is stored as an instance of this object. +** All rules are kept on a linked list sorted by rCost. +*/ +struct fuzzer_rule { + fuzzer_rule *pNext; /* Next rule in order of increasing rCost */ + char *zFrom; /* Transform from */ + fuzzer_cost rCost; /* Cost of this transformation */ + fuzzer_len nFrom, nTo; /* Length of the zFrom and zTo strings */ + fuzzer_ruleid iRuleset; /* The rule set to which this rule belongs */ + char zTo[4]; /* Transform to (extra space appended) */ +}; + +/* +** A stem object is used to generate variants. It is also used to record +** previously generated outputs. +** +** Every stem is added to a hash table as it is output. Generation of +** duplicate stems is suppressed. +** +** Active stems (those that might generate new outputs) are kepts on a linked +** list sorted by increasing cost. The cost is the sum of rBaseCost and +** pRule->rCost. +*/ +struct fuzzer_stem { + char *zBasis; /* Word being fuzzed */ + const fuzzer_rule *pRule; /* Current rule to apply */ + fuzzer_stem *pNext; /* Next stem in rCost order */ + fuzzer_stem *pHash; /* Next stem with same hash on zBasis */ + fuzzer_cost rBaseCost; /* Base cost of getting to zBasis */ + fuzzer_cost rCostX; /* Precomputed rBaseCost + pRule->rCost */ + fuzzer_len nBasis; /* Length of the zBasis string */ + fuzzer_len n; /* Apply pRule at this character offset */ +}; + +/* +** A fuzzer virtual-table object +*/ +struct fuzzer_vtab { + sqlite3_vtab base; /* Base class - must be first */ + char *zClassName; /* Name of this class. Default: "fuzzer" */ + fuzzer_rule *pRule; /* All active rules in this fuzzer */ + int nCursor; /* Number of active cursors */ +}; + +#define FUZZER_HASH 4001 /* Hash table size */ +#define FUZZER_NQUEUE 20 /* Number of slots on the stem queue */ + +/* A fuzzer cursor object */ +struct fuzzer_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + sqlite3_int64 iRowid; /* The rowid of the current word */ + fuzzer_vtab *pVtab; /* The virtual table this cursor belongs to */ + fuzzer_cost rLimit; /* Maximum cost of any term */ + fuzzer_stem *pStem; /* Stem with smallest rCostX */ + fuzzer_stem *pDone; /* Stems already processed to completion */ + fuzzer_stem *aQueue[FUZZER_NQUEUE]; /* Queue of stems with higher rCostX */ + int mxQueue; /* Largest used index in aQueue[] */ + char *zBuf; /* Temporary use buffer */ + int nBuf; /* Bytes allocated for zBuf */ + int nStem; /* Number of stems allocated */ + int iRuleset; /* Only process rules from this ruleset */ + fuzzer_rule nullRule; /* Null rule used first */ + fuzzer_stem *apHash[FUZZER_HASH]; /* Hash of previously generated terms */ +}; + +/* +** The two input rule lists are both sorted in order of increasing +** cost. Merge them together into a single list, sorted by cost, and +** return a pointer to the head of that list. +*/ +static fuzzer_rule *fuzzerMergeRules(fuzzer_rule *pA, fuzzer_rule *pB){ + fuzzer_rule head; + fuzzer_rule *pTail; + + pTail = &head; + while( pA && pB ){ + if( pA->rCost<=pB->rCost ){ + pTail->pNext = pA; + pTail = pA; + pA = pA->pNext; + }else{ + pTail->pNext = pB; + pTail = pB; + pB = pB->pNext; + } + } + if( pA==0 ){ + pTail->pNext = pB; + }else{ + pTail->pNext = pA; + } + return head.pNext; +} + +/* +** Statement pStmt currently points to a row in the fuzzer data table. This +** function allocates and populates a fuzzer_rule structure according to +** the content of the row. +** +** If successful, *ppRule is set to point to the new object and SQLITE_OK +** is returned. Otherwise, *ppRule is zeroed, *pzErr may be set to point +** to an error message and an SQLite error code returned. +*/ +static int fuzzerLoadOneRule( + fuzzer_vtab *p, /* Fuzzer virtual table handle */ + sqlite3_stmt *pStmt, /* Base rule on statements current row */ + fuzzer_rule **ppRule, /* OUT: New rule object */ + char **pzErr /* OUT: Error message */ +){ + sqlite3_int64 iRuleset = sqlite3_column_int64(pStmt, 0); + const char *zFrom = (const char *)sqlite3_column_text(pStmt, 1); + const char *zTo = (const char *)sqlite3_column_text(pStmt, 2); + int nCost = sqlite3_column_int(pStmt, 3); + + int rc = SQLITE_OK; /* Return code */ + int nFrom; /* Size of string zFrom, in bytes */ + int nTo; /* Size of string zTo, in bytes */ + fuzzer_rule *pRule = 0; /* New rule object to return */ + + if( zFrom==0 ) zFrom = ""; + if( zTo==0 ) zTo = ""; + nFrom = (int)strlen(zFrom); + nTo = (int)strlen(zTo); + + /* Silently ignore null transformations */ + if( strcmp(zFrom, zTo)==0 ){ + *ppRule = 0; + return SQLITE_OK; + } + + if( nCost<=0 || nCost>FUZZER_MX_COST ){ + *pzErr = sqlite3_mprintf("%s: cost must be between 1 and %d", + p->zClassName, FUZZER_MX_COST + ); + rc = SQLITE_ERROR; + }else + if( nFrom>FUZZER_MX_LENGTH || nTo>FUZZER_MX_LENGTH ){ + *pzErr = sqlite3_mprintf("%s: maximum string length is %d", + p->zClassName, FUZZER_MX_LENGTH + ); + rc = SQLITE_ERROR; + }else + if( iRuleset<0 || iRuleset>FUZZER_MX_RULEID ){ + *pzErr = sqlite3_mprintf("%s: ruleset must be between 0 and %d", + p->zClassName, FUZZER_MX_RULEID + ); + rc = SQLITE_ERROR; + }else{ + + pRule = sqlite3_malloc( sizeof(*pRule) + nFrom + nTo ); + if( pRule==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pRule, 0, sizeof(*pRule)); + pRule->zFrom = pRule->zTo; + pRule->zFrom += nTo + 1; + pRule->nFrom = nFrom; + memcpy(pRule->zFrom, zFrom, nFrom+1); + memcpy(pRule->zTo, zTo, nTo+1); + pRule->nTo = nTo; + pRule->rCost = nCost; + pRule->iRuleset = (int)iRuleset; + } + } + + *ppRule = pRule; + return rc; +} + +/* +** Load the content of the fuzzer data table into memory. +*/ +static int fuzzerLoadRules( + sqlite3 *db, /* Database handle */ + fuzzer_vtab *p, /* Virtual fuzzer table to configure */ + const char *zDb, /* Database containing rules data */ + const char *zData, /* Table containing rules data */ + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; /* Return code */ + char *zSql; /* SELECT used to read from rules table */ + fuzzer_rule *pHead = 0; + + zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zData); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + int rc2; /* finalize() return code */ + sqlite3_stmt *pStmt = 0; + rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); + if( rc!=SQLITE_OK ){ + *pzErr = sqlite3_mprintf("%s: %s", p->zClassName, sqlite3_errmsg(db)); + }else if( sqlite3_column_count(pStmt)!=4 ){ + *pzErr = sqlite3_mprintf("%s: %s has %d columns, expected 4", + p->zClassName, zData, sqlite3_column_count(pStmt) + ); + rc = SQLITE_ERROR; + }else{ + while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ + fuzzer_rule *pRule = 0; + rc = fuzzerLoadOneRule(p, pStmt, &pRule, pzErr); + if( pRule ){ + pRule->pNext = pHead; + pHead = pRule; + } + } + } + rc2 = sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ) rc = rc2; + } + sqlite3_free(zSql); + + /* All rules are now in a singly linked list starting at pHead. This + ** block sorts them by cost and then sets fuzzer_vtab.pRule to point to + ** point to the head of the sorted list. + */ + if( rc==SQLITE_OK ){ + unsigned int i; + fuzzer_rule *pX; + fuzzer_rule *a[15]; + for(i=0; ipNext; + pX->pNext = 0; + for(i=0; a[i] && ipRule = fuzzerMergeRules(p->pRule, pX); + }else{ + /* An error has occurred. Setting p->pRule to point to the head of the + ** allocated list ensures that the list will be cleaned up in this case. + */ + assert( p->pRule==0 ); + p->pRule = pHead; + } + + return rc; +} + +/* +** This function converts an SQL quoted string into an unquoted string +** and returns a pointer to a buffer allocated using sqlite3_malloc() +** containing the result. The caller should eventually free this buffer +** using sqlite3_free. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static char *fuzzerDequote(const char *zIn){ + int nIn; /* Size of input string, in bytes */ + char *zOut; /* Output (dequoted) string */ + + nIn = (int)strlen(zIn); + zOut = sqlite3_malloc(nIn+1); + if( zOut ){ + char q = zIn[0]; /* Quote character (if any ) */ + + if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ + memcpy(zOut, zIn, nIn+1); + }else{ + int iOut = 0; /* Index of next byte to write to output */ + int iIn; /* Index of next byte to read from input */ + + if( q=='[' ) q = ']'; + for(iIn=1; iInnCursor==0 ); + while( p->pRule ){ + fuzzer_rule *pRule = p->pRule; + p->pRule = pRule->pNext; + sqlite3_free(pRule); + } + sqlite3_free(p); + return SQLITE_OK; +} + +/* +** xConnect/xCreate method for the fuzzer module. Arguments are: +** +** argv[0] -> module name ("fuzzer") +** argv[1] -> database name +** argv[2] -> table name +** argv[3] -> fuzzer rule table name +*/ +static int fuzzerConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + int rc = SQLITE_OK; /* Return code */ + fuzzer_vtab *pNew = 0; /* New virtual table */ + const char *zModule = argv[0]; + const char *zDb = argv[1]; + + if( argc!=4 ){ + *pzErr = sqlite3_mprintf( + "%s: wrong number of CREATE VIRTUAL TABLE arguments", zModule + ); + rc = SQLITE_ERROR; + }else{ + int nModule; /* Length of zModule, in bytes */ + + nModule = (int)strlen(zModule); + pNew = sqlite3_malloc( sizeof(*pNew) + nModule + 1); + if( pNew==0 ){ + rc = SQLITE_NOMEM; + }else{ + char *zTab; /* Dequoted name of fuzzer data table */ + + memset(pNew, 0, sizeof(*pNew)); + pNew->zClassName = (char*)&pNew[1]; + memcpy(pNew->zClassName, zModule, nModule+1); + + zTab = fuzzerDequote(argv[3]); + if( zTab==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = fuzzerLoadRules(db, pNew, zDb, zTab, pzErr); + sqlite3_free(zTab); + } + + if( rc==SQLITE_OK ){ + rc = sqlite3_declare_vtab(db, "CREATE TABLE x(word,distance,ruleset)"); + } + if( rc!