diff options
98 files changed, 1621 insertions, 764 deletions
diff --git a/.bzrignore b/.bzrignore index 27f32040bc8..920522c7592 100644 --- a/.bzrignore +++ b/.bzrignore @@ -591,6 +591,7 @@ sql/sql_yacc.cc sql/sql_yacc.h sql/sql_yacc.output sql/sql_yacc.yy.orig +sql/udf_example.so sql_error.cc sql_prepare.cc stamp-h diff --git a/heap/hp_info.c b/heap/hp_info.c index 3122a665fac..e10d140e1f6 100644 --- a/heap/hp_info.c +++ b/heap/hp_info.c @@ -44,8 +44,7 @@ ulong heap_position_old(HP_INFO *info) /* Note that heap_info does NOT return information about the current position anymore; Use heap_position instead */ -int heap_info(reg1 HP_INFO *info,reg2 HEAPINFO *x, - int flag __attribute__((unused))) +int heap_info(reg1 HP_INFO *info,reg2 HEAPINFO *x, int flag ) { DBUG_ENTER("heap_info"); x->records = info->s->records; diff --git a/include/ft_global.h b/include/ft_global.h index 4aa916e91cf..df6860109e4 100644 --- a/include/ft_global.h +++ b/include/ft_global.h @@ -51,15 +51,18 @@ extern const char *ft_precompiled_stopwords[]; extern ulong ft_min_word_len; extern ulong ft_max_word_len; -extern ulong ft_max_word_len_for_sort; +extern ulong ft_query_expansion_limit; extern const char *ft_boolean_syntax; int ft_init_stopwords(void); void ft_free_stopwords(void); -#define FT_NL 0 -#define FT_BOOL 1 -FT_INFO *ft_init_search(uint,void *, uint, byte *, uint, my_bool); +#define FT_NL 0 +#define FT_BOOL 1 +#define FT_SORTED 2 +#define FT_EXPAND 4 /* query expansion */ + +FT_INFO *ft_init_search(uint,void *, uint, byte *, uint, byte *); #ifdef __cplusplus } diff --git a/include/m_ctype.h b/include/m_ctype.h index e344c434f81..a896afe750a 100644 --- a/include/m_ctype.h +++ b/include/m_ctype.h @@ -310,6 +310,7 @@ int my_wildcmp_8bit(CHARSET_INFO *, uint my_numchars_8bit(CHARSET_INFO *, const char *b, const char *e); uint my_charpos_8bit(CHARSET_INFO *, const char *b, const char *e, uint pos); +int my_mbcharlen_8bit(CHARSET_INFO *, uint c); /* Functions for multibyte charsets */ diff --git a/include/my_bitmap.h b/include/my_bitmap.h index ca0037addfb..5b3da011f54 100644 --- a/include/my_bitmap.h +++ b/include/my_bitmap.h @@ -25,29 +25,36 @@ typedef struct st_bitmap { uchar *bitmap; uint bitmap_size; - my_bool thread_safe; /* set if several threads access the bitmap */ /* mutex will be acquired for the duration of each bitmap operation if - thread_safe flag is set. Otherwise, we optimize by not acquiring the - mutex + thread_safe flag in bitmap_init was set. Otherwise, we optimize by not + acquiring the mutex */ #ifdef THREAD - pthread_mutex_t mutex; + pthread_mutex_t *mutex; #endif } MY_BITMAP; #ifdef __cplusplus extern "C" { #endif - extern my_bool bitmap_init(MY_BITMAP *bitmap, uint bitmap_size, - my_bool thread_safe); - extern void bitmap_free(MY_BITMAP *bitmap); - extern void bitmap_set_bit(MY_BITMAP *bitmap, uint bitmap_bit); - extern uint bitmap_set_next(MY_BITMAP *bitmap); - extern void bitmap_set_all(MY_BITMAP* bitmap); - extern my_bool bitmap_is_set(MY_BITMAP* bitmap, uint bitmap_bit); - extern void bitmap_clear_all(MY_BITMAP* bitmap); - extern void bitmap_clear_bit(MY_BITMAP *bitmap, uint bitmap_bit); +extern my_bool bitmap_cmp(const MY_BITMAP *map1, const MY_BITMAP *map2); +extern my_bool bitmap_init(MY_BITMAP *map, uchar *buf, uint bitmap_size, my_bool thread_safe); +extern my_bool bitmap_is_clear_all(const MY_BITMAP *map); +extern my_bool bitmap_is_prefix(const MY_BITMAP *map, uint prefix_size); +extern my_bool bitmap_is_set(const MY_BITMAP *map, uint bitmap_bit); +extern my_bool bitmap_is_set_all(const MY_BITMAP *map); +extern my_bool bitmap_is_subset(const MY_BITMAP *map1, const MY_BITMAP *map2); +extern uint bitmap_set_next(MY_BITMAP *map); +extern void bitmap_clear_all(MY_BITMAP *map); +extern void bitmap_clear_bit(MY_BITMAP *map, uint bitmap_bit); +extern void bitmap_free(MY_BITMAP *map); +extern void bitmap_intersect(MY_BITMAP *map, const MY_BITMAP *map2); +extern void bitmap_set_all(MY_BITMAP *map); +extern void bitmap_set_bit(MY_BITMAP *map, uint bitmap_bit); +extern void bitmap_set_prefix(MY_BITMAP *map, uint prefix_size); +extern void bitmap_subtract(MY_BITMAP *map, const MY_BITMAP *map2); +extern void bitmap_union(MY_BITMAP *map, const MY_BITMAP *map2); #ifdef __cplusplus } #endif diff --git a/include/mysqld_error.h b/include/mysqld_error.h index 0498196a6c6..60204349de2 100644 --- a/include/mysqld_error.h +++ b/include/mysqld_error.h @@ -230,7 +230,7 @@ #define ER_NO_PERMISSION_TO_CREATE_USER 1211 #define ER_UNION_TABLES_IN_DIFFERENT_DIR 1212 #define ER_LOCK_DEADLOCK 1213 -#define ER_TABLE_CANT_HANDLE_FULLTEXT 1214 +#define ER_TABLE_CANT_HANDLE_FT 1214 #define ER_CANNOT_ADD_FOREIGN 1215 #define ER_NO_REFERENCED_ROW 1216 #define ER_ROW_IS_REFERENCED 1217 @@ -297,4 +297,5 @@ #define ER_UNTIL_COND_IGNORED 1278 #define ER_WRONG_INDEX_NAME 1279 #define ER_WARN_QC_RESIZE 1280 -#define ER_ERROR_MESSAGES 281 +#define ER_BAD_FT_COLUMN 1281 +#define ER_ERROR_MESSAGES 282 diff --git a/myisam/ft_boolean_search.c b/myisam/ft_boolean_search.c index 2fbd6490cfd..95d0149c67d 100644 --- a/myisam/ft_boolean_search.c +++ b/myisam/ft_boolean_search.c @@ -20,8 +20,6 @@ #define FT_CORE #include "ftdefs.h" -#include <queues.h> -#include <assert.h> /* for DBUG_ASSERT() */ /* search with boolean queries */ @@ -144,7 +142,7 @@ static void _ftb_parse_query(FTB *ftb, byte **start, byte *end, param.prev=' '; param.quot=up->quot; - while ((res=ft_get_word(start,end,&w,¶m))) + while ((res=ft_get_word(ftb->charset,start,end,&w,¶m))) { int r=param.plusminus; float weight= (float) (param.pmsign ? nwghts : wghts)[(r>5)?5:((r<-5)?-5:r)]; @@ -168,7 +166,11 @@ static void _ftb_parse_query(FTB *ftb, byte **start, byte *end, ftbw->word[0]=w.len; if (param.yesno > 0) up->ythresh++; queue_insert(& ftb->queue, (byte *)ftbw); +#ifdef TO_BE_REMOVED + /* after removing the following line, + ftb->with_scan handling can be simplified (no longer a bitmap) */ ftb->with_scan|=(param.trunc & FTB_FLAG_TRUNC); +#endif break; case 2: /* left bracket */ ftbe=(FTB_EXPR *)alloc_root(&ftb->mem_root, sizeof(FTB_EXPR)); @@ -336,8 +338,7 @@ static void _ftb_init_index_search(FT_INFO *ftb) FT_INFO * ft_init_boolean_search(MI_INFO *info, uint keynr, byte *query, - uint query_len, - my_bool presort __attribute__((unused))) + uint query_len) { FTB *ftb; FTB_EXPR *ftbe; @@ -350,8 +351,7 @@ FT_INFO * ft_init_boolean_search(MI_INFO *info, uint keynr, byte *query, ftb->info=info; ftb->keynr=keynr; ftb->charset= ((keynr==NO_SUCH_KEY) ? - default_charset_info : - info->s->keyinfo[keynr].seg->charset); + default_charset_info : info->s->keyinfo[keynr].seg->charset); ftb->with_scan=0; ftb->lastpos=HA_POS_ERROR; bzero(& ftb->no_dupes, sizeof(TREE)); @@ -387,25 +387,34 @@ FT_INFO * ft_init_boolean_search(MI_INFO *info, uint keynr, byte *query, } -/* returns 1 if str0 contain str1 */ +/* returns 1 if str0 ~= /\<str1\>/ */ static int _ftb_strstr(const byte *s0, const byte *e0, const byte *s1, const byte *e1, CHARSET_INFO *cs) { - const byte *p; + const byte *p0, *p1; + my_bool s_after, e_before; - while (s0 < e0) + s_after=true_word_char(cs, s1[0]); + e_before=true_word_char(cs, e1[-1]); + p0=s0; + + while (p0 < e0) { - while (s0 < e0 && cs->to_upper[(uint) (uchar) *s0++] != + while (p0 < e0 && cs->to_upper[(uint) (uchar) *p0++] != cs->to_upper[(uint) (uchar) *s1]) /* no-op */; - if (s0 >= e0) + if (p0 >= e0) return 0; - p=s1+1; - while (s0 < e0 && p < e1 && cs->to_upper[(uint) (uchar) *s0] == - cs->to_upper[(uint) (uchar) *p]) - s0++, p++; - if (p >= e1) + + if (s_after && p0-1 > s0 && true_word_char(cs, p0[-2])) + continue; + + p1=s1+1; + while (p0 < e0 && p1 < e1 && cs->to_upper[(uint) (uchar) *p0] == + cs->to_upper[(uint) (uchar) *p1]) + p0++, p1++; + if (p1 == e1 && (!e_before || p0 == e0 || !true_word_char(cs, p0[0]))) return 1; } return 0; @@ -596,7 +605,8 @@ float ft_boolean_find_relevance(FT_INFO *ftb, byte *record, uint length) continue; end=ftsi.pos+ftsi.len; - while (ft_simple_get_word((byte **) &ftsi.pos,(byte *) end, &word)) + while (ft_simple_get_word(ftb->charset, + (byte **) &ftsi.pos, (byte *) end, &word)) { int a, b, c; for (a=0, b=ftb->queue.elements, c=(a+b)/2; b-a>1; c=(a+b)/2) diff --git a/myisam/ft_nlq_search.c b/myisam/ft_nlq_search.c index 690ad0fc453..a38c52704f7 100644 --- a/myisam/ft_nlq_search.c +++ b/myisam/ft_nlq_search.c @@ -167,17 +167,28 @@ static int walk_and_copy(FT_SUPERDOC *from, DBUG_RETURN(0); } +static int walk_and_push(FT_SUPERDOC *from, + uint32 count __attribute__((unused)), QUEUE *best) +{ + DBUG_ENTER("walk_and_copy"); + from->doc.weight+=from->tmp_weight*from->word_ptr->weight; + set_if_smaller(best->elements, ft_query_expansion_limit-1) + queue_insert(best, (byte *)& from->doc); + DBUG_RETURN(0); +} -static int FT_DOC_cmp(FT_DOC *a, FT_DOC *b) + +static int FT_DOC_cmp(void *unused __attribute__((unused)), + FT_DOC *a, FT_DOC *b) { return sgn(b->weight - a->weight); } FT_INFO *ft_init_nlq_search(MI_INFO *info, uint keynr, byte *query, - uint query_len, my_bool presort) + uint query_len, uint flags, byte *record) { - TREE allocated_wtree, *wtree=&allocated_wtree; + TREE wtree; ALL_IN_ONE aio; FT_DOC *dptr; FT_INFO *dlist=NULL; @@ -196,24 +207,47 @@ FT_INFO *ft_init_nlq_search(MI_INFO *info, uint keynr, byte *query, aio.charset=info->s->keyinfo[keynr].seg->charset; aio.keybuff=info->lastkey+info->s->base.max_key_length; - bzero(&allocated_wtree,sizeof(allocated_wtree)); + bzero(&wtree,sizeof(wtree)); init_tree(&aio.dtree,0,0,sizeof(FT_SUPERDOC),(qsort_cmp2)&FT_SUPERDOC_cmp,0, NULL, NULL); - ft_parse_init(&allocated_wtree, aio.charset); - if (ft_parse(&allocated_wtree,query,query_len)) + ft_parse_init(&wtree, aio.charset); + if (ft_parse(&wtree,query,query_len,0)) goto err; - if (tree_walk(wtree, (tree_walk_action)&walk_and_match, &aio, + if (tree_walk(&wtree, (tree_walk_action)&walk_and_match, &aio, left_root_right)) - goto err2; + goto err; + + if (flags & FT_EXPAND && ft_query_expansion_limit) + { + QUEUE best; + init_queue(&best,ft_query_expansion_limit,0,0, &FT_DOC_cmp, 0); + tree_walk(&aio.dtree, (tree_walk_action) &walk_and_push, + &best, left_root_right); + while (best.elements) + { + my_off_t docid=((FT_DOC *)queue_remove(& best, 0))->dpos; + if (!