=SQLITE_OK ){ + fuzzerDisconnect((sqlite3_vtab *)pNew); + pNew = 0; + } + } + } + + *ppVtab = (sqlite3_vtab *)pNew; + return rc; +} + +/* +** Open a new fuzzer cursor. +*/ +static int fuzzerOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + fuzzer_vtab *p = (fuzzer_vtab*)pVTab; + fuzzer_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVtab = p; + *ppCursor = &pCur->base; + p->nCursor++; + return SQLITE_OK; +} + +/* +** Free all stems in a list. +*/ +static void fuzzerClearStemList(fuzzer_stem *pStem){ + while( pStem ){ + fuzzer_stem *pNext = pStem->pNext; + sqlite3_free(pStem); + pStem = pNext; + } +} + +/* +** Free up all the memory allocated by a cursor. Set it rLimit to 0 +** to indicate that it is at EOF. +*/ +static void fuzzerClearCursor(fuzzer_cursor *pCur, int clearHash){ + int i; + fuzzerClearStemList(pCur->pStem); + fuzzerClearStemList(pCur->pDone); + for(i=0; iaQueue[i]); + pCur->rLimit = (fuzzer_cost)0; + if( clearHash && pCur->nStem ){ + pCur->mxQueue = 0; + pCur->pStem = 0; + pCur->pDone = 0; + memset(pCur->aQueue, 0, sizeof(pCur->aQueue)); + memset(pCur->apHash, 0, sizeof(pCur->apHash)); + } + pCur->nStem = 0; +} + +/* +** Close a fuzzer cursor. +*/ +static int fuzzerClose(sqlite3_vtab_cursor *cur){ + fuzzer_cursor *pCur = (fuzzer_cursor *)cur; + fuzzerClearCursor(pCur, 0); + sqlite3_free(pCur->zBuf); + pCur->pVtab->nCursor--; + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Compute the current output term for a fuzzer_stem. +*/ +static int fuzzerRender( + fuzzer_stem *pStem, /* The stem to be rendered */ + char **pzBuf, /* Write results into this buffer. realloc if needed */ + int *pnBuf /* Size of the buffer */ +){ + const fuzzer_rule *pRule = pStem->pRule; + int n; /* Size of output term without nul-term */ + char *z; /* Buffer to assemble output term in */ + + n = pStem->nBasis + pRule->nTo - pRule->nFrom; + if( (*pnBuf)n; + z = *pzBuf; + if( n<0 ){ + memcpy(z, pStem->zBasis, pStem->nBasis+1); + }else{ + memcpy(z, pStem->zBasis, n); + memcpy(&z[n], pRule->zTo, pRule->nTo); + memcpy(&z[n+pRule->nTo], &pStem->zBasis[n+pRule->nFrom], + pStem->nBasis-n-pRule->nFrom+1); + } + + assert( z[pStem->nBasis + pRule->nTo - pRule->nFrom]==0 ); + return SQLITE_OK; +} + +/* +** Compute a hash on zBasis. +*/ +static unsigned int fuzzerHash(const char *z){ + unsigned int h = 0; + while( *z ){ h = (h<<3) ^ (h>>29) ^ *(z++); } + return h % FUZZER_HASH; +} + +/* +** Current cost of a stem +*/ +static fuzzer_cost fuzzerCost(fuzzer_stem *pStem){ + return pStem->rCostX = pStem->rBaseCost + pStem->pRule->rCost; +} + +#if 0 +/* +** Print a description of a fuzzer_stem on stderr. +*/ +static void fuzzerStemPrint( + const char *zPrefix, + fuzzer_stem *pStem, + const char *zSuffix +){ + if( pStem->n<0 ){ + fprintf(stderr, "%s[%s](%d)-->self%s", + zPrefix, + pStem->zBasis, pStem->rBaseCost, + zSuffix + ); + }else{ + char *zBuf = 0; + int nBuf = 0; + if( fuzzerRender(pStem, &zBuf, &nBuf)!=SQLITE_OK ) return; + fprintf(stderr, "%s[%s](%d)-->{%s}(%d)%s", + zPrefix, + pStem->zBasis, pStem->rBaseCost, zBuf, pStem->, + zSuffix + ); + sqlite3_free(zBuf); + } +} +#endif + +/* +** Return 1 if the string to which the cursor is point has already +** been emitted. Return 0 if not. Return -1 on a memory allocation +** failures. +*/ +static int fuzzerSeen(fuzzer_cursor *pCur, fuzzer_stem *pStem){ + unsigned int h; + fuzzer_stem *pLookup; + + if( fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ + return -1; + } + h = fuzzerHash(pCur->zBuf); + pLookup = pCur->apHash[h]; + while( pLookup && strcmp(pLookup->zBasis, pCur->zBuf)!=0 ){ + pLookup = pLookup->pHash; + } + return pLookup!=0; +} + +/* +** If argument pRule is NULL, this function returns false. +** +** Otherwise, it returns true if rule pRule should be skipped. A rule +** should be skipped if it does not belong to rule-set iRuleset, or if +** applying it to stem pStem would create a string longer than +** FUZZER_MX_OUTPUT_LENGTH bytes. +*/ +static int fuzzerSkipRule( + const fuzzer_rule *pRule, /* Determine whether or not to skip this */ + fuzzer_stem *pStem, /* Stem rule may be applied to */ + int iRuleset /* Rule-set used by the current query */ +){ + return pRule && ( + (pRule->iRuleset!=iRuleset) + || (pStem->nBasis + pRule->nTo - pRule->nFrom)>FUZZER_MX_OUTPUT_LENGTH + ); +} + +/* +** Advance a fuzzer_stem to its next value. Return 0 if there are +** no more values that can be generated by this fuzzer_stem. Return +** -1 on a memory allocation failure. +*/ +static int fuzzerAdvance(fuzzer_cursor *pCur, fuzzer_stem *pStem){ + const fuzzer_rule *pRule; + while( (pRule = pStem->pRule)!=0 ){ + assert( pRule==&pCur->nullRule || pRule->iRuleset==pCur->iRuleset ); + while( pStem->n < pStem->nBasis - pRule->nFrom ){ + pStem->n++; + if( pRule->nFrom==0 + || memcmp(&pStem->zBasis[pStem->n], pRule->zFrom, pRule->nFrom)==0 + ){ + /* Found a rewrite case. Make sure it is not a duplicate */ + int rc = fuzzerSeen(pCur, pStem); + if( rc<0 ) return -1; + if( rc==0 ){ + fuzzerCost(pStem); + return 1; + } + } + } + pStem->n = -1; + do{ + pRule = pRule->pNext; + }while( fuzzerSkipRule(pRule, pStem, pCur->iRuleset) ); + pStem->pRule = pRule; + if( pRule && fuzzerCost(pStem)>pCur->rLimit ) pStem->pRule = 0; + } + return 0; +} + +/* +** The two input stem lists are both sorted in order of increasing +** rCostX. Merge them together into a single list, sorted by rCostX, and +** return a pointer to the head of that new list. +*/ +static fuzzer_stem *fuzzerMergeStems(fuzzer_stem *pA, fuzzer_stem *pB){ + fuzzer_stem head; + fuzzer_stem *pTail; + + pTail = &head; + while( pA && pB ){ + if( pA->rCostX<=pB->rCostX ){ + pTail->pNext = pA; + pTail = pA; + pA = pA->pNext; + }else{ + pTail->pNext = pB; + pTail = pB; + pB = pB->pNext; + } + } + if( pA==0 ){ + pTail->pNext = pB; + }else{ + pTail->pNext = pA; + } + return head.pNext; +} + +/* +** Load pCur->pStem with the lowest-cost stem. Return a pointer +** to the lowest-cost stem. +*/ +static fuzzer_stem *fuzzerLowestCostStem(fuzzer_cursor *pCur){ + fuzzer_stem *pBest, *pX; + int iBest; + int i; + + if( pCur->pStem==0 ){ + iBest = -1; + pBest = 0; + for(i=0; i<=pCur->mxQueue; i++){ + pX = pCur->aQueue[i]; + if( pX==0 ) continue; + if( pBest==0 || pBest->rCostX>pX->rCostX ){ + pBest = pX; + iBest = i; + } + } + if( pBest ){ + pCur->aQueue[iBest] = pBest->pNext; + pBest->pNext = 0; + pCur->pStem = pBest; + } + } + return pCur->pStem; +} + +/* +** Insert pNew into queue of pending stems. Then find the stem +** with the lowest rCostX and move it into pCur->pStem. +** list. The insert is done such the pNew is in the correct order +** according to fuzzer_stem.zBaseCost+fuzzer_stem.pRule->rCost. +*/ +static fuzzer_stem *fuzzerInsert(fuzzer_cursor *pCur, fuzzer_stem *pNew){ + fuzzer_stem *pX; + int i; + + /* If pCur->pStem exists and is greater than pNew, then make pNew + ** the new pCur->pStem and insert the old pCur->pStem instead. + */ + if( (pX = pCur->pStem)!=0 && pX->rCostX>pNew->rCostX ){ + pNew->pNext = 0; + pCur->pStem = pNew; + pNew = pX; + } + + /* Insert the new value */ + pNew->pNext = 0; + pX = pNew; + for(i=0; i<=pCur->mxQueue; i++){ + if( pCur->aQueue[i] ){ + pX = fuzzerMergeStems(pX, pCur->aQueue[i]); + pCur->aQueue[i] = 0; + }else{ + pCur->aQueue[i] = pX; + break; + } + } + if( i>pCur->mxQueue ){ + if( imxQueue = i; + pCur->aQueue[i] = pX; + }else{ + assert( pCur->mxQueue==FUZZER_NQUEUE-1 ); + pX = fuzzerMergeStems(pX, pCur->aQueue[FUZZER_NQUEUE-1]); + pCur->aQueue[FUZZER_NQUEUE-1] = pX; + } + } + + return fuzzerLowestCostStem(pCur); +} + +/* +** Allocate a new fuzzer_stem. Add it to the hash table but do not +** link it into either the pCur->pStem or pCur->pDone lists. +*/ +static fuzzer_stem *fuzzerNewStem( + fuzzer_cursor *pCur, + const char *zWord, + fuzzer_cost rBaseCost +){ + fuzzer_stem *pNew; + fuzzer_rule *pRule; + unsigned int h; + + pNew = sqlite3_malloc( sizeof(*pNew) + (int)strlen(zWord) + 1 ); + if( pNew==0 ) return 0; + memset(pNew, 0, sizeof(*pNew)); + pNew->zBasis = (char*)&pNew[1]; + pNew->nBasis = (fuzzer_len)strlen(zWord); + memcpy(pNew->zBasis, zWord, pNew->nBasis+1); + pRule = pCur->pVtab->pRule; + while( fuzzerSkipRule(pRule, pNew, pCur->iRuleset) ){ + pRule = pRule->pNext; + } + pNew->pRule = pRule; + pNew->n = -1; + pNew->rBaseCost = pNew->rCostX = rBaseCost; + h = fuzzerHash(pNew->zBasis); + pNew->pHash = pCur->apHash[h]; + pCur->apHash[h] = pNew; + pCur->nStem++; + return pNew; +} + + +/* +** Advance a cursor to its next row of output +*/ +static int fuzzerNext(sqlite3_vtab_cursor *cur){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + int rc; + fuzzer_stem *pStem, *pNew; + + pCur->iRowid++; + + /* Use the element the cursor is currently point to to create + ** a new stem and insert the new stem into the priority queue. + */ + pStem = pCur->pStem; + if( pStem->rCostX>0 ){ + rc = fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf); + if( rc==SQLITE_NOMEM ) return SQLITE_NOMEM; + pNew = fuzzerNewStem(pCur, pCur->zBuf, pStem->rCostX); + if( pNew ){ + if( fuzzerAdvance(pCur, pNew)==0 ){ + pNew->pNext = pCur->pDone; + pCur->pDone = pNew; + }else{ + if( fuzzerInsert(pCur, pNew)==pNew ){ + return SQLITE_OK; + } + } + }else{ + return SQLITE_NOMEM; + } + } + + /* Adjust the priority queue so that the first element of the + ** stem list is the next lowest cost word. + */ + while( (pStem = pCur->pStem)!=0 ){ + int res = fuzzerAdvance(pCur, pStem); + if( res<0 ){ + return SQLITE_NOMEM; + }else if( res>0 ){ + pCur->pStem = 0; + pStem = fuzzerInsert(pCur, pStem); + if( (rc = fuzzerSeen(pCur, pStem))!