(*info->read_record)(info,docid,record)) + { + info->update|= HA_STATE_AKTIV; + _mi_ft_parse(&wtree, info, keynr, record,1); + } + } + delete_queue(&best); + reset_tree(&aio.dtree); + if (tree_walk(&wtree, (tree_walk_action)&walk_and_match, &aio, + left_root_right)) + goto err; + + } dlist=(FT_INFO *)my_malloc(sizeof(FT_INFO)+ sizeof(FT_DOC)*(aio.dtree.elements_in_tree-1), MYF(0)); - if(!dlist) - goto err2; + if (!dlist) + goto err; dlist->please= (struct _ft_vft *) & _ft_vft_nlq; dlist->ndocs=aio.dtree.elements_in_tree; @@ -224,14 +258,12 @@ FT_INFO *ft_init_nlq_search(MI_INFO *info, uint keynr, byte *query, tree_walk(&aio.dtree, (tree_walk_action) &walk_and_copy, &dptr, left_root_right); - if (presort) - qsort(dlist->doc, dlist->ndocs, sizeof(FT_DOC), (qsort_cmp)&FT_DOC_cmp); - -err2: - delete_tree(wtree); - delete_tree(&aio.dtree); + if (flags & FT_SORTED) + qsort2(dlist->doc, dlist->ndocs, sizeof(FT_DOC), (qsort2_cmp)&FT_DOC_cmp, 0); err: + delete_tree(&aio.dtree); + delete_tree(&wtree); info->lastpos=saved_lastpos; DBUG_RETURN(dlist); } diff --git a/myisam/ft_parser.c b/myisam/ft_parser.c index 14c67333734..0318bcbed10 100644 --- a/myisam/ft_parser.c +++ b/myisam/ft_parser.c @@ -73,25 +73,17 @@ FT_WORD * ft_linearize(TREE *wtree) DBUG_RETURN(wlist); } -#define true_word_char(s,X) (my_isalnum(s,X) || (X)=='_') -#ifdef HYPHEN_IS_DELIM -#define misc_word_char(X) ((X)=='\'') -#else -#define misc_word_char(X) ((X)=='\'' || (X)=='-') -#endif -#define word_char(s,X) (true_word_char(s,X) || misc_word_char(X)) - - /* returns: * 0 - eof * 1 - word found * 2 - left bracket * 3 - right bracket */ -byte ft_get_word(byte **start, byte *end, FT_WORD *word, FTB_PARAM *param) +byte ft_get_word(CHARSET_INFO *cs, byte **start, byte *end, + FT_WORD *word, FTB_PARAM *param) { byte *doc=*start; - int mwc; + uint mwc, length; param->yesno=(FTB_YES==' ') ? 1 : (param->quot != 0); param->plusminus=param->pmsign=0; @@ -100,11 +92,7 @@ byte ft_get_word(byte **start, byte *end, FT_WORD *word, FTB_PARAM *param) { for (;doc<end;doc++) { - /* - BAR TODO: discuss with Serge how to remove - default_charset_info correctly - */ - if (true_word_char(default_charset_info,*doc)) break; + if (true_word_char(cs,*doc)) break; if (*doc == FTB_RQUOT && param->quot) { param->quot=doc; *start=doc+1; @@ -132,9 +120,9 @@ byte ft_get_word(byte **start, byte *end, FT_WORD *word, FTB_PARAM *param) param->plusminus=param->pmsign=0; } - mwc=0; - for (word->pos=doc; doc<end; doc++) - if (true_word_char(default_charset_info,*doc)) + mwc=length=0; + for (word->pos=doc; doc<end; length++, doc+=my_mbcharlen(cs, *(uchar *)doc)) + if (true_word_char(cs,*doc)) mwc=0; else if (!misc_word_char(*doc) || mwc++) break; @@ -144,8 +132,8 @@ byte ft_get_word(byte **start, byte *end, FT_WORD *word, FTB_PARAM *param) if ((param->trunc=(doc<end && *doc == FTB_TRUNC))) doc++; - if (((word->len >= ft_min_word_len && !is_stopword(word->pos, word->len)) - || param->trunc) && word->len < ft_max_word_len) + if (((length >= ft_min_word_len && !is_stopword(word->pos, word->len)) + || param->trunc) && length < ft_max_word_len) { *start=doc; return 1; @@ -154,32 +142,33 @@ byte ft_get_word(byte **start, byte *end, FT_WORD *word, FTB_PARAM *param) return 0; } -byte ft_simple_get_word(byte **start, byte *end, FT_WORD *word) +byte ft_simple_get_word(CHARSET_INFO *cs, byte **start, byte *end, + FT_WORD *word) { - byte *doc=*start; - int mwc; + byte *doc= *start; + uint mwc, length; DBUG_ENTER("ft_simple_get_word"); while (doc<end) { for (;doc<end;doc++) { - if (true_word_char(default_charset_info,*doc)) break; + if (true_word_char(cs,*doc)) break; } - mwc=0; - for(word->pos=doc; doc<end; doc++) - if (true_word_char(default_charset_info,*doc)) - mwc=0; + mwc= length= 0; + for (word->pos=doc; doc<end; length++, doc+=my_mbcharlen(cs, *(uchar *)doc)) + if (true_word_char(cs,*doc)) + mwc= 0; else if (!misc_word_char(*doc) || mwc++) break; word->len= (uint)(doc-word->pos) - mwc; - if (word->len >= ft_min_word_len && word->len < ft_max_word_len && + if (length >= ft_min_word_len && length < ft_max_word_len && !is_stopword(word->pos, word->len)) { - *start=doc; + *start= doc; DBUG_RETURN(1); } } @@ -194,14 +183,23 @@ void ft_parse_init(TREE *wtree, CHARSET_INFO *cs) DBUG_VOID_RETURN; } -int ft_parse(TREE *wtree, byte *doc, int doclen) +int ft_parse(TREE *wtree, byte *doc, int doclen, my_bool with_alloc) { byte *end=doc+doclen; FT_WORD w; DBUG_ENTER("ft_parse"); - while (ft_simple_get_word(&doc,end,&w)) + while (ft_simple_get_word(wtree->custom_arg, &doc,end,&w)) { + if (with_alloc) + { + byte *ptr; + /* allocating the data in the tree - to avoid mallocs and frees */ + DBUG_ASSERT(wtree->with_delete==0); + ptr=(byte *)alloc_root(& wtree->mem_root,w.len); + memcpy(ptr, w.pos, w.len); + w.pos=ptr; + } if (!tree_insert(wtree, &w, 0, wtree->custom_arg)) goto err; } diff --git a/myisam/ft_static.c b/myisam/ft_static.c index 44e80847fd7..2edb86679e9 100644 --- a/myisam/ft_static.c +++ b/myisam/ft_static.c @@ -20,7 +20,7 @@ ulong ft_min_word_len=4; ulong ft_max_word_len=HA_FT_MAXLEN; -ulong ft_max_word_len_for_sort=20; +ulong ft_query_expansion_limit=5; const char *ft_boolean_syntax="+ -><()~*:\"\"&|"; const HA_KEYSEG ft_keysegs[FT_SEGS]={ @@ -53,14 +53,13 @@ const struct _ft_vft _ft_vft_boolean = { ft_boolean_get_relevance, ft_boolean_reinit_search }; -FT_INFO *(*_ft_init_vft[2])(MI_INFO *, uint, byte *, uint, my_bool) = -{ ft_init_nlq_search, ft_init_boolean_search }; - -FT_INFO *ft_init_search(uint mode, void *info, uint keynr, - byte *query, uint query_len, my_bool presort) +FT_INFO *ft_init_search(uint flags, void *info, uint keynr, + byte *query, uint query_len, byte *record) { - return (*_ft_init_vft[mode])((MI_INFO *)info, keynr, - query, query_len, presort); + if (flags & FT_BOOL) + ft_init_boolean_search((MI_INFO *)info, keynr, query, query_len); + else + ft_init_nlq_search((MI_INFO *)info, keynr, query, query_len, flags, record); } const char *ft_stopword_file = 0; diff --git a/myisam/ft_stopwords.c b/myisam/ft_stopwords.c index 5163aca60bd..6682de18c65 100644 --- a/myisam/ft_stopwords.c +++ b/myisam/ft_stopwords.c @@ -81,7 +81,7 @@ int ft_init_stopwords() goto err0; len=my_read(fd, buffer, len, MYF(MY_WME)); end=start+len; - while (ft_simple_get_word(&start, end, &w)) + while (ft_simple_get_word(default_charset_info, &start, end, &w)) { if (ft_add_stopword(my_strdup_with_length(w.pos, w.len, MYF(0)))) goto err1; diff --git a/myisam/ft_update.c b/myisam/ft_update.c index cdf3b306087..2dc397e864f 100644 --- a/myisam/ft_update.c +++ b/myisam/ft_update.c @@ -21,13 +21,6 @@ #include "ftdefs.h" #include <math.h> -/************************************************************** - This is to make ft-code to ignore keyseg.length at all * - and to index the whole VARCHAR/BLOB instead... */ -#undef set_if_smaller -#define set_if_smaller(A,B) /* no op */ -/**************************************************************/ - void _mi_ft_segiterator_init(MI_INFO *info, uint keynr, const byte *record, FT_SEG_ITERATOR *ftsi) { @@ -88,7 +81,6 @@ uint _mi_ft_segiterator(register FT_SEG_ITERATOR *ftsi) { ftsi->len=uint2korr(ftsi->pos); ftsi->pos+=2; /* Skip VARCHAR length */ - set_if_smaller(ftsi->len,ftsi->seg->length); DBUG_RETURN(1); } if (ftsi->seg->flag & HA_BLOB_PART) @@ -96,7 +88,6 @@ uint _mi_ft_segiterator(register FT_SEG_ITERATOR *ftsi) ftsi->len=_mi_calc_blob_length(ftsi->seg->bit_start,ftsi->pos); memcpy_fixed((char*) &ftsi->pos, ftsi->pos+ftsi->seg->bit_start, sizeof(char*)); - set_if_smaller(ftsi->len,ftsi->seg->length); DBUG_RETURN(1); } ftsi->len=ftsi->seg->length; @@ -106,7 +97,8 @@ uint _mi_ft_segiterator(register FT_SEG_ITERATOR *ftsi) /* parses a document i.e. calls ft_parse for every keyseg */ -uint _mi_ft_parse(TREE *parsed, MI_INFO *info, uint keynr, const byte *record) +uint _mi_ft_parse(TREE *parsed, MI_INFO *info, uint keynr, + const byte *record, my_bool with_alloc) { FT_SEG_ITERATOR ftsi; DBUG_ENTER("_mi_ft_parse"); @@ -117,21 +109,19 @@ uint _mi_ft_parse(TREE *parsed, MI_INFO *info, uint keynr, const byte *record) while (_mi_ft_segiterator(&ftsi)) { if (ftsi.pos) - if (ft_parse(parsed, (byte *)ftsi.pos, ftsi.len)) + if (ft_parse(parsed, (byte *)ftsi.pos, ftsi.len, with_alloc)) DBUG_RETURN(1); } DBUG_RETURN(0); } -FT_WORD * _mi_ft_parserecord(MI_INFO *info, uint keynr, - byte *keybuf __attribute__((unused)), - const byte *record) +FT_WORD * _mi_ft_parserecord(MI_INFO *info, uint keynr, const byte *record) { TREE ptree; DBUG_ENTER("_mi_ft_parserecord"); bzero((char*) &ptree, sizeof(ptree)); - if (_mi_ft_parse(&ptree, info, keynr, record)) + if (_mi_ft_parse(&ptree, info, keynr, record,0)) DBUG_RETURN(NULL); DBUG_RETURN(ft_linearize(&ptree)); @@ -208,9 +198,9 @@ int _mi_ft_update(MI_INFO *info, uint keynr, byte *keybuf, int cmp, cmp2; DBUG_ENTER("_mi_ft_update"); - if (!(old_word=oldlist=_mi_ft_parserecord(info, keynr, keybuf, oldrec))) + if (!(old_word=oldlist=_mi_ft_parserecord(info, keynr, oldrec))) goto err0; - if (!(new_word=newlist=_mi_ft_parserecord(info, keynr, keybuf, newrec))) + if (!