=0 ){ + if( rc<0 ) return SQLITE_NOMEM; + continue; + } + return SQLITE_OK; /* New word found */ + } + pCur->pStem = 0; + pStem->pNext = pCur->pDone; + pCur->pDone = pStem; + if( fuzzerLowestCostStem(pCur) ){ + rc = fuzzerSeen(pCur, pCur->pStem); + if( rc<0 ) return SQLITE_NOMEM; + if( rc==0 ){ + return SQLITE_OK; + } + } + } + + /* Reach this point only if queue has been exhausted and there is + ** nothing left to be output. */ + pCur->rLimit = (fuzzer_cost)0; + return SQLITE_OK; +} + +/* +** 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 fuzzerColumn, fuzzerRowid, or fuzzerEof call. +*/ +static int fuzzerFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + fuzzer_cursor *pCur = (fuzzer_cursor *)pVtabCursor; + const char *zWord = ""; + fuzzer_stem *pStem; + int idx; + + fuzzerClearCursor(pCur, 1); + pCur->rLimit = 2147483647; + idx = 0; + if( idxNum & 1 ){ + zWord = (const char*)sqlite3_value_text(argv[0]); + idx++; + } + if( idxNum & 2 ){ + pCur->rLimit = (fuzzer_cost)sqlite3_value_int(argv[idx]); + idx++; + } + if( idxNum & 4 ){ + pCur->iRuleset = (fuzzer_cost)sqlite3_value_int(argv[idx]); + idx++; + } + pCur->nullRule.pNext = pCur->pVtab->pRule; + pCur->nullRule.rCost = 0; + pCur->nullRule.nFrom = 0; + pCur->nullRule.nTo = 0; + pCur->nullRule.zFrom = ""; + pCur->iRowid = 1; + assert( pCur->pStem==0 ); + + /* If the query term is longer than FUZZER_MX_OUTPUT_LENGTH bytes, this + ** query will return zero rows. */ + if( (int)strlen(zWord)pStem = pStem = fuzzerNewStem(pCur, zWord, (fuzzer_cost)0); + if( pStem==0 ) return SQLITE_NOMEM; + pStem->pRule = &pCur->nullRule; + pStem->n = pStem->nBasis; + }else{ + pCur->rLimit = 0; + } + + return SQLITE_OK; +} + +/* +** Only the word and distance columns have values. All other columns +** return NULL +*/ +static int fuzzerColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + if( i==0 ){ + /* the "word" column */ + if( fuzzerRender(pCur->pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ + return SQLITE_NOMEM; + } + sqlite3_result_text(ctx, pCur->zBuf, -1, SQLITE_TRANSIENT); + }else if( i==1 ){ + /* the "distance" column */ + sqlite3_result_int(ctx, pCur->pStem->rCostX); + }else{ + /* All other columns are NULL */ + sqlite3_result_null(ctx); + } + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int fuzzerRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + *pRowid = pCur->iRowid; + return SQLITE_OK; +} + +/* +** When the fuzzer_cursor.rLimit value is 0 or less, that is a signal +** that the cursor has nothing more to output. +*/ +static int fuzzerEof(sqlite3_vtab_cursor *cur){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + return pCur->rLimit<=(fuzzer_cost)0; +} + +/* +** Search for terms of these forms: +** +** (A) word MATCH $str +** (B1) distance < $value +** (B2) distance <= $value +** (C) ruleid == $ruleid +** +** The distance< and distance<= are both treated as distance<=. +** The query plan number is a bit vector: +** +** bit 1: Term of the form (A) found +** bit 2: Term like (B1) or (B2) found +** bit 3: Term like (C) found +** +** If bit-1 is set, $str is always in filter.argv[0]. If bit-2 is set +** then $value is in filter.argv[0] if bit-1 is clear and is in +** filter.argv[1] if bit-1 is set. If bit-3 is set, then $ruleid is +** in filter.argv[0] if bit-1 and bit-2 are both zero, is in +** filter.argv[1] if exactly one of bit-1 and bit-2 are set, and is in +** filter.argv[2] if both bit-1 and bit-2 are set. +*/ +static int fuzzerBestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ + int iPlan = 0; + int iDistTerm = -1; + int iRulesetTerm = -1; + int i; + int seenMatch = 0; + const struct sqlite3_index_constraint *pConstraint; + double rCost = 1e12; + + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->iColumn==0 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH ){ + seenMatch = 1; + } + if( pConstraint->usable==0 ) continue; + if( (iPlan & 1)==0 + && pConstraint->iColumn==0 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + rCost /= 1e6; + } + if( (iPlan & 2)==0 + && pConstraint->iColumn==1 + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) + ){ + iPlan |= 2; + iDistTerm = i; + rCost /= 10.0; + } + if( (iPlan & 4)==0 + && pConstraint->iColumn==2 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 4; + pIdxInfo->aConstraintUsage[i].omit = 1; + iRulesetTerm = i; + rCost /= 10.0; + } + } + if( iPlan & 2 ){ + pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = 1+((iPlan&1)!=0); + } + if( iPlan & 4 ){ + int idx = 1; + if( iPlan & 1 ) idx++; + if( iPlan & 2 ) idx++; + pIdxInfo->aConstraintUsage[iRulesetTerm].argvIndex = idx; + } + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==1 + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + if( seenMatch && (iPlan&1)==0 ) rCost = 1e99; + pIdxInfo->estimatedCost = rCost; + + return SQLITE_OK; +} + +/* +** A virtual table module that implements the "fuzzer". +*/ +static const sqlite3_module fuzzerModule = { + 0, /* iVersion */ + fuzzerConnect, + fuzzerConnect, + fuzzerBestIndex, + fuzzerDisconnect, + fuzzerDisconnect, + fuzzerOpen, /* xOpen - open a cursor */ + fuzzerClose, /* xClose - close a cursor */ + fuzzerFilter, /* xFilter - configure scan constraints */ + fuzzerNext, /* xNext - advance a cursor */ + fuzzerEof, /* xEof - check for end of scan */ + fuzzerColumn, /* xColumn - read data */ + fuzzerRowid, /* xRowid - read data */ + 0, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ +}; + +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_fuzzer_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "fuzzer", &fuzzerModule, 0); +#endif + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "fuzzer", &fuzzerModule, 0); +#endif + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/icu.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/icu.c 2016-01-06 16:30:35.371713500 +0100 @@ -0,0 +1,514 @@ +/* +** 2007 May 6 +** +** 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. +** +************************************************************************* +** $Id: icu.c,v 1.7 2007/12/13 21:54:11 drh Exp $ +** +** This file implements an integration between the ICU library +** ("International Components for Unicode", an open-source library +** for handling unicode data) and SQLite. The integration uses +** ICU to provide the following to SQLite: +** +** * An implementation of the SQL regexp() function (and hence REGEXP +** operator) using the ICU uregex_XX() APIs. +** +** * Implementations of the SQL scalar upper() and lower() functions +** for case mapping. +** +** * Integration of ICU and SQLite collation sequences. +** +** * An implementation of the LIKE operator that uses ICU to +** provide case-independent matching. +*/ + +#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_ICU) + +/* Include ICU headers */ +#include +#include +#include +#include + +#include + +#ifndef SQLITE_CORE + #include "sqlite3ext.h" + SQLITE_EXTENSION_INIT1 +#else + #include "sqlite3.h" +#endif + +/* +** Maximum length (in bytes) of the pattern in a LIKE or GLOB +** operator. +*/ +#ifndef SQLITE_MAX_LIKE_PATTERN_LENGTH +# define SQLITE_MAX_LIKE_PATTERN_LENGTH 50000 +#endif + +/* +** Version of sqlite3_free() that is always a function, never a macro. +*/ +static void xFree(void *p){ + sqlite3_free(p); +} + +/* +** Compare two UTF-8 strings for equality where the first string is +** a "LIKE" expression. Return true (1) if they are the same and +** false (0) if they are different. +*/ +static int icuLikeCompare( + const uint8_t *zPattern, /* LIKE pattern */ + const uint8_t *zString, /* The UTF-8 string to compare against */ + const UChar32 uEsc /* The escape character */ +){ + static const int MATCH_ONE = (UChar32)'_'; + static const int MATCH_ALL = (UChar32)'%'; + + int iPattern = 0; /* Current byte index in zPattern */ + int iString = 0; /* Current byte index in zString */ + + int prevEscape = 0; /* True if the previous character was uEsc */ + + while( zPattern[iPattern]!=0 ){ + + /* Read (and consume) the next character from the input pattern. */ + UChar32 uPattern; + U8_NEXT_UNSAFE(zPattern, iPattern, uPattern); + + /* There are now 4 possibilities: + ** + ** 1. uPattern is an unescaped match-all character "%", + ** 2. uPattern is an unescaped match-one character "_", + ** 3. uPattern is an unescaped escape character, or + ** 4. uPattern is to be handled as an ordinary character + */ + if( !prevEscape && uPattern==MATCH_ALL ){ + /* Case 1. */ + uint8_t c; + + /* Skip any MATCH_ALL or MATCH_ONE characters that follow a + ** MATCH_ALL. For each MATCH_ONE, skip one character in the + ** test string. + */ + while( (c=zPattern[iPattern]) == MATCH_ALL || c == MATCH_ONE ){ + if( c==MATCH_ONE ){ + if( zString[iString]==0 ) return 0; + U8_FWD_1_UNSAFE(zString, iString); + } + iPattern++; + } + + if( zPattern[iPattern]==0 ) return 1; + + while( zString[iString] ){ + if( icuLikeCompare(&zPattern[iPattern], &zString[iString], uEsc) ){ + return 1; + } + U8_FWD_1_UNSAFE(zString, iString); + } + return 0; + + }else if( !prevEscape && uPattern==MATCH_ONE ){ + /* Case 2. */ + if( zString[iString]==0 ) return 0; + U8_FWD_1_UNSAFE(zString, iString); + + }else if( !prevEscape && uPattern==uEsc){ + /* Case 3. */ + prevEscape = 1; + + }else{ + /* Case 4. */ + UChar32 uString; + U8_NEXT_UNSAFE(zString, iString, uString); + uString = u_foldCase(uString, U_FOLD_CASE_DEFAULT); + uPattern = u_foldCase(uPattern, U_FOLD_CASE_DEFAULT); + if( uString!