(new_word=newlist=_mi_ft_parserecord(info, keynr, newrec))) goto err1; error=0; @@ -258,7 +248,7 @@ int _mi_ft_add(MI_INFO *info, uint keynr, byte *keybuf, const byte *record, FT_WORD *wlist; DBUG_ENTER("_mi_ft_add"); - if ((wlist=_mi_ft_parserecord(info, keynr, keybuf, record))) + if ((wlist=_mi_ft_parserecord(info, keynr, record))) { error=_mi_ft_store(info,keynr,keybuf,wlist,pos); my_free((char*) wlist,MYF(0)); @@ -277,7 +267,7 @@ int _mi_ft_del(MI_INFO *info, uint keynr, byte *keybuf, const byte *record, DBUG_ENTER("_mi_ft_del"); DBUG_PRINT("enter",("keynr: %d",keynr)); - if ((wlist=_mi_ft_parserecord(info, keynr, keybuf, record))) + if ((wlist=_mi_ft_parserecord(info, keynr, record))) { error=_mi_ft_erase(info,keynr,keybuf,wlist,pos); my_free((char*) wlist,MYF(0)); @@ -305,3 +295,53 @@ uint _ft_make_key(MI_INFO *info, uint keynr, byte *keybuf, FT_WORD *wptr, memcpy(buf+HA_FT_WLEN+2,wptr->pos,wptr->len); DBUG_RETURN(_mi_make_key(info,keynr,(uchar*) keybuf,buf,filepos)); } + +/* + convert key value to ft2 +*/ +uint _mi_ft_convert_to_ft2(MI_INFO *info, uint keynr, uchar *key) +{ + my_off_t root; + DYNAMIC_ARRAY *da=info->ft1_to_ft2; + MI_KEYDEF *keyinfo=&info->s->ft2_keyinfo; + uchar *key_ptr=dynamic_array_ptr(da, 0), *end; + uint length, key_length; + DBUG_ENTER("_mi_ft_convert_to_ft2"); + + /* we'll generate one pageful at once, and insert the rest one-by-one */ + /* calculating the length of this page ...*/ + length=(keyinfo->block_length-2) / keyinfo->keylength; + set_if_smaller(length, da->elements); + length=length * keyinfo->keylength; + + get_key_full_length_rdonly(key_length, key); + while (_mi_ck_delete(info, keynr, key, key_length) == 0) + /* nothing to do here. + _mi_ck_delete() will populate info->ft1_to_ft2 with deleted keys + */; + + /* creating pageful of keys */ + mi_putint(info->buff,length+2,0); + memcpy(info->buff+2, key_ptr, length); + info->buff_used=info->page_changed=1; /* info->buff is used */ + if ((root= _mi_new(info,keyinfo)) == HA_OFFSET_ERROR || + _mi_write_keypage(info,keyinfo,root,info->buff)) + DBUG_RETURN(-1); + + /* inserting the rest of key values */ + end=dynamic_array_ptr(da, da->elements); + for (key_ptr+=length; key_ptr < end; key_ptr+=keyinfo->keylength) + if(_mi_ck_real_write_btree(info, keyinfo, key_ptr, 0, &root, SEARCH_SAME)) + DBUG_RETURN(-1); + + /* now, writing the word key entry */ + ft_intXstore(key+key_length, -da->elements); + _mi_dpointer(info, key+key_length+HA_FT_WLEN, root); + + DBUG_RETURN(_mi_ck_real_write_btree(info, + info->s->keyinfo+keynr, + key, 0, + &info->s->state.key_root[keynr], + SEARCH_SAME)); +} + diff --git a/myisam/ftbench/Ecompare.pl b/myisam/ftbench/Ecompare.pl index 4bcee1fb588..265534e704d 100755 --- a/myisam/ftbench/Ecompare.pl +++ b/myisam/ftbench/Ecompare.pl @@ -73,14 +73,14 @@ for $file (split) { for (0..$#l1) { $pp[$_]+=$p[$_]; $mm[$_]+=$m[$_]; - $total.=rep($file, ($#l1 ? $_ : undef), $p[$_], $m[$_]); + $total[$_].=rep($file, ($#l1 ? $_ : undef), $p[$_], $m[$_]); } close OUT1; close OUT2; } for (0..$#l1) { - rep($total, ($#l1 ? $_ : undef), $pp[$_], $mm[$_]); + rep($total[$_], ($#l1 ? $_ : undef), $pp[$_], $mm[$_]); } sub rep { diff --git a/myisam/ftbench/Ecreate.pl b/myisam/ftbench/Ecreate.pl new file mode 100755 index 00000000000..d90a6f7a0ad --- /dev/null +++ b/myisam/ftbench/Ecreate.pl @@ -0,0 +1,44 @@ +#!/usr/bin/perl + +$test=shift || die "Usage $0 testname [option]"; +$option=shift; + +open(D, "<data/$test.d") || die "Cannot open(<data/$test.d): $!"; +open(Q, "<data/$test.q") || die "Cannot open(<data/$test.q): $!"; + +$N=0; + +print <<__HEADER__; +DROP TABLE IF EXISTS $test; +CREATE TABLE $test ( + id int(10) unsigned NOT NULL, + text text NOT NULL, + FULLTEXT KEY text (text) +) TYPE=MyISAM CHARSET=latin1; + +ALTER TABLE $test DISABLE KEYS; +__HEADER__ + +while (<D>) { chomp; + s/'/\\'/g; ++$N; + print "INSERT $test VALUES ($N, '$_');\n"; +} + +print <<__PREP__; +ALTER TABLE $test ENABLE KEYS; +SELECT $N; +__PREP__ + +$N=0; + +while (<Q>) { chomp; + s/'/\\'/g; ++$N; + $_="MATCH text AGAINST ('$_' $option)"; + print "SELECT $N, id, $_ FROM $test WHERE $_;\n"; +} + +print <<__FOOTER__; +DROP TABLE $test; +__FOOTER__ + + diff --git a/myisam/ftbench/Ereport.pl b/myisam/ftbench/Ereport.pl index aac06503849..5969304da09 100755 --- a/myisam/ftbench/Ereport.pl +++ b/myisam/ftbench/Ereport.pl @@ -29,7 +29,7 @@ sub Favg { my $a=shift; $Pavg*$Ravg ? 1/($a/$Pavg+(1-$a)/$Ravg) : 0; } # F0 : a=0 -- ignore precision # F5 : a=0.5 # F1 : a=1 -- ignore recall - while($eout_str =~ /^$qid\s+(\d+)\s+(\d+\.\d+)/) { + while($eout_str =~ /^$qid\s+(\d+)\s+(\d+(?:\.\d+)?)/) { $B++; $AB++ if $dq{$1+0}; $Ravg+=$AB; @@ -41,11 +41,7 @@ sub Favg { my $a=shift; $Pavg*$Ravg ? 1/($a/$Pavg+(1-$a)/$Ravg) : 0; } $Ravg/=$B*$A if $B; $Pavg/=$B if $B; - write; -format= -@##### @#.####### -$qid, Favg(0.5) -. + printf "%5d %1.12f %1.12f %1.12f\n", $qid, Favg(0),Favg(0.5),Favg(1); } exit 0; diff --git a/myisam/ftbench/README b/myisam/ftbench/README new file mode 100644 index 00000000000..b1f8b66b15f --- /dev/null +++ b/myisam/ftbench/README @@ -0,0 +1,43 @@ +1. should be run from myisam/ftbench/ +2. myisam/ftdefs.h should NOT be locked (bk get, not bk edit!) +3. there should be ./data/ subdir with test collections, files: + test1.d + test1.q + test1.r + test2.d + test2.q + test2.r + where test1, test2, etc - are arbitrary test names + + *.[dq] files contain documents/queries one item per line. + + *.r files have the structure: + 1 16 .....blablabla + 1 09 .....blablabla + 2 116 .....blablabla + ... + + that is /^\d+\s+\d+/ + and are sorted by the first number (not necessarily by the second) + +4. there should be ./t/ subdir with test directories + + ./t + ./t/BEST/ + ./t/testdir1/ + ./t/testdir2/ + ... + + there *must* be ./t/BEST/ subdir or a symlink to one of other dirs in ./t + all other names (besides BEST) can be arbitrary + + all test results are compared with BEST results. + + test directories may contain ftdefs.h, my.cnf, ft_mode + (the last one is used as in ... MATCH ... AGAINST ("..." $ft_mode) ...) + NOTE: all *.out files in test directories will NOT be overwritten! + delete them to re-test + +5. run ./ft-test-run.sh +6. go make some coffee + diff --git a/myisam/ftbench/ft-test-run.sh b/myisam/ftbench/ft-test-run.sh index 4c81cac6d4f..ceba818fa5c 100755 --- a/myisam/ftbench/ft-test-run.sh +++ b/myisam/ftbench/ft-test-run.sh @@ -1,4 +1,4 @@ -#!/bin/sh -x +#!/bin/sh if [ ! -x ./ft-test-run.sh ] ; then echo "Usage: ./ft-test-run.sh" @@ -15,12 +15,7 @@ SOCK=$DATA/mysql.sock PID=$DATA/mysql.pid H=../ftdefs.h OPTS="--no-defaults --socket=$SOCK --character-sets-dir=$ROOT/sql/share/charsets" - -# --ft_min_word_len=# -# --ft_max_word_len=# -# --ft_max_word_len_for_sort=# -# --ft_stopword_file=name -# --key_buffer_size=# +DELAY=10 stop_myslqd() { @@ -38,44 +33,66 @@ if [ -w $H ] ; then exit 1 fi -for batch in t/BEST t/* ; do - A=`ls $batch/*.out` - [ ! -d $batch -o -n "$A" ] && continue +stop_myslqd +rm -rf var > /dev/null 2>&1 +mkdir var +mkdir var/test + +for batch in t/* ; do + [ ! -d $batch ] && continue + [ $batch -ef t/BEST -a $batch != t/BEST ] && continue + + rm -rf var/test/* > /dev/null 2>&1 rm -f $H - ln -s $BASE/$batch/ftdefs.h $H - touch $H + if [ -f $BASE/$batch/ftdefs.h ] ; then + cat $BASE/$batch/ftdefs.h > $H + chmod a-wx $H + else + bk get -q $H + fi OPTS="--defaults-file=$BASE/$batch/my.cnf --socket=$SOCK --character-sets-dir=$ROOT/sql/share/charsets" stop_myslqd - rm $MYSQLD - (cd $ROOT; gmake) + rm -f $MYSQLD + echo "building $batch" + echo "============== $batch ===============" >> var/ft_test.log + (cd $ROOT; gmake) >> var/ft_test.log 2>&1 for prog in $MYSQLD $MYSQL $MYSQLADMIN ; do if [ ! -x $prog ] ; then - echo "No $prog" + echo "build failed: no $prog" exit 1 fi done - rm -rf var 2>&1 >/dev/null - mkdir var - mkdir var/test - + echo "=====================================" >> var/ft_test.log $MYSQLD $OPTS --basedir=$BASE --skip-bdb --pid-file=$PID \ --language=$ROOT/sql/share/english \ --skip-grant-tables --skip-innodb \ - --skip-networking --tmpdir=$DATA & + --skip-networking --tmpdir=$DATA >> var/ft_test.log 2>&1 & - $MYSQLADMIN $OPTS --connect_timeout=60 ping + sleep $DELAY + $MYSQLADMIN $OPTS ping if [ $? != 0 ] ; then echo "$MYSQLD refused to start" exit 1 fi - for test in `cd data; echo *.test|sed "s/\.test\>//g"` ; do - $MYSQL $OPTS --skip-column-names test <data/$test.test >var/$test.eval - ./Ereport.pl var/$test.eval data/$test.relj > $batch/$test.out || exit + for test in `cd data; echo *.r|sed "s/\.r//g"` ; do + if [ -f $batch/$test.out ] ; then + echo "skipping $batch/$test.out" + continue + fi + echo "testing $batch/$test" + FT_MODE=`cat $batch/ft_mode 2>/dev/null` + ./Ecreate.pl $test "$FT_MODE" | $MYSQL $OPTS --skip-column-names test >var/$test.eval + echo "reporting $batch/$test" + ./Ereport.pl var/$test.eval data/$test.r > $batch/$test.out || exit done stop_myslqd rm -f $H - [ $batch -ef t/BEST ] || ./Ecompare.pl t/BEST $batch >> t/BEST/report.txt + bk get -q $H + if [ ! $batch -ef t/BEST ] ; then + echo "comparing $batch" + ./Ecompare.pl t/BEST $batch >> t/BEST/report.txt + fi done diff --git a/myisam/ftdefs.h b/myisam/ftdefs.h index 88d7e79937b..a97485ec021 100644 --- a/myisam/ftdefs.h +++ b/myisam/ftdefs.h @@ -21,9 +21,14 @@ #include "fulltext.h" #include <m_ctype.h> #include <my_tree.h> +#include <queues.h> +#include <assert.h> -#define HYPHEN_IS_DELIM -#define HYPHEN_IS_CONCAT /* not used for now */ +#define true_word_char(s,X) (my_isalnum(s,X) || (X)=='_') +#define misc_word_char(X) ((X)=='\'') +#define word_char(s,X) (true_word_char(s,X) || misc_word_char(X)) + +#define FT_MAX_WORD_LEN_FOR_SORT 20 #define COMPILE_STOPWORDS_IN @@ -107,8 +112,8 @@ int is_stopword(char *word, uint len); uint _ft_make_key(MI_INFO *, uint , byte *, FT_WORD *, my_off_t); -byte ft_get_word(byte **, byte *, FT_WORD *, FTB_PARAM *); -byte ft_simple_get_word(byte **, byte *, FT_WORD *); +byte ft_get_word(CHARSET_INFO *, byte **, byte *, FT_WORD *, FTB_PARAM *); +byte ft_simple_get_word(CHARSET_INFO *, byte **, byte *, FT_WORD *); typedef struct _st_ft_seg_iterator { uint num, len; @@ -121,13 +126,15 @@ void _mi_ft_segiterator_dummy_init(const byte *, uint, FT_SEG_ITERATOR *); uint _mi_ft_segiterator(FT_SEG_ITERATOR *); void ft_parse_init(TREE *, CHARSET_INFO *); -int ft_parse(TREE *, byte *, int); +int ft_parse(TREE *, byte *, int, my_bool); FT_WORD * ft_linearize(TREE *); -FT_WORD * _mi_ft_parserecord(MI_INFO *, uint, byte *, const byte *); -uint _mi_ft_parse(TREE *parsed, MI_INFO *info, uint keynr, const byte *record); +FT_WORD * _mi_ft_parserecord(MI_INFO *, uint, const byte *); +uint _mi_ft_parse(TREE *, MI_INFO *, uint, const byte *, my_bool); + +FT_INFO *ft_init_nlq_search(MI_INFO *, uint, byte *, uint, uint, byte *); +FT_INFO *ft_init_boolean_search(MI_INFO *, uint, byte *, uint); extern const struct _ft_vft _ft_vft_nlq; -FT_INFO *ft_init_nlq_search(MI_INFO *, uint, byte *, uint, my_bool); int ft_nlq_read_next(FT_INFO *, char *); float ft_nlq_find_relevance(FT_INFO *, byte *, uint); void ft_nlq_close_search(FT_INFO *); @@ -136,10 +143,10 @@ my_off_t ft_nlq_get_docid(FT_INFO *); void ft_nlq_reinit_search(FT_INFO *); extern const struct _ft_vft _ft_vft_boolean; -FT_INFO *ft_init_boolean_search(MI_INFO *, uint, byte *, uint, my_bool); int ft_boolean_read_next(FT_INFO *, char *); float ft_boolean_find_relevance(FT_INFO *, byte *, uint); void ft_boolean_close_search(FT_INFO *); float ft_boolean_get_relevance(FT_INFO *); my_off_t ft_boolean_get_docid(FT_INFO *); void ft_boolean_reinit_search(FT_INFO *); + diff --git a/myisam/fulltext.h b/myisam/fulltext.h index ec267eb3e86..d8c74d4e94b 100644 --- a/myisam/fulltext.h +++ b/myisam/fulltext.h @@ -34,3 +34,5 @@ int _mi_ft_cmp(MI_INFO *, uint, const byte *, const byte *); int _mi_ft_add(MI_INFO *, uint, byte *, const byte *, my_off_t); int _mi_ft_del(MI_INFO *, uint, byte *, const byte *, my_off_t); +uint _mi_ft_convert_to_ft2(MI_INFO *, uint, uchar *); + diff --git a/myisam/mi_check.c b/myisam/mi_check.c index fe7ff1af407..a75ac75a2e9 100644 --- a/myisam/mi_check.c +++ b/myisam/mi_check.c @@ -20,7 +20,6 @@ #include <m_ctype.h> #include <stdarg.h> #include <my_getopt.h> -#include <assert.h> #ifdef HAVE_SYS_VADVISE_H #include <sys/vadvise.h> #endif @@ -1961,11 +1960,11 @@ int mi_repair_by_sort(MI_CHECK *param, register MI_INFO *info, if (sort_param.keyinfo->flag & HA_FULLTEXT) { sort_info.max_records= - (ha_rows) (sort_info.filelength/ft_max_word_len_for_sort+1); + (ha_rows) (sort_info.filelength/FT_MAX_WORD_LEN_FOR_SORT+1); sort_param.key_read=sort_ft_key_read; sort_param.key_write=sort_ft_key_write; - sort_param.key_length+=ft_max_word_len_for_sort-HA_FT_MAXLEN; + sort_param.key_length+=FT_MAX_WORD_LEN_FOR_SORT-HA_FT_MAXLEN; } else { @@ -2356,7 +2355,7 @@ int mi_repair_parallel(MI_CHECK *param, register MI_INFO *info, total_key_length+=sort_param[i].key_length; if (sort_param[i].keyinfo->flag & HA_FULLTEXT) - sort_param[i].key_length+=ft_max_word_len_for_sort-ft_max_word_len; + sort_param[i].key_length+=FT_MAX_WORD_LEN_FOR_SORT-ft_max_word_len; } sort_info.total_keys=i; sort_param[0].master= 1; @@ -2581,8 +2580,7 @@ static int sort_ft_key_read(MI_SORT_PARAM *sort_param, void *key) my_free((char*) wptr, MYF(MY_ALLOW_ZERO_PTR)); if ((error=sort_get_next_record(sort_param))) DBUG_RETURN(error); - if (!(wptr=_mi_ft_parserecord(info,sort_param->key, - key,sort_param->record))) + if (!(wptr=_mi_ft_parserecord(info,sort_param->key,sort_param->record))) DBUG_RETURN(1); if (wptr->pos) break; @@ -3883,7 +3881,7 @@ static my_bool mi_too_big_key_for_sort(MI_KEYDEF *key, ha_rows rows) { uint key_maxlength=key->maxlength; if (key->flag & HA_FULLTEXT) - key_maxlength+=ft_max_word_len_for_sort-HA_FT_MAXLEN; + key_maxlength+=FT_MAX_WORD_LEN_FOR_SORT-HA_FT_MAXLEN; return (key->flag & (HA_BINARY_PACK_KEY | HA_VAR_LENGTH_KEY | HA_FULLTEXT) && ((ulonglong) rows * key_maxlength > (ulonglong) myisam_max_temp_length)); @@ -3899,7 +3897,7 @@ void mi_disable_non_unique_index(MI_INFO *info, ha_rows rows) MI_KEYDEF *key=share->keyinfo; for (i=0 ; i < share->base.keys ; i++,key++) { - if (!(key->flag & (HA_NOSAME | HA_SPATIAL | HA_AUTO_KEY)) && + if (!(key->flag & (HA_NOSAME | HA_SPATIAL | HA_AUTO_KEY)) && ! mi_too_big_key_for_sort(key,rows) && info->s->base.auto_key != i+1) { share->state.key_map&= ~ ((ulonglong) 1 << i); diff --git a/myisam/mi_delete.c b/myisam/mi_delete.c index 5635e0a2139..f21b0e928b6 100644 --- a/myisam/mi_delete.c +++ b/myisam/mi_delete.c @@ -18,6 +18,7 @@ #include "fulltext.h" #include "rt_index.h" +#include <assert.h> #ifdef __WIN__ #include <errno.h> @@ -232,13 +233,22 @@ static int d_search(register MI_INFO *info, register MI_KEYDEF *keyinfo, get_key_full_length_rdonly(off, lastkey); subkeys=ft_sintXkorr(lastkey+off); + DBUG_ASSERT(info->ft1_to_ft2==0 || subkeys >=0); comp_flag=SEARCH_SAME; if (subkeys >= 0) { /* normal word, one-level tree structure */ - DBUG_PRINT("info",("FT1")); - flag=(*keyinfo->bin_search)(info,keyinfo,anc_buff,key,USE_WHOLE_KEY, - comp_flag, &keypos, lastkey, &last_key); + if (info->ft1_to_ft2) + { + /* we're in ft1->ft2 conversion mode. Saving key data */ + insert_dynamic(info->ft1_to_ft2, lastkey+off); + } + else + { + /* we need exact match only if not in ft1->ft2 conversion mode */ + flag=(*keyinfo->bin_search)(info,keyinfo,anc_buff,key,USE_WHOLE_KEY, + comp_flag, &keypos, lastkey, &last_key); + } /* fall through to normal delete */ } else @@ -253,13 +263,11 @@ static int d_search(register MI_INFO *info, register MI_KEYDEF *keyinfo, if (subkeys == -1) { /* the last entry in sub-tree */ - DBUG_PRINT("info",("FT2: the last entry")); _mi_dispose(info, keyinfo, root,DFLT_INIT_HITS); /* fall through to normal delete */ } else { - DBUG_PRINT("info",("FT2: going down")); keyinfo=&info->s->ft2_keyinfo; kpos-=keyinfo->keylength+nod_flag; /* we'll modify key entry 'in vivo' */ key+=off; diff --git a/myisam/mi_open.c b/myisam/mi_open.c index c4b24acdb77..744bb9bb3b6 100644 --- a/myisam/mi_open.c +++ b/myisam/mi_open.c @@ -513,8 +513,8 @@ MI_INFO *mi_open(const char *name, int mode, uint open_flags) NullS)) goto err; errpos=6; - - if (!have_rtree) + + if (!have_rtree) info.rtree_recursion_state= NULL; strmov(info.filename,org_name); @@ -536,6 +536,7 @@ MI_INFO *mi_open(const char *name, int mode, uint open_flags) info.lock_type=F_UNLCK; info.quick_mode=0; info.bulk_insert=0; + info.ft1_to_ft2=0; info.errkey= -1; info.page_changed=1; pthread_mutex_lock(&share->intern_lock); @@ -1112,7 +1113,7 @@ char *mi_recinfo_read(char *ptr, MI_COLUMNDEF *recinfo) /************************************************************************** Open data file with or without RAID -We can't use dup() here as the data file descriptors need to have different +We can't use dup() here as the data file descriptors need to have different active seek-positions. The argument file_to_dup is here for the future if there would on some OS diff --git a/myisam/mi_write.c b/myisam/mi_write.c index 1fd5cddec62..86c45e0692f 100644 --- a/myisam/mi_write.c +++ b/myisam/mi_write.c @@ -18,6 +18,7 @@ #include "fulltext.h" #include "rt_index.h" +#include <assert.h> #ifdef __WIN__ #include <errno.h> @@ -124,7 +125,7 @@ int mi_write(MI_INFO *info, byte *record) else { if (share->keyinfo[i].ck_insert(info,i,buff, - _mi_make_key(info,i,buff,record,filepos))) + _mi_make_key(info,i,buff,record,filepos))) { if (local_lock_tree) rw_unlock(&share->key_root_lock[i]); @@ -264,13 +265,32 @@ int _mi_ck_write_btree(register MI_INFO *info, uint keynr, uchar *key, else comp_flag=SEARCH_SAME; /* Keys in rec-pos order */ + error=_mi_ck_real_write_btree(info, keyinfo, key, key_length, + root, comp_flag); + if (info->ft1_to_ft2) + { + if (!error) + error= _mi_ft_convert_to_ft2(info, keynr, key); + delete_dynamic(info->ft1_to_ft2); + my_free((gptr)info->ft1_to_ft2, MYF(0)); + info->ft1_to_ft2=0; + } + DBUG_RETURN(error); +} /* _mi_ck_write_btree */ + +int _mi_ck_real_write_btree(MI_INFO *info, MI_KEYDEF *keyinfo, + uchar *key, uint key_length, my_off_t *root, uint comp_flag) +{ + int error; + DBUG_ENTER("_mi_ck_real_write_btree"); + /* key_length parameter is used only if comp_flag is SEARCH_FIND */ if (*root == HA_OFFSET_ERROR || (error=w_search(info, keyinfo, comp_flag, key, key_length, *root, (uchar *) 0, (uchar*) 0, (my_off_t) 0, 1)) > 0) error=_mi_enlarge_root(info,keyinfo,key,root); DBUG_RETURN(error); -} /* _mi_ck_write_btree */ +} /* _mi_ck_real_write_btree */ /* Make a new root with key as only pointer */ @@ -359,13 +379,11 @@ static int w_search(register MI_INFO *info, register MI_KEYDEF *keyinfo, keyinfo=&info->s->ft2_keyinfo; key+=off; keypos-=keyinfo->keylength; /* we'll modify key entry 'in vivo' */ - if ((error=w_search(info, keyinfo, comp_flag, key, HA_FT_WLEN, root, - (uchar *) 0, (uchar*) 0, (my_off_t) 0, 1)) > 0) - { - error=_mi_enlarge_root(info, keyinfo, key, &root); - _mi_dpointer(info, keypos+HA_FT_WLEN, root); - } + error=_mi_ck_real_write_btree(info, keyinfo, key, 0, + &root, comp_flag); + _mi_dpointer(info, keypos+HA_FT_WLEN, root); subkeys--; /* should there be underflow protection ? */ + DBUG_ASSERT(subkeys < 0); ft_intXstore(keypos, subkeys); if (!error) error=_mi_write_keypage(info,keyinfo,page,DFLT_INIT_HITS,temp_buff); @@ -410,7 +428,6 @@ int _mi_insert(register MI_INFO *info, register MI_KEYDEF *keyinfo, uchar *key, uchar *anc_buff, uchar *key_pos, uchar *key_buff, uchar *father_buff, uchar *father_key_pos, my_off_t father_page, my_bool insert_last) - { uint a_length,nod_flag; int t_length; @@ -464,8 +481,56 @@ int _mi_insert(register MI_INFO *info, register MI_KEYDEF *keyinfo, a_length+=t_length; mi_putint(anc_buff,a_length,nod_flag); if (a_length <= keyinfo->block_length) + { + if (keyinfo->block_length - a_length < 32 && + keyinfo->flag & HA_FULLTEXT && key_pos == endpos && + info->s->base.key_reflength <= info->s->base.rec_reflength && + info->s->options & (HA_OPTION_PACK_RECORD | HA_OPTION_COMPRESS_RECORD)) + { + /* + Normal word. One-level tree. Page is almost full. + Let's consider converting. + We'll compare 'key' and the first key at anc_buff + */ + uchar *a=key, *b=anc_buff+2+nod_flag; + uint alen, blen, ft2len=info->s->ft2_keyinfo.keylength; + /* the very first key on the page is always unpacked */ + DBUG_ASSERT((*b & 128) == 0); +#if HA_FT_MAXLEN >= 127 + blen= mi_uint2korr(b); b+=2; +#else + blen= *b++; +#endif + get_key_length(alen,a); + DBUG_ASSERT(info->ft1_to_ft2==0); + if (alen == blen && + mi_compare_text(keyinfo->seg->charset, a, alen, b, blen, 0)==0) + { + /* yup. converting */ + info->ft1_to_ft2=(DYNAMIC_ARRAY *) + my_malloc(sizeof(DYNAMIC_ARRAY), MYF(MY_WME)); + my_init_dynamic_array(info->ft1_to_ft2, ft2len, 300, 50); + + /* + now, adding all keys from the page to dynarray + if the page is a leaf (if not keys will be deleted later) + */ + if (!nod_flag) + { + /* let's leave the first key on the page, though, because + we cannot easily dispatch an empty page here */ + b+=blen+ft2len+2; + for (a=anc_buff+a_length ; b < a ; b+=ft2len+2) + insert_dynamic(info->ft1_to_ft2, b); + + /* fixing the page's length - it contains only one key now */ + mi_putint(anc_buff,2+blen+ft2len+2,0); + } + /* the rest will be done when we're back from recursion */ + } + } DBUG_RETURN(0); /* There is room on page */ - + } /* Page is full */ if (nod_flag) insert_last=0; diff --git a/myisam/myisamchk.c b/myisam/myisamchk.c index b03f76ea8e8..5a3768c9932 100644 --- a/myisam/myisamchk.c +++ b/myisam/myisamchk.c @@ -321,9 +321,6 @@ static struct my_option my_long_options[] = { "ft_max_word_len", OPT_FT_MAX_WORD_LEN, "", (gptr*) &ft_max_word_len, (gptr*) &ft_max_word_len, 0, GET_ULONG, REQUIRED_ARG, HA_FT_MAXLEN, 10, HA_FT_MAXLEN, 0, 1, 0}, - { "ft_max_word_len_for_sort", OPT_FT_MAX_WORD_LEN_FOR_SORT, "", - (gptr*) &ft_max_word_len_for_sort, (gptr*) &ft_max_word_len_for_sort, 0, - GET_ULONG, REQUIRED_ARG, 20, 4, HA_FT_MAXLEN, 0, 1, 0}, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0} }; diff --git a/myisam/myisamdef.h b/myisam/myisamdef.h index fe9cd151c57..3b1dd1fbf92 100644 --- a/myisam/myisamdef.h +++ b/myisam/myisamdef.h @@ -224,7 +224,8 @@ struct st_myisam_info { MI_BLOB *blobs; /* Pointer to blobs */ MI_BIT_BUFF bit_buff; /* accumulate indexfile changes between write's */ - TREE *bulk_insert; + TREE *bulk_insert; + DYNAMIC_ARRAY *ft1_to_ft2; /* used only in ft1->ft2 conversion */ char *filename; /* parameter to open filename */ uchar *buff, /* Temp area for key */ *lastkey,*lastkey2; /* Last used search key */ @@ -466,6 +467,9 @@ extern int _mi_delete_static_record(MI_INFO *info); extern int _mi_cmp_static_record(MI_INFO *info,const byte *record); extern int _mi_read_rnd_static_record(MI_INFO*, byte *,my_off_t, my_bool); extern int _mi_ck_write(MI_INFO *info,uint keynr,uchar *key,uint length); +extern int _mi_ck_real_write_btree(MI_INFO *info, MI_KEYDEF *keyinfo, + uchar *key, uint key_length, + my_off_t *root, uint comp_flag); extern int _mi_enlarge_root(MI_INFO *info,MI_KEYDEF *keyinfo,uchar *key, my_off_t *root); extern int _mi_insert(MI_INFO *info,MI_KEYDEF *keyinfo,uchar *key, uchar *anc_buff,uchar *key_pos,uchar *key_buff, diff --git a/mysql-test/r/fulltext.result b/mysql-test/r/fulltext.result index e2d35973383..31ed252b22d 100644 --- a/mysql-test/r/fulltext.result +++ b/mysql-test/r/fulltext.result @@ -18,11 +18,20 @@ Full-text indexes are called collections Only MyISAM tables support collections select * from t1 where MATCH(a,b) AGAINST ("only"); a b -select * from t1 where MATCH(a,b) AGAINST ("collections") UNION ALL select * from t1 where MATCH(a,b) AGAINST ("indexes"); +select * from t1 where MATCH(a,b) AGAINST ("collections" WITH QUERY EXPANSION); a b Only MyISAM tables support collections Full-text indexes are called collections +MySQL has now support for full-text search +select * from t1 where MATCH(a,b) AGAINST ("indexes" WITH QUERY EXPANSION); +a b +Full-text indexes are called collections +Only MyISAM tables support collections +select * from t1 where MATCH(a,b) AGAINST ("indexes collections" WITH QUERY EXPANSION); +a b Full-text indexes are called collections +Only MyISAM tables support collections +MySQL has now support for full-text search explain select * from t1 where MATCH(a,b) AGAINST ("collections"); id select_type table type possible_keys key key_len ref rows Extra 1 SIMPLE t1 fulltext a a 0 1 Using where @@ -119,7 +128,8 @@ a b MySQL has now support for full-text search select * from t1 where MATCH a,b AGAINST ('"text i"' IN BOOLEAN MODE); a b -Full-text indexes are called collections +select * from t1 where MATCH a,b AGAINST ('"xt indexes"' IN BOOLEAN MODE); +a b select * from t1 where MATCH a AGAINST ("search" IN BOOLEAN MODE); a b Full-text search in MySQL implements vector space model @@ -190,6 +200,12 @@ ticket inhalt select * from t2 having MATCH inhalt AGAINST ('foobar'); ticket inhalt 3 foobar +CREATE TABLE t3 (t int(11),i text,fulltext tix (t,i)); +ERROR HY000: Column 't' cannot be part of FULLTEXT index +CREATE TABLE t3 (t int(11),i text, +j varchar(200) CHARACTER SET latin2, +fulltext tix (i,j)); +ERROR HY000: Column 'j' cannot be part of FULLTEXT index CREATE TABLE t3 ( ticket int(11), inhalt text, @@ -272,3 +288,24 @@ select * from t1 join t2 using(`t1_id`) where match (t1.name, t2.name) against(' t1_id name t2_id t1_id name 1 data1 1 1 xxfoo drop table t1,t2; +SET NAMES latin1; +CREATE TABLE t1 (t text character set utf8 not null, fulltext(t)); +INSERT t1 VALUES ('Mit freundlichem Grüß'), ('aus Osnabrück'); +SET NAMES koi8r; +INSERT t1 VALUES ("üÔÏ ÍÙ - ÏÐÉÌËÉ"),("ïÔÌÅÚØ, ÇÎÉÄÁ!"), +("îÅ ×ÌÅÚÁÊ, ÕÂØÅÔ!"),("É ÂÕÄÅÔ ÐÒÁ×!"); +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('ïðéìëé'); +t charset(t) +üÔÏ ÍÙ - ÏÐÉÌËÉ utf8 +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('ðÒá*' IN BOOLEAN MODE); +t charset(t) +É ÂÕÄÅÔ ÐÒÁ×! utf8 +SELECT * FROM t1 WHERE MATCH t AGAINST ('ÜÔÏ' IN BOOLEAN MODE); +t +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('Osnabrück'); +t charset(t) +SET NAMES latin1; +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('Osnabrück'); +t charset(t) +aus Osnabrück utf8 +DROP TABLE t1; diff --git a/mysql-test/r/fulltext2.result b/mysql-test/r/fulltext2.result index a35be210ece..687bc6edfa0 100644 --- a/mysql-test/r/fulltext2.result +++ b/mysql-test/r/fulltext2.result @@ -103,4 +103,106 @@ count(*) select count(*) from t1 where match a against ('aaazzz' in boolean mode); count(*) 262 -DROP TABLE IF EXISTS t1; +drop table t1; +CREATE TABLE t1 ( +i int(10) unsigned not null auto_increment primary key, +a varchar(255) not null, +FULLTEXT KEY (a) +) TYPE=MyISAM; +select count(*) from t1 where match a against ('aaaxxx'); +count(*) +260 +select count(*) from t1 where match a against ('aaayyy'); +count(*) +250 +select count(*) from t1 where match a against ('aaazzz'); +count(*) +255 +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +count(*) +260 +select count(*) from t1 where match a against ('aaayyy' in boolean mode); +count(*) +250 +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +count(*) +255 +select count(*) from t1 where match a against ('aaaxxx aaayyy aaazzz'); +count(*) +765 +select count(*) from t1 where match a against ('aaaxxx aaayyy aaazzz' in boolean mode); +count(*) +765 +select count(*) from t1 where match a against ('aaax*' in boolean mode); +count(*) +260 +select count(*) from t1 where match a against ('aaay*' in boolean mode); +count(*) +250 +select count(*) from t1 where match a against ('aaa*' in boolean mode); +count(*) +765 +insert t1 (a) values ('aaaxxx'),('aaayyy'); +insert t1 (a) values ('aaazzz'),('aaazzz'),('aaazzz'),('aaazzz'),('aaazzz'); +select count(*) from t1 where match a against ('aaaxxx'); +count(*) +261 +select count(*) from t1 where match a against ('aaayyy'); +count(*) +251 +select count(*) from t1 where match a against ('aaazzz'); +count(*) +260 +insert t1 (a) values ('aaaxxx 000000'); +select count(*) from t1 where match a against ('000000'); +count(*) +1 +delete from t1 where match a against ('000000'); +select count(*) from t1 where match a against ('000000'); +count(*) +0 +select count(*) from t1 where match a against ('aaaxxx'); +count(*) +261 +delete from t1 where match a against ('aaazzz'); +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +count(*) +261 +select count(*) from t1 where match a against ('aaayyy' in boolean mode); +count(*) +251 +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +count(*) +0 +select count(*) from t1 where a = 'aaaxxx'; +count(*) +261 +select count(*) from t1 where a = 'aaayyy'; +count(*) +251 +select count(*) from t1 where a = 'aaazzz'; +count(*) +0 +insert t1 (a) values ('aaaxxx 000000'); +select count(*) from t1 where match a against ('000000'); +count(*) +1 +update t1 set a='aaazzz' where match a against ('000000'); +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +count(*) +261 +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +count(*) +1 +update t1 set a='aaazzz' where a = 'aaaxxx'; +update t1 set a='aaaxxx' where a = 'aaayyy'; +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +count(*) +251 +select count(*) from t1 where match a against ('aaayyy' in boolean mode); +count(*) +0 +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +count(*) +262 +drop table t1; diff --git a/mysql-test/r/fulltext_var.result b/mysql-test/r/fulltext_var.result index dda8e332fba..f45e18a9591 100644 --- a/mysql-test/r/fulltext_var.result +++ b/mysql-test/r/fulltext_var.result @@ -3,5 +3,5 @@ Variable_name Value ft_boolean_syntax + -><()~*:""&| ft_min_word_len 4 ft_max_word_len 254 -ft_max_word_len_for_sort 20 +ft_query_expansion_limit 20 ft_stopword_file (built-in) diff --git a/mysql-test/r/range.result b/mysql-test/r/range.result index db477fc662f..38d3d9b2b7d 100644 --- a/mysql-test/r/range.result +++ b/mysql-test/r/range.result @@ -229,11 +229,11 @@ id select_type table type possible_keys key key_len ref rows Extra explain select * from t1, t1 t2 where t1.y = 2 and t2.x between t1.y-1 and t1.y+1; id select_type table type possible_keys key key_len ref rows Extra 1 SIMPLE t1 ref y y 5 const 1 Using where -1 SIMPLE t2 ALL x NULL NULL NULL 9 Range checked for each record (index map: 1) +1 SIMPLE t2 ALL x NULL NULL NULL 9 Range checked for each record (index map: 0x1) explain select * from t1, t1 t2 where t1.y = 2 and t2.x >= t1.y-1 and t2.x <= t1.y+1; id select_type table type possible_keys key key_len ref rows Extra 1 SIMPLE t1 ref y y 5 const 1 Using where -1 SIMPLE t2 ALL x NULL NULL NULL 9 Range checked for each record (index map: 1) +1 SIMPLE t2 ALL x NULL NULL NULL 9 Range checked for each record (index map: 0x1) explain select * from t1, t1 t2 where t1.y = 2 and t2.x between 0 and t1.y; id select_type table type possible_keys key key_len ref rows Extra 1 SIMPLE t1 ref y y 5 const 1 Using where @@ -256,11 +256,11 @@ INSERT INTO t2 VALUES (0),(0),(1),(1),(2),(2); explain select * from t1, t2 where (t1.key1 <t2.keya + 1) and t2.keya=3; id select_type table type possible_keys key key_len ref rows Extra 1 SIMPLE t2 ref j1 j1 4 const 1 Using where; Using index -1 SIMPLE t1 ALL i1,i2 NULL NULL NULL 4 Range checked for each record (index map: 3) +1 SIMPLE t1 ALL i1,i2 NULL NULL NULL 4 Range checked for each record (index map: 0x3) explain select * from t1 force index(i2), t2 where (t1.key1 <t2.keya + 1) and t2.keya=3; id select_type table type possible_keys key key_len ref rows Extra 1 SIMPLE t2 ref j1 j1 4 const 1 Using where; Using index -1 SIMPLE t1 ALL i2 NULL NULL NULL 4 Range checked for each record (index map: 2) +1 SIMPLE t1 ALL i2 NULL NULL NULL 4 Range checked for each record (index map: 0x2) DROP TABLE t1,t2; create table t1 (id int(10) primary key); insert into t1 values (1),(2),(3),(4),(5),(6),(7),(8),(9); diff --git a/mysql-test/t/fulltext.test b/mysql-test/t/fulltext.test index 8c6bb97edf1..37913d2d747 100644 --- a/mysql-test/t/fulltext.test +++ b/mysql-test/t/fulltext.test @@ -20,9 +20,11 @@ select * from t1 where MATCH(a,b) AGAINST ("indexes"); select * from t1 where MATCH(a,b) AGAINST ("indexes collections"); select * from t1 where MATCH(a,b) AGAINST ("only"); -# UNION of fulltext's -select * from t1 where MATCH(a,b) AGAINST ("collections") UNION ALL select * from t1 where MATCH(a,b) AGAINST ("indexes"); +# query expansion +select * from t1 where MATCH(a,b) AGAINST ("collections" WITH QUERY EXPANSION); +select * from t1 where MATCH(a,b) AGAINST ("indexes" WITH QUERY EXPANSION); +select * from t1 where MATCH(a,b) AGAINST ("indexes collections" WITH QUERY EXPANSION); # add_ft_keys() tests @@ -58,6 +60,7 @@ select * from t1 where MATCH a,b AGAINST ('"text search" "now support"' IN BOOL select * from t1 where MATCH a,b AGAINST ('"text search" -"now support"' IN BOOLEAN MODE); select * from t1 where MATCH a,b AGAINST ('"text search" +"now support"' IN BOOLEAN MODE); select * from t1 where MATCH a,b AGAINST ('"text i"' IN BOOLEAN MODE); +select * from t1 where MATCH a,b AGAINST ('"xt indexes"' IN BOOLEAN MODE); # boolean w/o index: @@ -65,7 +68,6 @@ select * from t1 where MATCH a AGAINST ("search" IN BOOLEAN MODE); select * from t1 where MATCH b AGAINST ("sear*" IN BOOLEAN MODE); # UNION of fulltext's - select * from t1 where MATCH(a,b) AGAINST ("collections") UNION ALL select * from t1 where MATCH(a,b) AGAINST ("indexes"); #update/delete with fulltext index @@ -129,6 +131,13 @@ select * from t2 having MATCH inhalt AGAINST ('foobar'); # check of fulltext errors # +--error 1281 +CREATE TABLE t3 (t int(11),i text,fulltext tix (t,i)); +--error 1281 +CREATE TABLE t3 (t int(11),i text, + j varchar(200) CHARACTER SET latin2, + fulltext tix (i,j)); + CREATE TABLE t3 ( ticket int(11), inhalt text, @@ -218,3 +227,21 @@ insert into t2 values (2, 1, 'xxbar'); insert into t2 values (3, 1, 'xxbuz'); select * from t1 join t2 using(`t1_id`) where match (t1.name, t2.name) against('xxfoo' in boolean mode); drop table t1,t2; + +# +# UTF8 +# +SET NAMES latin1; +CREATE TABLE t1 (t text character set utf8 not null, fulltext(t)); +INSERT t1 VALUES ('Mit freundlichem Grüß'), ('aus Osnabrück'); +SET NAMES koi8r; +INSERT t1 VALUES ("üÔÏ ÍÙ - ÏÐÉÌËÉ"),("ïÔÌÅÚØ, ÇÎÉÄÁ!"), + ("îÅ ×ÌÅÚÁÊ, ÕÂØÅÔ!"),("É ÂÕÄÅÔ ÐÒÁ×!"); +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('ïðéìëé'); +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('ðÒá*' IN BOOLEAN MODE); +SELECT * FROM t1 WHERE MATCH t AGAINST ('ÜÔÏ' IN BOOLEAN MODE); +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('Osnabrück'); +SET NAMES latin1; +SELECT t, charset(t) FROM t1 WHERE MATCH t AGAINST ('Osnabrück'); +DROP TABLE t1; + diff --git a/mysql-test/t/fulltext2.test b/mysql-test/t/fulltext2.test index b739d60e3b3..a01ec3a59c9 100644 --- a/mysql-test/t/fulltext2.test +++ b/mysql-test/t/fulltext2.test @@ -94,5 +94,83 @@ select count(*) from t1 where match a against ('aaaxxx' in boolean mode); select count(*) from t1 where match a against ('aaayyy' in boolean mode); select count(*) from t1 where match a against ('aaazzz' in boolean mode); -DROP TABLE IF EXISTS t1; +drop table t1; + +CREATE TABLE t1 ( + i int(10) unsigned not null auto_increment primary key, + a varchar(255) not null, + FULLTEXT KEY (a) +) TYPE=MyISAM; + +# two-level entry, second-level tree with depth 2 +--disable_query_log +let $1=260; +while ($1) +{ + eval insert t1 (a) values ('aaaxxx'); + dec $1; +} +let $1=255; +while ($1) +{ + eval insert t1 (a) values ('aaazzz'); + dec $1; +} +let $1=250; +while ($1) +{ + eval insert t1 (a) values ('aaayyy'); + dec $1; +} +--enable_query_log + +select count(*) from t1 where match a against ('aaaxxx'); +select count(*) from t1 where match a against ('aaayyy'); +select count(*) from t1 where match a against ('aaazzz'); +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +select count(*) from t1 where match a against ('aaayyy' in boolean mode); +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +select count(*) from t1 where match a against ('aaaxxx aaayyy aaazzz'); +select count(*) from t1 where match a against ('aaaxxx aaayyy aaazzz' in boolean mode); + +select count(*) from t1 where match a against ('aaax*' in boolean mode); +select count(*) from t1 where match a against ('aaay*' in boolean mode); +select count(*) from t1 where match a against ('aaa*' in boolean mode); + +# mi_write: + +insert t1 (a) values ('aaaxxx'),('aaayyy'); +insert t1 (a) values ('aaazzz'),('aaazzz'),('aaazzz'),('aaazzz'),('aaazzz'); +select count(*) from t1 where match a against ('aaaxxx'); +select count(*) from t1 where match a against ('aaayyy'); +select count(*) from t1 where match a against ('aaazzz'); + +# mi_delete +insert t1 (a) values ('aaaxxx 000000'); +select count(*) from t1 where match a against ('000000'); +delete from t1 where match a against ('000000'); +select count(*) from t1 where match a against ('000000'); +select count(*) from t1 where match a against ('aaaxxx'); +delete from t1 where match a against ('aaazzz'); +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +select count(*) from t1 where match a against ('aaayyy' in boolean mode); +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +# double-check without index +select count(*) from t1 where a = 'aaaxxx'; +select count(*) from t1 where a = 'aaayyy'; +select count(*) from t1 where a = 'aaazzz'; + +# update +insert t1 (a) values ('aaaxxx 000000'); +select count(*) from t1 where match a against ('000000'); +update t1 set a='aaazzz' where match a against ('000000'); +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +select count(*) from t1 where match a against ('aaazzz' in boolean mode); +update t1 set a='aaazzz' where a = 'aaaxxx'; +update t1 set a='aaaxxx' where a = 'aaayyy'; +select count(*) from t1 where match a against ('aaaxxx' in boolean mode); +select count(*) from t1 where match a against ('aaayyy' in boolean mode); +select count(*) from t1 where match a against ('aaazzz' in boolean mode); + +drop table t1; diff --git a/mysys/Makefile.am b/mysys/Makefile.am index 6ec1e1ec40d..5ff682fbc80 100644 --- a/mysys/Makefile.am +++ b/mysys/Makefile.am @@ -39,7 +39,7 @@ libmysys_a_SOURCES = my_init.c my_getwd.c mf_getdate.c\ mf_format.c mf_same.c mf_dirname.c mf_fn_ext.c \ my_symlink.c my_symlink2.c \ mf_pack.c mf_unixpath.c mf_strip.c \ - mf_soundex.c mf_wcomp.c mf_wfile.c \ + mf_wcomp.c mf_wfile.c \ mf_qsort.c mf_qsort2.c mf_sort.c \ ptr_cmp.c mf_radix.c queues.c \ tree.c list.c hash.c array.c string.c typelib.c \ diff --git a/mysys/my_bitmap.c b/mysys/my_bitmap.c index 8834dda98e1..336bf27d559 100644 --- a/mysys/my_bitmap.c +++ b/mysys/my_bitmap.c @@ -16,7 +16,18 @@ /* Handling of uchar arrays as large bitmaps. - We assume that the size of the used bitmap is less than ~(uint) 0 + + API limitations (or, rather asserted safety assumptions, + to encourage correct programming) + + * the size of the used bitmap is less than ~(uint) 0 + * it's a multiple of 8 (for efficiency reasons) + * when arguments are a bitmap and a bit number, the number + must be within bitmap size + * bitmap_set_prefix() is an exception - one can use ~0 to set all bits + * when both arguments are bitmaps, they must be of the same size + * bitmap_intersect() is an exception :) + (for for Bitmap::intersect(ulonglong map2buff)) TODO: Make assembler THREAD safe versions of these using test-and-set instructions @@ -27,67 +38,73 @@ #include <assert.h> #include <m_string.h> -inline void bitmap_lock(MY_BITMAP* map) +inline void bitmap_lock(MY_BITMAP *map) { #ifdef THREAD - if (map->thread_safe) - pthread_mutex_lock(&map->mutex); + if (map->mutex) + pthread_mutex_lock(map->mutex); #endif } -inline void bitmap_unlock(MY_BITMAP* map) +inline void bitmap_unlock(MY_BITMAP *map) { #ifdef THREAD - if (map->thread_safe) - pthread_mutex_unlock(&map->mutex); + if (map->mutex) + pthread_mutex_unlock(map->mutex); #endif } -my_bool bitmap_init(MY_BITMAP *map, uint bitmap_size, my_bool thread_safe) +my_bool bitmap_init(MY_BITMAP *map, uchar *buf, uint bitmap_size, my_bool thread_safe) { - if (!