=uPattern ){ + return 0; + } + prevEscape = 0; + } + } + + return zString[iString]==0; +} + +/* +** Implementation of the like() SQL function. This function implements +** the build-in LIKE operator. The first argument to the function is the +** pattern and the second argument is the string. So, the SQL statements: +** +** A LIKE B +** +** is implemented as like(B, A). If there is an escape character E, +** +** A LIKE B ESCAPE E +** +** is mapped to like(B, A, E). +*/ +static void icuLikeFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zA = sqlite3_value_text(argv[0]); + const unsigned char *zB = sqlite3_value_text(argv[1]); + UChar32 uEsc = 0; + + /* Limit the length of the LIKE or GLOB pattern to avoid problems + ** of deep recursion and N*N behavior in patternCompare(). + */ + if( sqlite3_value_bytes(argv[0])>SQLITE_MAX_LIKE_PATTERN_LENGTH ){ + sqlite3_result_error(context, "LIKE or GLOB pattern too complex", -1); + return; + } + + + if( argc==3 ){ + /* The escape character string must consist of a single UTF-8 character. + ** Otherwise, return an error. + */ + size_t nE= sqlite3_value_bytes(argv[2]); + const unsigned char *zE = sqlite3_value_text(argv[2]); + int i = 0; + if( zE==0 ) return; + U8_NEXT(zE, i, nE, uEsc); + if( i!=nE){ + sqlite3_result_error(context, + "ESCAPE expression must be a single character", -1); + return; + } + } + + if( zA && zB ){ + sqlite3_result_int(context, icuLikeCompare(zA, zB, uEsc)); + } +} + +/* +** This function is called when an ICU function called from within +** the implementation of an SQL scalar function returns an error. +** +** The scalar function context passed as the first argument is +** loaded with an error message based on the following two args. +*/ +static void icuFunctionError( + sqlite3_context *pCtx, /* SQLite scalar function context */ + const char *zName, /* Name of ICU function that failed */ + UErrorCode e /* Error code returned by ICU function */ +){ + char zBuf[128]; + sqlite3_snprintf(128, zBuf, "ICU error: %s(): %s", zName, u_errorName(e)); + zBuf[127] = '\0'; + sqlite3_result_error(pCtx, zBuf, -1); +} + +/* +** Function to delete compiled regexp objects. Registered as +** a destructor function with sqlite3_set_auxdata(). +*/ +static void icuRegexpDelete(void *p){ + URegularExpression *pExpr = (URegularExpression *)p; + uregex_close(pExpr); +} + +/* +** Implementation of SQLite REGEXP operator. This scalar function takes +** two arguments. The first is a regular expression pattern to compile +** the second is a string to match against that pattern. If either +** argument is an SQL NULL, then NULL Is returned. Otherwise, the result +** is 1 if the string matches the pattern, or 0 otherwise. +** +** SQLite maps the regexp() function to the regexp() operator such +** that the following two are equivalent: +** +** zString REGEXP zPattern +** regexp(zPattern, zString) +** +** Uses the following ICU regexp APIs: +** +** uregex_open() +** uregex_matches() +** uregex_close() +*/ +static void icuRegexpFunc(sqlite3_context *p, int nArg, sqlite3_value **apArg){ + UErrorCode status = U_ZERO_ERROR; + URegularExpression *pExpr; + UBool res; + const UChar *zString = sqlite3_value_text16(apArg[1]); + + (void)nArg; /* Unused parameter */ + + /* If the left hand side of the regexp operator is NULL, + ** then the result is also NULL. + */ + if( !zString ){ + return; + } + + pExpr = sqlite3_get_auxdata(p, 0); + if( !pExpr ){ + const UChar *zPattern = sqlite3_value_text16(apArg[0]); + if( !zPattern ){ + return; + } + pExpr = uregex_open(zPattern, -1, 0, 0, &status); + + if( U_SUCCESS(status) ){ + sqlite3_set_auxdata(p, 0, pExpr, icuRegexpDelete); + }else{ + assert(!pExpr); + icuFunctionError(p, "uregex_open", status); + return; + } + } + + /* Configure the text that the regular expression operates on. */ + uregex_setText(pExpr, zString, -1, &status); + if( !U_SUCCESS(status) ){ + icuFunctionError(p, "uregex_setText", status); + return; + } + + /* Attempt the match */ + res = uregex_matches(pExpr, 0, &status); + if( !U_SUCCESS(status) ){ + icuFunctionError(p, "uregex_matches", status); + return; + } + + /* Set the text that the regular expression operates on to a NULL + ** pointer. This is not really necessary, but it is tidier than + ** leaving the regular expression object configured with an invalid + ** pointer after this function returns. + */ + uregex_setText(pExpr, 0, 0, &status); + + /* Return 1 or 0. */ + sqlite3_result_int(p, res ? 1 : 0); +} + +/* +** Implementations of scalar functions for case mapping - upper() and +** lower(). Function upper() converts its input to upper-case (ABC). +** Function lower() converts to lower-case (abc). +** +** ICU provides two types of case mapping, "general" case mapping and +** "language specific". Refer to ICU documentation for the differences +** between the two. +** +** To utilise "general" case mapping, the upper() or lower() scalar +** functions are invoked with one argument: +** +** upper('ABC') -> 'abc' +** lower('abc') -> 'ABC' +** +** To access ICU "language specific" case mapping, upper() or lower() +** should be invoked with two arguments. The second argument is the name +** of the locale to use. Passing an empty string ("") or SQL NULL value +** as the second argument is the same as invoking the 1 argument version +** of upper() or lower(). +** +** lower('I', 'en_us') -> 'i' +** lower('I', 'tr_tr') -> 'ı' (small dotless i) +** +** http://www.icu-project.org/userguide/posix.html#case_mappings +*/ +static void icuCaseFunc16(sqlite3_context *p, int nArg, sqlite3_value **apArg){ + const UChar *zInput; + UChar *zOutput; + size_t nInput; + size_t nOutput; + + UErrorCode status = U_ZERO_ERROR; + const char *zLocale = 0; + + assert(nArg==1 || nArg==2); + if( nArg==2 ){ + zLocale = (const char *)sqlite3_value_text(apArg[1]); + } + + zInput = sqlite3_value_text16(apArg[0]); + if( !zInput ){ + return; + } + nInput = sqlite3_value_bytes16(apArg[0]); + + nOutput = nInput * 2 + 2; + zOutput = sqlite3_malloc(nOutput); + if( !zOutput ){ + return; + } + + if( sqlite3_user_data(p) ){ + u_strToUpper(zOutput, nOutput/2, zInput, nInput/2, zLocale, &status); + }else{ + u_strToLower(zOutput, nOutput/2, zInput, nInput/2, zLocale, &status); + } + + if( !U_SUCCESS(status) ){ + icuFunctionError(p, "u_strToLower()/u_strToUpper", status); + return; + } + + sqlite3_result_text16(p, zOutput, -1, xFree); +} + +/* +** Collation sequence destructor function. The pCtx argument points to +** a UCollator structure previously allocated using ucol_open(). +*/ +static void icuCollationDel(void *pCtx){ + UCollator *p = (UCollator *)pCtx; + ucol_close(p); +} + +/* +** Collation sequence comparison function. The pCtx argument points to +** a UCollator structure previously allocated using ucol_open(). +*/ +static int icuCollationColl( + void *pCtx, + int nLeft, + const void *zLeft, + int nRight, + const void *zRight +){ + UCollationResult res; + UCollator *p = (UCollator *)pCtx; + res = ucol_strcoll(p, (UChar *)zLeft, nLeft/2, (UChar *)zRight, nRight/2); + switch( res ){ + case UCOL_LESS: return -1; + case UCOL_GREATER: return +1; + case UCOL_EQUAL: return 0; + } + assert(!"Unexpected return value from ucol_strcoll()"); + return 0; +} + +/* +** Implementation of the scalar function icu_load_collation(). +** +** This scalar function is used to add ICU collation based collation +** types to an SQLite database connection. It is intended to be called +** as follows: +** +** SELECT icu_load_collation(, ); +** +** Where is a string containing an ICU locale identifier (i.e. +** "en_AU", "tr_TR" etc.) and is the name of the +** collation sequence to create. +*/ +static void icuLoadCollation( + sqlite3_context *p, + int nArg, + sqlite3_value **apArg +){ + sqlite3 *db = (sqlite3 *)sqlite3_user_data(p); + UErrorCode status = U_ZERO_ERROR; + const char *zLocale; /* Locale identifier - (eg. "jp_JP") */ + const char *zName; /* SQL Collation sequence name (eg. "japanese") */ + UCollator *pUCollator; /* ICU library collation object */ + int rc; /* Return code from sqlite3_create_collation_x() */ + + assert(nArg==2); + (void)nArg; /* Unused parameter */ + zLocale = (const char *)sqlite3_value_text(apArg[0]); + zName = (const char *)sqlite3_value_text(apArg[1]); + + if( !zLocale || !zName ){ + return; + } + + pUCollator = ucol_open(zLocale, &status); + if( !U_SUCCESS(status) ){ + icuFunctionError(p, "ucol_open", status); + return; + } + assert(p); + + rc = sqlite3_create_collation_v2(db, zName, SQLITE_UTF16, (void *)pUCollator, + icuCollationColl, icuCollationDel + ); + if( rc!=SQLITE_OK ){ + ucol_close(pUCollator); + sqlite3_result_error(p, "Error registering collation function", -1); + } +} + +/* +** Register the ICU extension functions with database db. +*/ +int sqlite3IcuInit(sqlite3 *db){ + struct IcuScalar { + const char *zName; /* Function name */ + int nArg; /* Number of arguments */ + int enc; /* Optimal text encoding */ + void *pContext; /* sqlite3_user_data() context */ + void (*xFunc)(sqlite3_context*,int,sqlite3_value**); + } scalars[] = { + {"regexp", 2, SQLITE_UTF8, 0, icuRegexpFunc}, + + {"lower", 1, SQLITE_UTF16, 0, icuCaseFunc16}, + {"lower", 2, SQLITE_UTF16, 0, icuCaseFunc16}, + {"upper", 1, SQLITE_UTF16, (void*)1, icuCaseFunc16}, + {"upper", 2, SQLITE_UTF16, (void*)1, icuCaseFunc16}, + + {"lower", 1, SQLITE_UTF8, 0, icuCaseFunc16}, + {"lower", 2, SQLITE_UTF8, 0, icuCaseFunc16}, + {"upper", 1, SQLITE_UTF8, (void*)1, icuCaseFunc16}, + {"upper", 2, SQLITE_UTF8, (void*)1, icuCaseFunc16}, + + {"like", 2, SQLITE_UTF8, 0, icuLikeFunc}, + {"like", 3, SQLITE_UTF8, 0, icuLikeFunc}, + + {"icu_load_collation", 2, SQLITE_UTF8, (void*)db, icuLoadCollation}, + }; + + int rc = SQLITE_OK; + int i; + + for(i=0; rc==SQLITE_OK && i<(int)(sizeof(scalars)/sizeof(scalars[0])); i++){ + struct IcuScalar *p = &scalars[i]; + rc = sqlite3_create_function_v2( + db, p->zName, p->nArg, p->enc, p->pContext, p->xFunc, 0, 0, 0 + ); + } + + return rc; +} + +#if !SQLITE_CORE +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_icu_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + SQLITE_EXTENSION_INIT2(pApi) + return sqlite3IcuInit(db); +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + SQLITE_EXTENSION_INIT2(pApi) + return sqlite3IcuInit(db); +} +#endif +#endif + +#endif --- origsrc/sqlite-autoconf-3100000/ieee754.