(map->bitmap=(uchar*) my_malloc((bitmap_size+7)/8, - MYF(MY_WME | MY_ZEROFILL)))) + DBUG_ASSERT((bitmap_size & 7) == 0); + bitmap_size/=8; + if (!(map->bitmap=buf) && + !(map->bitmap=(uchar*)my_malloc(bitmap_size + + (thread_safe ? sizeof(pthread_mutex_t) : 0), + MYF(MY_WME | MY_ZEROFILL)))) return 1; - DBUG_ASSERT(bitmap_size != ~(uint) 0); + map->bitmap_size=bitmap_size; #ifdef THREAD - if ((map->thread_safe = thread_safe)) - pthread_mutex_init(&map->mutex, MY_MUTEX_INIT_FAST); + if (thread_safe) + { + map->mutex=(pthread_mutex_t *)(map->bitmap+bitmap_size); + pthread_mutex_init(map->mutex, MY_MUTEX_INIT_FAST); + } + else + map->mutex=0; #endif - map->bitmap_size=bitmap_size; return 0; } void bitmap_free(MY_BITMAP *map) { +#ifdef THREAD + if (map->mutex) + pthread_mutex_destroy(map->mutex); +#endif if (map->bitmap) { my_free((char*) map->bitmap, MYF(0)); map->bitmap=0; -#ifdef THREAD - if (map->thread_safe) - pthread_mutex_destroy(&map->mutex); -#endif } } void bitmap_set_bit(MY_BITMAP *map, uint bitmap_bit) { - if (bitmap_bit < map->bitmap_size) - { - bitmap_lock(map); - map->bitmap[bitmap_bit / 8] |= (1 << (bitmap_bit & 7)); - bitmap_unlock(map); - } + DBUG_ASSERT(map->bitmap && bitmap_bit < map->bitmap_size*8); + bitmap_lock(map); + map->bitmap[bitmap_bit / 8] |= (1 << (bitmap_bit & 7)); + bitmap_unlock(map); } - uint bitmap_set_next(MY_BITMAP *map) { uchar *bitmap=map->bitmap; uint bit_found = MY_BIT_NONE; - uint bitmap_size=map->bitmap_size; + uint bitmap_size=map->bitmap_size*8; uint i; + DBUG_ASSERT(map->bitmap); bitmap_lock(map); for (i=0; i < bitmap_size ; i++, bitmap++) { @@ -110,35 +127,180 @@ uint bitmap_set_next(MY_BITMAP *map) return bit_found; } - void bitmap_clear_bit(MY_BITMAP *map, uint bitmap_bit) { - if (bitmap_bit < map->bitmap_size) - { - bitmap_lock(map); - map->bitmap[bitmap_bit / 8] &= ~ (1 << (bitmap_bit & 7)); - bitmap_unlock(map); - } + DBUG_ASSERT(map->bitmap && bitmap_bit < map->bitmap_size*8); + bitmap_lock(map); + map->bitmap[bitmap_bit / 8] &= ~ (1 << (bitmap_bit & 7)); + bitmap_unlock(map); } +void bitmap_set_prefix(MY_BITMAP *map, uint prefix_size) +{ + uint prefix_bytes, prefix_bits; -void bitmap_set_all(MY_BITMAP* map) + DBUG_ASSERT(map->bitmap && + (prefix_size <= map->bitmap_size*8 || prefix_size == ~0)); + bitmap_lock(map); + set_if_smaller(prefix_size, map->bitmap_size*8); + if ((prefix_bytes= prefix_size / 8)) + memset(map->bitmap, 0xff, prefix_bytes); + if ((prefix_bits= prefix_size & 7)) + map->bitmap[prefix_bytes++]= (1 << prefix_bits)-1; + if (prefix_bytes < map->bitmap_size) + bzero(map->bitmap+prefix_bytes, map->bitmap_size-prefix_bytes); + bitmap_unlock(map); +} + +void bitmap_clear_all(MY_BITMAP *map) +{ + bitmap_set_prefix(map, 0); +} + +void bitmap_set_all(MY_BITMAP *map) +{ + bitmap_set_prefix(map, ~0); +} + +my_bool bitmap_is_prefix(const MY_BITMAP *map, uint prefix_size) { + uint prefix_bits= prefix_size & 7, res= 0; + uchar *m= map->bitmap, *end_prefix= map->bitmap+prefix_size/8, + *end= map->bitmap+map->bitmap_size; + + DBUG_ASSERT(map->bitmap && prefix_size <= map->bitmap_size*8); + bitmap_lock(map); - memset(map->bitmap, 0xff, (map->bitmap_size+7)/8); + while (m < end_prefix) + if (*m++ != 0xff) + goto ret; + + if (prefix_bits && *m++ != (1 << prefix_bits)-1) + goto ret; + + while (m < end) + if (*m++ != 0) + goto ret; + + res=1; +ret: bitmap_unlock(map); + return res; } -my_bool bitmap_is_set(MY_BITMAP* map, uint bitmap_bit) +my_bool bitmap_is_clear_all(const MY_BITMAP *map) { - return (bitmap_bit < map->bitmap_size) ? - (map->bitmap[bitmap_bit / 8] & (1 << (bitmap_bit & 7))) : - 0; + return bitmap_is_prefix(map, 0); } -void bitmap_clear_all(MY_BITMAP* map) +my_bool bitmap_is_set_all(const MY_BITMAP *map) { + return bitmap_is_prefix(map, map->bitmap_size*8); +} + +my_bool bitmap_is_set(const MY_BITMAP *map, uint bitmap_bit) +{ + DBUG_ASSERT(map->bitmap && bitmap_bit < map->bitmap_size*8); + return map->bitmap[bitmap_bit / 8] & (1 << (bitmap_bit & 7)); +} + +my_bool bitmap_is_subset(const MY_BITMAP *map1, const MY_BITMAP *map2) +{ + uint length, res=0; + uchar *m1=map1->bitmap, *m2=map2->bitmap, *end; + + DBUG_ASSERT(map1->bitmap && map2->bitmap && + map1->bitmap_size==map2->bitmap_size); + bitmap_lock(map1); + bitmap_lock(map2); + + end= m1+map1->bitmap_size; + + while (m1 < end) + if ((*m1++) & ~(*m2++)) + goto ret; + + res=1; +ret: + bitmap_unlock(map2); + bitmap_unlock(map1); + return res; +} + +my_bool bitmap_cmp(const MY_BITMAP *map1, const MY_BITMAP *map2) +{ + uint res; + + DBUG_ASSERT(map1->bitmap && map2->bitmap && + map1->bitmap_size==map2->bitmap_size); + bitmap_lock(map1); + bitmap_lock(map2); + + res= memcmp(map1->bitmap, map2->bitmap, map1->bitmap_size)==0; + + bitmap_unlock(map2); + bitmap_unlock(map1); + return res; +} + +void bitmap_intersect(MY_BITMAP *map, const MY_BITMAP *map2) +{ + uchar *to=map->bitmap, *from=map2->bitmap, *end; + uint len=map->bitmap_size, len2=map2->bitmap; + + DBUG_ASSERT(map->bitmap && map2->bitmap); + bitmap_lock(map); + bitmap_lock(map2); + + end= to+min(len,len2); + + while (to < end) + *to++ &= *from++; + + if (len2 < len) + { + end+=len-len2; + while (to < end) + *to++=0; + } + + bitmap_unlock(map2); + bitmap_unlock(map); +} + +void bitmap_subtract(MY_BITMAP *map, const MY_BITMAP *map2) +{ + uchar *to=map->bitmap, *from=map2->bitmap, *end; + + DBUG_ASSERT(map->bitmap && map2->bitmap && + map->bitmap_size==map2->bitmap_size); + bitmap_lock(map); + bitmap_lock(map2); + + end= to+map->bitmap_size; + + while (to < end) + *to++ &= ~(*from++); + + bitmap_unlock(map2); + bitmap_unlock(map); +} + +void bitmap_union(MY_BITMAP *map, const MY_BITMAP *map2) +{ + uchar *to=map->bitmap, *from=map2->bitmap, *end; + + DBUG_ASSERT(map->bitmap && map2->bitmap && + map->bitmap_size==map2->bitmap_size); bitmap_lock(map); - bzero(map->bitmap,(map->bitmap_size+7)/8); + bitmap_lock(map2); + + end= to+map->bitmap_size; + + while (to < end) + *to++ |= *from++; + + bitmap_unlock(map2); bitmap_unlock(map); } + diff --git a/sql/field.cc b/sql/field.cc index b025f65a798..c30c06c6395 100644 --- a/sql/field.cc +++ b/sql/field.cc @@ -181,7 +181,7 @@ Field::Field(char *ptr_arg,uint32 length_arg,uchar *null_ptr_arg, :ptr(ptr_arg),null_ptr(null_ptr_arg), table(table_arg),table_name(table_arg ? table_arg->table_name : 0), field_name(field_name_arg), - query_id(0),key_start(0),part_of_key(0),part_of_sortkey(0), + query_id(0), key_start(0), part_of_key(0), part_of_sortkey(0), unireg_check(unireg_check_arg), field_length(length_arg),null_bit(null_bit_arg),abs_offset(0) { diff --git a/sql/field.h b/sql/field.h index fe5141e9d80..e7d4bb2416a 100644 --- a/sql/field.h +++ b/sql/field.h @@ -151,7 +151,9 @@ public: if (tmp->table->maybe_null) tmp->flags&= ~NOT_NULL_FLAG; tmp->table= new_table; - tmp->key_start= tmp->part_of_key= tmp->part_of_sortkey= 0; + tmp->key_start.init(0); + tmp->part_of_key.init(0); + tmp->part_of_sortkey.init(0); tmp->unireg_check=Field::NONE; tmp->flags&= (NOT_NULL_FLAG | BLOB_FLAG | UNSIGNED_FLAG | ZEROFILL_FLAG | BINARY_FLAG | ENUM_FLAG | SET_FLAG); diff --git a/sql/ha_berkeley.cc b/sql/ha_berkeley.cc index 5b0fc95442c..e87e31708f3 100644 --- a/sql/ha_berkeley.cc +++ b/sql/ha_berkeley.cc @@ -844,7 +844,7 @@ int ha_berkeley::write_row(byte * record) ulong thd_options = table->in_use ? table->in_use->options : 0; for (uint retry=0 ; retry < berkeley_trans_retry ; retry++) { - key_map changed_keys = 0; + key_map changed_keys(0); if (using_ignore && (thd_options & OPTION_INTERNAL_SUBTRANSACTIONS)) { if ((error=txn_begin(db_env, transaction, &sub_trans, 0))) /* purecov: deadcode */ @@ -855,7 +855,7 @@ int ha_berkeley::write_row(byte * record) key_buff, record), &row, key_type[primary_key]))) { - changed_keys |= (key_map) 1 << primary_key; + changed_keys.set_bit(primary_key); for (uint keynr=0 ; keynr < table->keys ; keynr++) { if (keynr == primary_key) @@ -868,7 +868,7 @@ int ha_berkeley::write_row(byte * record) last_dup_key=keynr; break; } - changed_keys |= (key_map) 1 << keynr; + changed_keys.set_bit(keynr); } } else @@ -885,12 +885,13 @@ int ha_berkeley::write_row(byte * record) DBUG_PRINT("trans",("aborting subtransaction")); /* purecov: deadcode */ new_error=txn_abort(sub_trans); /* purecov: deadcode */ } - else if (changed_keys) + else if (!changed_keys.is_clear_all()) { new_error = 0; - for (uint keynr=0; changed_keys; keynr++, changed_keys >>= 1) + for (uint keynr=0 ; keynr < table->keys+test(hidden_primary_key) ; + keynr++) { - if (changed_keys & 1) + if (changed_keys.is_set(keynr)) { if ((new_error = remove_key(sub_trans, keynr, record, &prim_key))) @@ -1012,7 +1013,7 @@ int ha_berkeley::update_primary_key(DB_TXN *trans, bool primary_key_changed, Clobbers keybuff2 */ -int ha_berkeley::restore_keys(DB_TXN *trans, key_map