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/ieee754.c 2016-01-06 16:30:42.504121500 +0100 @@ -0,0 +1,149 @@ +/* +** 2013-04-17 +** +** 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 SQLite extension implements functions for the exact display +** and input of IEEE754 Binary64 floating-point numbers. +** +** ieee754(X) +** ieee754(Y,Z) +** +** In the first form, the value X should be a floating-point number. +** The function will return a string of the form 'ieee754(Y,Z)' where +** Y and Z are integers such that X==Y*pow(2,Z). +** +** In the second form, Y and Z are integers which are the mantissa and +** base-2 exponent of a new floating point number. The function returns +** a floating-point value equal to Y*pow(2,Z). +** +** Examples: +** +** ieee754(2.0) -> 'ieee754(2,0)' +** ieee754(45.25) -> 'ieee754(181,-2)' +** ieee754(2, 0) -> 2.0 +** ieee754(181, -2) -> 45.25 +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include +#include + +/* +** Implementation of the ieee754() function +*/ +static void ieee754func( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + if( argc==1 ){ + sqlite3_int64 m, a; + double r; + int e; + int isNeg; + char zResult[100]; + assert( sizeof(m)==sizeof(r) ); + if( sqlite3_value_type(argv[0])!=SQLITE_FLOAT ) return; + r = sqlite3_value_double(argv[0]); + if( r<0.0 ){ + isNeg = 1; + r = -r; + }else{ + isNeg = 0; + } + memcpy(&a,&r,sizeof(a)); + if( a==0 ){ + e = 0; + m = 0; + }else{ + e = a>>52; + m = a & ((((sqlite3_int64)1)<<52)-1); + m |= ((sqlite3_int64)1)<<52; + while( e<1075 && m>0 && (m&1)==0 ){ + m >>= 1; + e++; + } + if( isNeg ) m = -m; + } + sqlite3_snprintf(sizeof(zResult), zResult, "ieee754(%lld,%d)", + m, e-1075); + sqlite3_result_text(context, zResult, -1, SQLITE_TRANSIENT); + }else if( argc==2 ){ + sqlite3_int64 m, e, a; + double r; + int isNeg = 0; + m = sqlite3_value_int64(argv[0]); + e = sqlite3_value_int64(argv[1]); + if( m<0 ){ + isNeg = 1; + m = -m; + if( m<0 ) return; + }else if( m==0 && e>1000 && e<1000 ){ + sqlite3_result_double(context, 0.0); + return; + } + while( (m>>32)&0xffe00000 ){ + m >>= 1; + e++; + } + while( m!=0 && ((m>>32)&0xfff00000)==0 ){ + m <<= 1; + e--; + } + e += 1075; + if( e<0 ) e = m = 0; + if( e>0x7ff ) e = 0x7ff; + a = m & ((((sqlite3_int64)1)<<52)-1); + a |= e<<52; + if( isNeg ) a |= ((sqlite3_uint64)1)<<63; + memcpy(&r, &a, sizeof(r)); + sqlite3_result_double(context, r); + } +} + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_ieee_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "ieee754", 1, SQLITE_UTF8, 0, + ieee754func, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "ieee754", 2, SQLITE_UTF8, 0, + ieee754func, 0, 0); + } + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "ieee754", 1, SQLITE_UTF8, 0, + ieee754func, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "ieee754", 2, SQLITE_UTF8, 0, + ieee754func, 0, 0); + } + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/nextchar.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/nextchar.c 2016-01-06 16:30:42.508121700 +0100 @@ -0,0 +1,333 @@ +/* +** 2013-02-28 +** +** 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 file contains code to implement the next_char(A,T,F,W,C) SQL function. +** +** The next_char(A,T,F,W,C) function finds all valid "next" characters for +** string A given the vocabulary in T.F. If the W value exists and is a +** non-empty string, then it is an SQL expression that limits the entries +** in T.F that will be considered. If C exists and is a non-empty string, +** then it is the name of the collating sequence to use for comparison. If +** +** Only the first three arguments are required. If the C parameter is +** omitted or is NULL or is an empty string, then the default collating +** sequence of T.F is used for comparision. If the W parameter is omitted +** or is NULL or is an empty string, then no filtering of the output is +** done. +** +** The T.F column should be indexed using collation C or else this routine +** will be quite slow. +** +** For example, suppose an application has a dictionary like this: +** +** CREATE TABLE dictionary(word TEXT UNIQUE); +** +** Further suppose that for user keypad entry, it is desired to disable +** (gray out) keys that are not valid as the next character. If the +** the user has previously entered (say) 'cha' then to find all allowed +** next characters (and thereby determine when keys should not be grayed +** out) run the following query: +** +** SELECT next_char('cha','dictionary','word'); +** +** IMPLEMENTATION NOTES: +** +** The next_char function is implemented using recursive SQL that makes +** use of the table name and column name as part of a query. If either +** the table name or column name are keywords or contain special characters, +** then they should be escaped. For example: +** +** SELECT next_char('cha','[dictionary]','[word]'); +** +** This also means that the table name can be a subquery: +** +** SELECT next_char('cha','(SELECT word AS w FROM dictionary)','w'); +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include + +/* +** A structure to hold context of the next_char() computation across +** nested function calls. +*/ +typedef struct nextCharContext nextCharContext; +struct nextCharContext { + sqlite3 *db; /* Database connection */ + sqlite3_stmt *pStmt; /* Prepared statement used to query */ + const unsigned char *zPrefix; /* Prefix to scan */ + int nPrefix; /* Size of zPrefix in bytes */ + int nAlloc; /* Space allocated to aResult */ + int nUsed; /* Space used in aResult */ + unsigned int *aResult; /* Array of next characters */ + int mallocFailed; /* True if malloc fails */ + int otherError; /* True for any other failure */ +}; + +/* +** Append a result character if the character is not already in the +** result. +*/ +static void nextCharAppend(nextCharContext *p, unsigned c){ + int i; + for(i=0; inUsed; i++){ + if( p->aResult[i]==c ) return; + } + if( p->nUsed+1 > p->nAlloc ){ + unsigned int *aNew; + int n = p->nAlloc*2 + 30; + aNew = sqlite3_realloc(p->aResult, n*sizeof(unsigned int)); + if( aNew==0 ){ + p->mallocFailed = 1; + return; + }else{ + p->aResult = aNew; + p->nAlloc = n; + } + } + p->aResult[p->nUsed++] = c; +} + +/* +** Write a character into z[] as UTF8. Return the number of bytes needed +** to hold the character +*/ +static int writeUtf8(unsigned char *z, unsigned c){ + if( c<0x00080 ){ + z[0] = (unsigned char)(c&0xff); + return 1; + } + if( c<0x00800 ){ + z[0] = 0xC0 + (unsigned char)((c>>6)&0x1F); + z[1] = 0x80 + (unsigned char)(c & 0x3F); + return 2; + } + if( c<0x10000 ){ + z[0] = 0xE0 + (unsigned char)((c>>12)&0x0F); + z[1] = 0x80 + (unsigned char)((c>>6) & 0x3F); + z[2] = 0x80 + (unsigned char)(c & 0x3F); + return 3; + } + z[0] = 0xF0 + (unsigned char)((c>>18) & 0x07); + z[1] = 0x80 + (unsigned char)((c>>12) & 0x3F); + z[2] = 0x80 + (unsigned char)((c>>6) & 0x3F); + z[3] = 0x80 + (unsigned char)(c & 0x3F); + return 4; +} + +/* +** Read a UTF8 character out of z[] and write it into *pOut. Return +** the number of bytes in z[] that were used to construct the character. +*/ +static int readUtf8(const unsigned char *z, unsigned *pOut){ + static const unsigned char validBits[] = { + 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, + }; + unsigned c = z[0]; + if( c<0xc0 ){ + *pOut = c; + return 1; + }else{ + int n = 1; + c = validBits[c-0xc0]; + while( (z[n] & 0xc0)==0x80 ){ + c = (c<<6) + (0x3f & z[n++]); + } + if( c<0x80 || (c&0xFFFFF800)==0xD800 || (c&0xFFFFFFFE)==0xFFFE ){ + c = 0xFFFD; + } + *pOut = c; + return n; + } +} + +/* +** The nextCharContext structure has been set up. Add all "next" characters +** to the result set. +*/ +static void findNextChars(nextCharContext *p){ + unsigned cPrev = 0; + unsigned char zPrev[8]; + int n, rc; + + for(;;){ + sqlite3_bind_text(p->pStmt, 1, (char*)p->zPrefix, p->nPrefix, + SQLITE_STATIC); + n = writeUtf8(zPrev, cPrev+1); + sqlite3_bind_text(p->pStmt, 2, (char*)zPrev, n, SQLITE_STATIC); + rc = sqlite3_step(p->pStmt); + if( rc==SQLITE_DONE ){ + sqlite3_reset(p->pStmt); + return; + }else if( rc!=SQLITE_ROW ){ + p->otherError = rc; + return; + }else{ + const unsigned char *zOut = sqlite3_column_text(p->pStmt, 0); + unsigned cNext; + n = readUtf8(zOut+p->nPrefix, &cNext); + sqlite3_reset(p->pStmt); + nextCharAppend(p, cNext); + cPrev = cNext; + if( p->mallocFailed ) return; + } + } +} + + +/* +** next_character(A,T,F,W) +** +** Return a string composted of all next possible characters after +** A for elements of T.F. If W is supplied, then it is an SQL expression +** that limits the elements in T.F that are considered. +*/ +static void nextCharFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + nextCharContext c; + const unsigned char *zTable = sqlite3_value_text(argv[1]); + const unsigned char *zField = sqlite3_value_text(argv[2]); + const unsigned char *zWhere; + const unsigned char *zCollName; + char *zWhereClause = 0; + char *zColl = 0; + char *zSql; + int rc; + + memset(&c, 0, sizeof(c)); + c.db = sqlite3_context_db_handle(context); + c.zPrefix = sqlite3_value_text(argv[0]); + c.nPrefix = sqlite3_value_bytes(argv[0]); + if( zTable==0 || zField==0 || c.zPrefix==0 ) return; + if( argc>=4 + && (zWhere = sqlite3_value_text(argv[3]))!=0 + && zWhere[0]!=0 + ){ + zWhereClause = sqlite3_mprintf("AND (%s)", zWhere); + if( zWhereClause==0 ){ + sqlite3_result_error_nomem(context); + return; + } + }else{ + zWhereClause = ""; + } + if( argc>=5 + && (zCollName = sqlite3_value_text(argv[4]))!=0 + && zCollName[0]!=0 + ){ + zColl = sqlite3_mprintf("collate \"%w\"", zCollName); + if( zColl==0 ){ + sqlite3_result_error_nomem(context); + if( zWhereClause[0] ) sqlite3_free(zWhereClause); + return; + } + }else{ + zColl = ""; + } + zSql = sqlite3_mprintf( + "SELECT %s FROM %s" + " WHERE %s>=(?1 || ?2) %s" + " AND %s<=(?1 || char(1114111)) %s" /* 1114111 == 0x10ffff */ + " %s" + " ORDER BY 1 %s ASC LIMIT 1", + zField, zTable, zField, zColl, zField, zColl, zWhereClause, zColl + ); + if( zWhereClause[0] ) sqlite3_free(zWhereClause); + if( zColl[0] ) sqlite3_free(zColl); + if( zSql==0 ){ + sqlite3_result_error_nomem(context); + return; + } + + rc = sqlite3_prepare_v2(c.db, zSql, -1, &c.pStmt, 0); + sqlite3_free(zSql); + if( rc ){ + sqlite3_result_error(context, sqlite3_errmsg(c.db), -1); + return; + } + findNextChars(&c); + if( c.mallocFailed ){ + sqlite3_result_error_nomem(context); + }else{ + unsigned char *pRes; + pRes = sqlite3_malloc( c.nUsed*4 + 1 ); + if( pRes==0 ){ + sqlite3_result_error_nomem(context); + }else{ + int i; + int n = 0; + for(i=0; i +#include +#include + +/* The following object is the session context for a single percentile() +** function. We have to remember all input Y values until the very end. +** Those values are accumulated in the Percentile.a[] array. +*/ +typedef struct Percentile Percentile; +struct Percentile { + unsigned nAlloc; /* Number of slots allocated for a[] */ + unsigned nUsed; /* Number of slots actually used in a[] */ + double rPct; /* 1.0 more than the value for P */ + double *a; /* Array of Y values */ +}; + +/* +** Return TRUE if the input floating-point number is an infinity. +*/ +static int isInfinity(double r){ + sqlite3_uint64 u; + assert( sizeof(u)==sizeof(r) ); + memcpy(&u, &r, sizeof(u)); + return ((u>>52)&0x7ff)==0x7ff; +} + +/* +** Return TRUE if two doubles differ by 0.001 or less +*/ +static int sameValue(double a, double b){ + a -= b; + return a>=-0.001 && a<=0.001; +} + +/* +** The "step" function for percentile(Y,P) is called once for each +** input row. +*/ +static void percentStep(sqlite3_context *pCtx, int argc, sqlite3_value **argv){ + Percentile *p; + double rPct; + int eType; + double y; + assert( argc==2 ); + + /* Requirement 3: P must be a number between 0 and 100 */ + eType = sqlite3_value_numeric_type(argv[1]); + rPct = sqlite3_value_double(argv[1]); + if( (eType!=SQLITE_INTEGER && eType!=SQLITE_FLOAT) || + ((rPct = sqlite3_value_double(argv[1]))<0.0 || rPct>100.0) ){ + sqlite3_result_error(pCtx, "2nd argument to percentile() is not " + "a number between 0.0 and 100.0", -1); + return; + } + + /* Allocate the session context. */ + p = (Percentile*)sqlite3_aggregate_context(pCtx, sizeof(*p)); + if( p==0 ) return; + + /* Remember the P value. Throw an error if the P value is different + ** from any prior row, per Requirement (2). */ + if( p->rPct==0.0 ){ + p->rPct = rPct+1.0; + }else if( !sameValue(p->rPct,rPct+1.0) ){ + sqlite3_result_error(pCtx, "2nd argument to percentile() is not the " + "same for all input rows", -1); + return; + } + + /* Ignore rows for which Y is NULL */ + eType = sqlite3_value_type(argv[0]); + if( eType==SQLITE_NULL ) return; + + /* If not NULL, then Y must be numeric. Otherwise throw an error. + ** Requirement 4 */ + if( eType!=SQLITE_INTEGER && eType!=SQLITE_FLOAT ){ + sqlite3_result_error(pCtx, "1st argument to percentile() is not " + "numeric", -1); + return; + } + + /* Throw an error if the Y value is infinity or NaN */ + y = sqlite3_value_double(argv[0]); + if( isInfinity(y) ){ + sqlite3_result_error(pCtx, "Inf input to percentile()", -1); + return; + } + + /* Allocate and store the Y */ + if( p->nUsed>=p->nAlloc ){ + unsigned n = p->nAlloc*2 + 250; + double *a = sqlite3_realloc(p->a, sizeof(double)*n); + if( a==0 ){ + sqlite3_free(p->a); + memset(p, 0, sizeof(*p)); + sqlite3_result_error_nomem(pCtx); + return; + } + p->nAlloc = n; + p->a = a; + } + p->a[p->nUsed++] = y; +} + +/* +** Compare to doubles for sorting using qsort() +*/ +static int doubleCmp(const void *pA, const void *pB){ + double a = *(double*)pA; + double b = *(double*)pB; + if( a==b ) return 0; + if( aa==0 ) return; + if( p->nUsed ){ + qsort(p->a, p->nUsed, sizeof(double), doubleCmp); + ix = (p->rPct-1.0)*(p->nUsed-1)*0.01; + i1 = (unsigned)ix; + i2 = ix==(double)i1 || i1==p->nUsed-1 ? i1 : i1+1; + v1 = p->a[i1]; + v2 = p->a[i2]; + vx = v1 + (v2-v1)*(ix-i1); + sqlite3_result_double(pCtx, vx); + } + sqlite3_free(p->a); + memset(p, 0, sizeof(*p)); +} + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_percentile_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "percentile", 2, SQLITE_UTF8, 0, + 0, percentStep, percentFinal); + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "percentile", 2, SQLITE_UTF8, 0, + 0, percentStep, percentFinal); + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/rbu.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/rbu.c 2016-01-06 16:48:54.403574500 +0100 @@ -0,0 +1,125 @@ +/* +** 2014 August 30 +** +** 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 file contains a command-line application that uses the RBU +** extension. See the usage() function below for an explanation. +*/ + +#include "sqlite3rbu.h" +#include +#include +#include + +/* +** Print a usage message and exit. +*/ +void usage(const char *zArgv0){ + fprintf(stderr, +"Usage: %s [-step NSTEP] TARGET-DB RBU-DB\n" +"\n" +" Argument RBU-DB must be an RBU database containing an update suitable for\n" +" target database TARGET-DB. If NSTEP is set to less than or equal to zero\n" +" (the default value), this program attempts to apply the entire update to\n" +" the target database.\n" +"\n" +" If NSTEP is greater than zero, then a maximum of NSTEP calls are made\n" +" to sqlite3rbu_step(). If the RBU update has not been completely applied\n" +" after the NSTEP'th call is made, the state is saved in the database RBU-DB\n" +" and the program exits. Subsequent invocations of this (or any other RBU)\n" +" application will use this state to resume applying the RBU update to the\n" +" target db.\n" +"\n" +, zArgv0); + exit(1); +} + +void report_default_vfs(){ + sqlite3_vfs *pVfs = sqlite3_vfs_find(0); + fprintf(stdout, "default vfs is \"%s\"\n", pVfs->zName); +} + +void report_rbu_vfs(sqlite3rbu *pRbu){ + sqlite3 *db = sqlite3rbu_db(pRbu, 0); + if( db ){ + char *zName = 0; + sqlite3_file_control(db, "main", SQLITE_FCNTL_VFSNAME, &zName); + if( zName ){ + fprintf(stdout, "using vfs \"%s\"\n", zName); + }else{ + fprintf(stdout, "vfs name not available\n"); + } + sqlite3_free(zName); + } +} + +int main(int argc, char **argv){ + int i; + const char *zTarget; /* Target database to apply RBU to */ + const char *zRbu; /* Database containing RBU */ + char zBuf[200]; /* Buffer for printf() */ + char *zErrmsg; /* Error message, if any */ + sqlite3rbu *pRbu; /* RBU handle */ + int nStep = 0; /* Maximum number of step() calls */ + int rc; + sqlite3_int64 nProgress = 0; + + /* Process command line arguments. Following this block local variables + ** zTarget, zRbu and nStep are all set. */ + if( argc==5 ){ + int nArg1 = strlen(argv[1]); + if( nArg1>5 || nArg1<2 || memcmp("-step", argv[1], nArg1) ) usage(argv[0]); + nStep = atoi(argv[2]); + }else if( argc!=3 ){ + usage(argv[0]); + } + zTarget = argv[argc-2]; + zRbu = argv[argc-1]; + + report_default_vfs(); + + /* Open an RBU handle. If nStep is less than or equal to zero, call + ** sqlite3rbu_step() until either the RBU has been completely applied + ** or an error occurs. Or, if nStep is greater than zero, call + ** sqlite3rbu_step() a maximum of nStep times. */ + pRbu = sqlite3rbu_open(zTarget, zRbu, 0); + report_rbu_vfs(pRbu); + for(i=0; (nStep<=0 || i +#include +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 + +/* +** The following #defines change the names of some functions implemented in +** this file to prevent name collisions with C-library functions of the +** same name. +*/ +#define re_match sqlite3re_match +#define re_compile sqlite3re_compile +#define re_free sqlite3re_free + +/* The end-of-input character */ +#define RE_EOF 0 /* End of input */ + +/* The NFA is implemented as sequence of opcodes taken from the following +** set. Each opcode has a single integer argument. +*/ +#define RE_OP_MATCH 1 /* Match the one character in the argument */ +#define RE_OP_ANY 2 /* Match any one character. (Implements ".") */ +#define RE_OP_ANYSTAR 3 /* Special optimized version of .* */ +#define RE_OP_FORK 4 /* Continue to both next and opcode at iArg */ +#define RE_OP_GOTO 5 /* Jump to opcode at iArg */ +#define RE_OP_ACCEPT 6 /* Halt and indicate a successful match */ +#define RE_OP_CC_INC 7 /* Beginning of a [...] character class */ +#define RE_OP_CC_EXC 8 /* Beginning of a [^...] character class */ +#define RE_OP_CC_VALUE 9 /* Single value in a character class */ +#define RE_OP_CC_RANGE 10 /* Range of values in a character class */ +#define RE_OP_WORD 11 /* Perl word character [A-Za-z0-9_] */ +#define RE_OP_NOTWORD 12 /* Not a perl word character */ +#define RE_OP_DIGIT 13 /* digit: [0-9] */ +#define RE_OP_NOTDIGIT 14 /* Not a digit */ +#define RE_OP_SPACE 15 /* space: [ \t\n\r\v\f] */ +#define RE_OP_NOTSPACE 16 /* Not a digit */ +#define RE_OP_BOUNDARY 17 /* Boundary between word and non-word */ + +/* Each opcode is a "state" in the NFA */ +typedef unsigned short ReStateNumber; + +/* Because this is an NFA and not a DFA, multiple states can be active at +** once. An instance of the following object records all active states in +** the NFA. The implementation is optimized for the common case where the +** number of actives states is small. +*/ +typedef struct ReStateSet { + unsigned nState; /* Number of current states */ + ReStateNumber *aState; /* Current states */ +} ReStateSet; + +/* An input string read one character at a time. +*/ +typedef struct ReInput ReInput; +struct ReInput { + const unsigned char *z; /* All text */ + int i; /* Next byte to read */ + int mx; /* EOF when i>=mx */ +}; + +/* A compiled NFA (or an NFA that is in the process of being compiled) is +** an instance of the following object. +*/ +typedef struct ReCompiled ReCompiled; +struct ReCompiled { + ReInput sIn; /* Regular expression text */ + const char *zErr; /* Error message to return */ + char *aOp; /* Operators for the virtual machine */ + int *aArg; /* Arguments to each operator */ + unsigned (*xNextChar)(ReInput*); /* Next character function */ + unsigned char zInit[12]; /* Initial text to match */ + int nInit; /* Number of characters in zInit */ + unsigned nState; /* Number of entries in aOp[] and aArg[] */ + unsigned nAlloc; /* Slots allocated for aOp[] and aArg[] */ +}; + +/* Add a state to the given state set if it is not already there */ +static void re_add_state(ReStateSet *pSet, int newState){ + unsigned i; + for(i=0; inState; i++) if( pSet->aState[i]==newState ) return; + pSet->aState[pSet->nState++] = newState; +} + +/* Extract the next unicode character from *pzIn and return it. Advance +** *pzIn to the first byte past the end of the character returned. To +** be clear: this routine converts utf8 to unicode. This routine is +** optimized for the common case where the next character is a single byte. +*/ +static unsigned re_next_char(ReInput *p){ + unsigned c; + if( p->i>=p->mx ) return 0; + c = p->z[p->i++]; + if( c>=0x80 ){ + if( (c&0xe0)==0xc0 && p->imx && (p->z[p->i]&0xc0)==0x80 ){ + c = (c&0x1f)<<6 | (p->z[p->i++]&0x3f); + if( c<0x80 ) c = 0xfffd; + }else if( (c&0xf0)==0xe0 && p->i+1mx && (p->z[p->i]&0xc0)==0x80 + && (p->z[p->i+1]&0xc0)==0x80 ){ + c = (c&0x0f)<<12 | ((p->z[p->i]&0x3f)<<6) | (p->z[p->i+1]&0x3f); + p->i += 2; + if( c<=0x3ff || (c>=0xd800 && c<=0xdfff) ) c = 0xfffd; + }else if( (c&0xf8)==0xf0 && p->i+3mx && (p->z[p->i]&0xc0)==0x80 + && (p->z[p->i+1]&0xc0)==0x80 && (p->z[p->i+2]&0xc0)==0x80 ){ + c = (c&0x07)<<18 | ((p->z[p->i]&0x3f)<<12) | ((p->z[p->i+1]&0x3f)<<6) + | (p->z[p->i+2]&0x3f); + p->i += 3; + if( c<=0xffff || c>0x10ffff ) c = 0xfffd; + }else{ + c = 0xfffd; + } + } + return c; +} +static unsigned re_next_char_nocase(ReInput *p){ + unsigned c = re_next_char(p); + if( c>='A' && c<='Z' ) c += 'a' - 'A'; + return c; +} + +/* Return true if c is a perl "word" character: [A-Za-z0-9_] */ +static int re_word_char(int c){ + return (c>='0' && c<='9') || (c>='a' && c<='z') + || (c>='A' && c<='Z') || c=='_'; +} + +/* Return true if c is a "digit" character: [0-9] */ +static int re_digit_char(int c){ + return (c>='0' && c<='9'); +} + +/* Return true if c is a perl "space" character: [ \t\r\n\v\f] */ +static int re_space_char(int c){ + return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f'; +} + +/* Run a compiled regular expression on the zero-terminated input +** string zIn[]. Return true on a match and false if there is no match. +*/ +int re_match(ReCompiled *pRe, const unsigned char *zIn, int nIn){ + ReStateSet aStateSet[2], *pThis, *pNext; + ReStateNumber aSpace[100]; + ReStateNumber *pToFree; + unsigned int i = 0; + unsigned int iSwap = 0; + int c = RE_EOF+1; + int cPrev = 0; + int rc = 0; + ReInput in; + + in.z = zIn; + in.i = 0; + in.mx = nIn>=0 ? nIn : (int)strlen((char const*)zIn); + + /* Look for the initial prefix match, if there is one. */ + if( pRe->nInit ){ + unsigned char x = pRe->zInit[0]; + while( in.i+pRe->nInit<=in.mx + && (zIn[in.i]!=x || + strncmp((const char*)zIn+in.i, (const char*)pRe->zInit, pRe->nInit)!=0) + ){ + in.i++; + } + if( in.i+pRe->nInit>in.mx ) return 0; + } + + if( pRe->nState<=(sizeof(aSpace)/(sizeof(aSpace[0])*2)) ){ + pToFree = 0; + aStateSet[0].aState = aSpace; + }else{ + pToFree = sqlite3_malloc( sizeof(ReStateNumber)*2*pRe->nState ); + if( pToFree==0 ) return -1; + aStateSet[0].aState = pToFree; + } + aStateSet[1].aState = &aStateSet[0].aState[pRe->nState]; + pNext = &aStateSet[1]; + pNext->nState = 0; + re_add_state(pNext, 0); + while( c!=RE_EOF && pNext->nState>0 ){ + cPrev = c; + c = pRe->xNextChar(&in); + pThis = pNext; + pNext = &aStateSet[iSwap]; + iSwap = 1 - iSwap; + pNext->nState = 0; + for(i=0; inState; i++){ + int x = pThis->aState[i]; + switch( pRe->aOp[x] ){ + case RE_OP_MATCH: { + if( pRe->aArg[x]==c ) re_add_state(pNext, x+1); + break; + } + case RE_OP_ANY: { + re_add_state(pNext, x+1); + break; + } + case RE_OP_WORD: { + if( re_word_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_NOTWORD: { + if( !re_word_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_DIGIT: { + if( re_digit_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_NOTDIGIT: { + if( !re_digit_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_SPACE: { + if( re_space_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_NOTSPACE: { + if( !re_space_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_BOUNDARY: { + if( re_word_char(c)!=re_word_char(cPrev) ) re_add_state(pThis, x+1); + break; + } + case RE_OP_ANYSTAR: { + re_add_state(pNext, x); + re_add_state(pThis, x+1); + break; + } + case RE_OP_FORK: { + re_add_state(pThis, x+pRe->aArg[x]); + re_add_state(pThis, x+1); + break; + } + case RE_OP_GOTO: { + re_add_state(pThis, x+pRe->aArg[x]); + break; + } + case RE_OP_ACCEPT: { + rc = 1; + goto re_match_end; + } + case RE_OP_CC_INC: + case RE_OP_CC_EXC: { + int j = 1; + int n = pRe->aArg[x]; + int hit = 0; + for(j=1; j>0 && jaOp[x+j]==RE_OP_CC_VALUE ){ + if( pRe->aArg[x+j]==c ){ + hit = 1; + j = -1; + } + }else{ + if( pRe->aArg[x+j]<=c && pRe->aArg[x+j+1]>=c ){ + hit = 1; + j = -1; + }else{ + j++; + } + } + } + if( pRe->aOp[x]==RE_OP_CC_EXC ) hit = !hit; + if( hit ) re_add_state(pNext, x+n); + break; + } + } + } + } + for(i=0; inState; i++){ + if( pRe->aOp[pNext->aState[i]]==RE_OP_ACCEPT ){ rc = 1; break; } + } +re_match_end: + sqlite3_free(pToFree); + return rc; +} + +/* Resize the opcode and argument arrays for an RE under construction. +*/ +static int re_resize(ReCompiled *p, int N){ + char *aOp; + int *aArg; + aOp = sqlite3_realloc(p->aOp, N*sizeof(p->aOp[0])); + if( aOp==0 ) return 1; + p->aOp = aOp; + aArg = sqlite3_realloc(p->aArg, N*sizeof(p->aArg[0])); + if( aArg==0 ) return 1; + p->aArg = aArg; + p->nAlloc = N; + return 0; +} + +/* Insert a new opcode and argument into an RE under construction. The +** insertion point is just prior to existing opcode iBefore. +*/ +static int re_insert(ReCompiled *p, int iBefore, int op, int arg){ + int i; + if( p->nAlloc<=p->nState && re_resize(p, p->nAlloc*2) ) return 0; + for(i=p->nState; i>iBefore; i--){ + p->aOp[i] = p->aOp[i-1]; + p->aArg[i] = p->aArg[i-1]; + } + p->nState++; + p->aOp[iBefore] = op; + p->aArg[iBefore] = arg; + return iBefore; +} + +/* Append a new opcode and argument to the end of the RE under construction. +*/ +static int re_append(ReCompiled *p, int op, int arg){ + return re_insert(p, p->nState, op, arg); +} + +/* Make a copy of N opcodes starting at iStart onto the end of the RE +** under construction. +*/ +static void re_copy(ReCompiled *p, int iStart, int N){ + if( p->nState+N>=p->nAlloc && re_resize(p, p->nAlloc*2+N) ) return; + memcpy(&p->aOp[p->nState], &p->aOp[iStart], N*sizeof(p->aOp[0])); + memcpy(&p->aArg[p->nState], &p->aArg[iStart], N*sizeof(p->aArg[0])); + p->nState += N; +} + +/* Return true if c is a hexadecimal digit character: [0-9a-fA-F] +** If c is a hex digit, also set *pV = (*pV)*16 + valueof(c). If +** c is not a hex digit *pV is unchanged. +*/ +static int re_hex(int c, int *pV){ + if( c>='0' && c<='9' ){ + c -= '0'; + }else if( c>='a' && c<='f' ){ + c -= 'a' - 10; + }else if( c>='A' && c<='F' ){ + c -= 'A' - 10; + }else{ + return 0; + } + *pV = (*pV)*16 + (c & 0xff); + return 1; +} + +/* A backslash character has been seen, read the next character and +** return its interpretation. +*/ +static unsigned re_esc_char(ReCompiled *p){ + static const char zEsc[] = "afnrtv\\()*.+?[$^{|}]"; + static const char zTrans[] = "\a\f\n\r\t\v"; + int i, v = 0; + char c; + if( p->sIn.i>=p->sIn.mx ) return 0; + c = p->sIn.z[p->sIn.i]; + if( c=='u' && p->sIn.i+4sIn.mx ){ + const unsigned char *zIn = p->sIn.z + p->sIn.i; + if( re_hex(zIn[1],&v) + && re_hex(zIn[2],&v) + && re_hex(zIn[3],&v) + && re_hex(zIn[4],&v) + ){ + p->sIn.i += 5; + return v; + } + } + if( c=='x' && p->sIn.i+2sIn.mx ){ + const unsigned char *zIn = p->sIn.z + p->sIn.i; + if( re_hex(zIn[1],&v) + && re_hex(zIn[2],&v) + ){ + p->sIn.i += 3; + return v; + } + } + for(i=0; zEsc[i] && zEsc[i]!=c; i++){} + if( zEsc[i] ){ + if( i<6 ) c = zTrans[i]; + p->sIn.i++; + }else{ + p->zErr = "unknown \\ escape"; + } + return c; +} + +/* Forward declaration */ +static const char *re_subcompile_string(ReCompiled*); + +/* Peek at the next byte of input */ +static unsigned char rePeek(ReCompiled *p){ + return p->sIn.isIn.mx ? p->sIn.z[p->sIn.i] : 0; +} + +/* Compile RE text into a sequence of opcodes. Continue up to the +** first unmatched ")" character, then return. If an error is found, +** return a pointer to the error message string. +*/ +static const char *re_subcompile_re(ReCompiled *p){ + const char *zErr; + int iStart, iEnd, iGoto; + iStart = p->nState; + zErr = re_subcompile_string(p); + if( zErr ) return zErr; + while( rePeek(p)=='|' ){ + iEnd = p->nState; + re_insert(p, iStart, RE_OP_FORK, iEnd + 2 - iStart); + iGoto = re_append(p, RE_OP_GOTO, 0); + p->sIn.i++; + zErr = re_subcompile_string(p); + if( zErr ) return zErr; + p->aArg[iGoto] = p->nState - iGoto; + } + return 0; +} + +/* Compile an element of regular expression text (anything that can be +** an operand to the "|" operator). Return NULL on success or a pointer +** to the error message if there is a problem. +*/ +static const char *re_subcompile_string(ReCompiled *p){ + int iPrev = -1; + int iStart; + unsigned c; + const char *zErr; + while( (c = p->xNextChar(&p->sIn))!=0 ){ + iStart = p->nState; + switch( c ){ + case '|': + case '$': + case ')': { + p->sIn.i--; + return 0; + } + case '(': { + zErr = re_subcompile_re(p); + if( zErr ) return zErr; + if( rePeek(p)!=')' ) return "unmatched '('"; + p->sIn.i++; + break; + } + case '.': { + if( rePeek(p)=='*' ){ + re_append(p, RE_OP_ANYSTAR, 0); + p->sIn.i++; + }else{ + re_append(p, RE_OP_ANY, 0); + } + break; + } + case '*': { + if( iPrev<0 ) return "'*' without operand"; + re_insert(p, iPrev, RE_OP_GOTO, p->nState - iPrev + 1); + re_append(p, RE_OP_FORK, iPrev - p->nState + 1); + break; + } + case '+': { + if( iPrev<0 ) return "'+' without operand"; + re_append(p, RE_OP_FORK, iPrev - p->nState); + break; + } + case '?': { + if( iPrev<0 ) return "'?' without operand"; + re_insert(p, iPrev, RE_OP_FORK, p->nState - iPrev+1); + break; + } + case '{': { + int m = 0, n = 0; + int sz, j; + if( iPrev<0 ) return "'{m,n}' without operand"; + while( (c=rePeek(p))>='0' && c<='9' ){ m = m*10 + c - '0'; p->sIn.i++; } + n = m; + if( c==',' ){ + p->sIn.i++; + n = 0; + while( (c=rePeek(p))>='0' && c<='9' ){ n = n*10 + c-'0'; p->sIn.i++; } + } + if( c!='}' ) return "unmatched '{'"; + if( n>0 && nsIn.i++; + sz = p->nState - iPrev; + if( m==0 ){ + if( n==0 ) return "both m and n are zero in '{m,n}'"; + re_insert(p, iPrev, RE_OP_FORK, sz+1); + n--; + }else{ + for(j=1; j0 ){ + re_append(p, RE_OP_FORK, -sz); + } + break; + } + case '[': { + int iFirst = p->nState; + if( rePeek(p)=='^' ){ + re_append(p, RE_OP_CC_EXC, 0); + p->sIn.i++; + }else{ + re_append(p, RE_OP_CC_INC, 0); + } + while( (c = p->xNextChar(&p->sIn))!=0 ){ + if( c=='[' && rePeek(p)==':' ){ + return "POSIX character classes not supported"; + } + if( c=='\\' ) c = re_esc_char(p); + if( rePeek(p)=='-' ){ + re_append(p, RE_OP_CC_RANGE, c); + p->sIn.i++; + c = p->xNextChar(&p->sIn); + if( c=='\\' ) c = re_esc_char(p); + re_append(p, RE_OP_CC_RANGE, c); + }else{ + re_append(p, RE_OP_CC_VALUE, c); + } + if( rePeek(p)==']' ){ p->sIn.i++; break; } + } + if( c==0 ) return "unclosed '['"; + p->aArg[iFirst] = p->nState - iFirst; + break; + } + case '\\': { + int specialOp = 0; + switch( rePeek(p) ){ + case 'b': specialOp = RE_OP_BOUNDARY; break; + case 'd': specialOp = RE_OP_DIGIT; break; + case 'D': specialOp = RE_OP_NOTDIGIT; break; + case 's': specialOp = RE_OP_SPACE; break; + case 'S': specialOp = RE_OP_NOTSPACE; break; + case 'w': specialOp = RE_OP_WORD; break; + case 'W': specialOp = RE_OP_NOTWORD; break; + } + if( specialOp ){ + p->sIn.i++; + re_append(p, specialOp, 0); + }else{ + c = re_esc_char(p); + re_append(p, RE_OP_MATCH, c); + } + break; + } + default: { + re_append(p, RE_OP_MATCH, c); + break; + } + } + iPrev = iStart; + } + return 0; +} + +/* Free and reclaim all the memory used by a previously compiled +** regular expression. Applications should invoke this routine once +** for every call to re_compile() to avoid memory leaks. +*/ +void re_free(ReCompiled *pRe){ + if( pRe ){ + sqlite3_free(pRe->aOp); + sqlite3_free(pRe->aArg); + sqlite3_free(pRe); + } +} + +/* +** Compile a textual regular expression in zIn[] into a compiled regular +** expression suitable for us by re_match() and return a pointer to the +** compiled regular expression in *ppRe. Return NULL on success or an +** error message if something goes wrong. +*/ +const char *re_compile(ReCompiled **ppRe, const char *zIn, int noCase){ + ReCompiled *pRe; + const char *zErr; + int i, j; + + *ppRe = 0; + pRe = sqlite3_malloc( sizeof(*pRe) ); + if( pRe==0 ){ + return "out of memory"; + } + memset(pRe, 0, sizeof(*pRe)); + pRe->xNextChar = noCase ? re_next_char_nocase : re_next_char; + if( re_resize(pRe, 30) ){ + re_free(pRe); + return "out of memory"; + } + if( zIn[0]=='^' ){ + zIn++; + }else{ + re_append(pRe, RE_OP_ANYSTAR, 0); + } + pRe->sIn.z = (unsigned char*)zIn; + pRe->sIn.i = 0; + pRe->sIn.mx = (int)strlen(zIn); + zErr = re_subcompile_re(pRe); + if( zErr ){ + re_free(pRe); + return zErr; + } + if( rePeek(pRe)=='$' && pRe->sIn.i+1>=pRe->sIn.mx ){ + re_append(pRe, RE_OP_MATCH, RE_EOF); + re_append(pRe, RE_OP_ACCEPT, 0); + *ppRe = pRe; + }else if( pRe->sIn.i>=pRe->sIn.mx ){ + re_append(pRe, RE_OP_ACCEPT, 0); + *ppRe = pRe; + }else{ + re_free(pRe); + return "unrecognized character"; + } + + /* The following is a performance optimization. If the regex begins with + ** ".*" (if the input regex lacks an initial "^") and afterwards there are + ** one or more matching characters, enter those matching characters into + ** zInit[]. The re_match() routine can then search ahead in the input + ** string looking for the initial match without having to run the whole + ** regex engine over the string. Do not worry able trying to match + ** unicode characters beyond plane 0 - those are very rare and this is + ** just an optimization. */ + if( pRe->aOp[0]==RE_OP_ANYSTAR ){ + for(j=0, i=1; jzInit)-2 && pRe->aOp[i]==RE_OP_MATCH; i++){ + unsigned x = pRe->aArg[i]; + if( x<=127 ){ + pRe->zInit[j++] = x; + }else if( x<=0xfff ){ + pRe->zInit[j++] = 0xc0 | (x>>6); + pRe->zInit[j++] = 0x80 | (x&0x3f); + }else if( x<=0xffff ){ + pRe->zInit[j++] = 0xd0 | (x>>12); + pRe->zInit[j++] = 0x80 | ((x>>6)&0x3f); + pRe->zInit[j++] = 0x80 | (x&0x3f); + }else{ + break; + } + } + if( j>0 && pRe->zInit[j-1]==0 ) j--; + pRe->nInit = j; + } + return pRe->zErr; +} + +/* +** Implementation of the regexp() SQL function. This function implements +** the build-in REGEXP operator. The first argument to the function is the +** pattern and the second argument is the string. So, the SQL statements: +** +** A REGEXP B +** +** is implemented as regexp(B,A). +*/ +static void re_sql_func( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + ReCompiled *pRe; /* Compiled regular expression */ + const char *zPattern; /* The regular expression */ + const unsigned char *zStr;/* String being searched */ + const char *zErr; /* Compile error message */ + int setAux = 0; /* True to invoke sqlite3_set_auxdata() */ + + pRe = sqlite3_get_auxdata(context, 0); + if( pRe==0 ){ + zPattern = (const char*)sqlite3_value_text(argv[0]); + if( zPattern==0 ) return; + zErr = re_compile(&pRe, zPattern, 0); + if( zErr ){ + re_free(pRe); + sqlite3_result_error(context, zErr, -1); + return; + } + if( pRe==0 ){ + sqlite3_result_error_nomem(context); + return; + } + setAux = 1; + } + zStr = (const unsigned char*)sqlite3_value_text(argv[1]); + if( zStr!=0 ){ + sqlite3_result_int(context, re_match(pRe, zStr, -1)); + } + if( setAux ){ + sqlite3_set_auxdata(context, 0, pRe, (void(*)(void*))re_free); + } +} + +/* +** Invoke this routine to register the regexp() function with the +** SQLite database connection. +*/ +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_regexp_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + rc = sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, 0, + re_sql_func, 0, 0); + return rc; +} +#if !defined(_WIN32) && !defined(SQLITE_TEST) +int sqlite3_extension_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + rc = sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, 0, + re_sql_func, 0, 0); + return rc; +} +#endif --- origsrc/sqlite-autoconf-3100000/rot13.c 1970-01-01 01:00:00.000000000 +0100 +++ src/sqlite-autoconf-3100000/rot13.c 2016-01-06 16:30:42.512121900 +0100 @@ -0,0 +1,131 @@ +/* +** 2013-05-15 +** +** 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 SQLite extension implements a rot13() function and a rot13 +** collating sequence. +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include +#include + +/* +** Perform rot13 encoding on a single ASCII character. +*/ +static unsigned char rot13(unsigned char c){ + if( c>='a' && c<='z' ){ + c += 13; + if( c>'z' ) c -= 26; + }else if( c>='A' && c<='Z' ){ + c += 13; + if( c>'Z' ) c -= 26; + } + return c; +} + +/* +** Implementation of the rot13() function. +** +** Rotate ASCII alphabetic characters by 13 character positions. +** Non-ASCII characters are unchanged. rot13(rot13(X)) should always +** equal X. +*/ +static void rot13func( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn; + int nIn; + unsigned char *zOut; + char *zToFree = 0; + int i; + char zTemp[100]; + assert( argc==1 ); + if( sqlite3_value_type(argv[0])==SQLITE_NULL ) return; + zIn = (const unsigned char*)sqlite3_value_text(argv[0]); + nIn = sqlite3_value_bytes(argv[0]); + if( nIn