changed_keys, +int ha_berkeley::restore_keys(DB_TXN *trans, key_map *changed_keys, uint primary_key, const byte *old_row, DBT *old_key, const byte *new_row, DBT *new_key, @@ -1034,18 +1035,21 @@ int ha_berkeley::restore_keys(DB_TXN *trans, key_map changed_keys, rolled back. The last key set in changed_keys is the one that triggered the duplicate key error (it wasn't inserted), so for that one just put back the old value. */ - for (keynr=0; changed_keys; keynr++, changed_keys >>= 1) + if (!changed_keys->is_clear_all()) { - if (changed_keys & 1) + for (keynr=0 ; keynr < table->keys+test(hidden_primary_key) ; keynr++) { - if (changed_keys != 1 && - (error = remove_key(trans, keynr, new_row, new_key))) - break; /* purecov: inspected */ - if ((error = key_file[keynr]->put(key_file[keynr], trans, - create_key(&tmp_key, keynr, key_buff2, - old_row), - old_key, key_type[keynr]))) - break; /* purecov: inspected */ + if (changed_keys->is_set(keynr)) + { + if (changed_keys->is_prefix(1) && + (error = remove_key(trans, keynr, new_row, new_key))) + break; /* purecov: inspected */ + if ((error = key_file[keynr]->put(key_file[keynr], trans, + create_key(&tmp_key, keynr, key_buff2, + old_row), + old_key, key_type[keynr]))) + break; /* purecov: inspected */ + } } } @@ -1090,7 +1094,7 @@ int ha_berkeley::update_row(const byte * old_row, byte * new_row) sub_trans = transaction; for (uint retry=0 ; retry < berkeley_trans_retry ; retry++) { - key_map changed_keys = 0; + key_map changed_keys(0); if (using_ignore && (thd_options & OPTION_INTERNAL_SUBTRANSACTIONS)) { if ((error=txn_begin(db_env, transaction, &sub_trans, 0))) /* purecov: deadcode */ @@ -1123,7 +1127,7 @@ int ha_berkeley::update_row(const byte * old_row, byte * new_row) } DBUG_RETURN(error); // Fatal error /* purecov: inspected */ } - changed_keys |= (key_map)1 << keynr; + changed_keys.set_bit(keynr); if ((error=key_file[keynr]->put(key_file[keynr], sub_trans, create_key(&key, keynr, key_buff2, new_row), @@ -1147,8 +1151,8 @@ int ha_berkeley::update_row(const byte * old_row, byte * new_row) DBUG_PRINT("trans",("aborting subtransaction")); /* purecov: deadcode */ new_error=txn_abort(sub_trans); /* purecov: deadcode */ } - else if (changed_keys) - new_error=restore_keys(transaction, changed_keys, primary_key, + else if (!changed_keys.is_clear_all()) + new_error=restore_keys(transaction, &changed_keys, primary_key, old_row, &old_prim_key, new_row, &prim_key, thd_options); if (new_error) @@ -1229,12 +1233,12 @@ int ha_berkeley::remove_key(DB_TXN *trans, uint keynr, const byte *record, /* Delete all keys for new_record */ int ha_berkeley::remove_keys(DB_TXN *trans, const byte *record, - DBT *new_record, DBT *prim_key, key_map keys) + DBT *new_record, DBT *prim_key, key_map *keys) { int result = 0; - for (uint keynr=0; keys; keynr++, keys>>=1) + for (uint keynr=0 ; keynr < table->keys+test(hidden_primary_key) ; keynr++) { - if (keys & 1) + if (keys->is_set(keynr)) { int new_error=remove_key(trans, keynr, record, prim_key); if (new_error) @@ -1261,7 +1265,7 @@ int ha_berkeley::delete_row(const byte * record) DBUG_RETURN((error)); /* purecov: inspected */ create_key(&prim_key, primary_key, key_buff, record); if (hidden_primary_key) - keys|= (key_map) 1 << primary_key; + keys.set_bit(primary_key); /* Subtransactions may be used in order to retry the delete in case we get a DB_LOCK_DEADLOCK error. */ @@ -1274,7 +1278,7 @@ int ha_berkeley::delete_row(const byte * record) break; /* purecov: deadcode */ DBUG_PRINT("trans",("starting sub transaction")); /* purecov: deadcode */ } - error=remove_keys(sub_trans, record, &row, &prim_key, keys); + error=remove_keys(sub_trans, record, &row, &prim_key, &keys); if (!error && (thd_options & OPTION_INTERNAL_SUBTRANSACTIONS)) { DBUG_PRINT("trans",("ending sub transaction")); /* purecov: deadcode */ diff --git a/sql/ha_berkeley.h b/sql/ha_berkeley.h index f1669e9b6c7..39293717c27 100644 --- a/sql/ha_berkeley.h +++ b/sql/ha_berkeley.h @@ -72,8 +72,8 @@ class ha_berkeley: public handler uint key_length); int remove_key(DB_TXN *trans, uint keynr, const byte *record, DBT *prim_key); int remove_keys(DB_TXN *trans,const byte *record, DBT *new_record, - DBT *prim_key, key_map keys); - int restore_keys(DB_TXN *trans, key_map changed_keys, uint primary_key, + DBT *prim_key, key_map *keys); + int restore_keys(DB_TXN *trans, key_map *changed_keys, uint primary_key, const byte *old_row, DBT *old_key, const byte *new_row, DBT *new_key, ulong thd_options); @@ -107,7 +107,7 @@ class ha_berkeley: public handler uint extra_rec_buf_length() { return BDB_HIDDEN_PRIMARY_KEY_LENGTH; } ha_rows estimate_number_of_rows(); bool fast_key_read() { return 1;} - key_map keys_to_use_for_scanning() { return ~(key_map) 0; } + const key_map *keys_to_use_for_scanning() { return &key_map_full; } bool has_transactions() { return 1;} int open(const char *name, int mode, uint test_if_locked); diff --git a/sql/ha_innodb.h b/sql/ha_innodb.h index 0c89a9d29ce..2fa70f27128 100644 --- a/sql/ha_innodb.h +++ b/sql/ha_innodb.h @@ -125,7 +125,7 @@ class ha_innobase: public handler uint max_key_length() const { return((MAX_KEY_LENGTH <= 3500) ? MAX_KEY_LENGTH : 3500);} bool fast_key_read() { return 1;} - key_map keys_to_use_for_scanning() { return ~(key_map) 0; } + const key_map *keys_to_use_for_scanning() { return &key_map_full; } bool has_transactions() { return 1;} int open(const char *name, int mode, uint test_if_locked); diff --git a/sql/ha_isam.cc b/sql/ha_isam.cc index 3acc385d251..8025e5169c8 100644 --- a/sql/ha_isam.cc +++ b/sql/ha_isam.cc @@ -202,7 +202,7 @@ void ha_isam::info(uint flag) sortkey = info.sortkey; block_size=nisam_block_size; table->keys = min(table->keys,info.keys); - table->keys_in_use= set_bits(key_map,table->keys); + table->keys_in_use.set_prefix(table->keys); table->db_options_in_use= info.options; table->db_record_offset= (table->db_options_in_use & diff --git a/sql/ha_myisam.cc b/sql/ha_myisam.cc index aa889b649d0..6c46fc141f3 100644 --- a/sql/ha_myisam.cc +++ b/sql/ha_myisam.cc @@ -836,20 +836,21 @@ int ha_myisam::preload_keys(THD* thd, HA_CHECK_OPT *check_opt) DBUG_ENTER("ha_myisam::preload_keys"); - /* Check validity of the index references */ + /* Check validity of the index references */ if (table_list->use_index) { - key_map kmap= get_key_map_from_key_list(table, table_list->use_index); - if (kmap == ~(key_map) 0) + key_map kmap; + get_key_map_from_key_list(&kmap, table, table_list->use_index); + if (kmap.is_set_all()) { errmsg= thd->net.last_error; error= HA_ADMIN_FAILED; goto err; } - if (kmap) - map= kmap; + if (!kmap.is_clear_all()) + map= kmap.to_ulonglong(); } - + mi_extra(file, HA_EXTRA_PRELOAD_BUFFER_SIZE, (void *) &thd->variables.preload_buff_size); @@ -862,16 +863,16 @@ int ha_myisam::preload_keys(THD* thd, HA_CHECK_OPT *check_opt) case HA_ERR_OUT_OF_MEM: errmsg= "Failed to allocate buffer"; break; - default: + default: char buf[ERRMSGSIZE+20]; - my_snprintf(buf, ERRMSGSIZE, + my_snprintf(buf, ERRMSGSIZE, "Failed to read from index file (errno: %d)", my_errno); errmsg= buf; } error= HA_ADMIN_FAILED; goto err; } - + DBUG_RETURN(HA_ADMIN_OK); err: @@ -1153,9 +1154,10 @@ void ha_myisam::info(uint flag) ref_length=info.reflength; table->db_options_in_use = info.options; block_size=myisam_block_size; - table->keys_in_use= (set_bits(key_map, table->keys) & - (key_map) info.key_map); - table->keys_for_keyread= table->keys_in_use & ~table->read_only_keys; + table->keys_in_use.set_prefix(table->keys); + table->keys_in_use.intersect(info.key_map); + table->keys_for_keyread= table->keys_in_use; + table->keys_for_keyread.subtract(table->read_only_keys); table->db_record_offset=info.record_offset; if (table->key_parts) memcpy((char*) table->key_info[0].rec_per_key, diff --git a/sql/ha_myisam.h b/sql/ha_myisam.h index e8413ef0bd2..e6297373fea 100644 --- a/sql/ha_myisam.h +++ b/sql/ha_myisam.h @@ -89,9 +89,8 @@ class ha_myisam: public handler ft_handler->please->reinit_search(ft_handler); return 0; } - FT_INFO *ft_init_ext(uint mode, uint inx,const byte *key, uint keylen, - bool presort) - { return ft_init_search(mode, file,inx,(byte*) key,keylen,presort); } + FT_INFO *ft_init_ext(uint flags, uint inx,const byte *key, uint keylen) + { return ft_init_search(flags,file,inx,(byte*) key,keylen, table->record[0]); } int ft_read(byte *buf); int rnd_init(bool scan=1); int rnd_next(byte *buf); diff --git a/sql/ha_myisammrg.cc b/sql/ha_myisammrg.cc index d6d4f58b31c..05da214c1d4 100644 --- a/sql/ha_myisammrg.cc +++ b/sql/ha_myisammrg.cc @@ -234,7 +234,7 @@ void ha_myisammrg::info(uint flag) #endif data_file_length=info.data_file_length; errkey = info.errkey; - table->keys_in_use= set_bits(key_map, table->keys); + table->keys_in_use.set_prefix(table->keys); table->db_options_in_use = info.options; table->is_view=1; mean_rec_length=info.reclength; diff --git a/sql/handler.h b/sql/handler.h index 8f0f2ef4e55..c24bf777f01 100644 --- a/sql/handler.h +++ b/sql/handler.h @@ -241,7 +241,7 @@ public: virtual double read_time(uint index, uint ranges, ha_rows rows) { return rows2double(ranges+rows); } virtual bool fast_key_read() { return 0;} - virtual key_map keys_to_use_for_scanning() { return 0; } + virtual const key_map *keys_to_use_for_scanning() { return &key_map_empty; } virtual bool has_transactions(){ return 0;} virtual uint extra_rec_buf_length() { return 0; } virtual ha_rows estimate_number_of_rows() { return records+EXTRA_RECORDS; } @@ -271,8 +271,7 @@ public: } virtual int ft_init() { return -1; } - virtual FT_INFO *ft_init_ext(uint mode,uint inx,const byte *key, uint keylen, - bool presort) + virtual FT_INFO *ft_init_ext(uint flags,uint inx,const byte *key, uint keylen) { return NULL; } virtual int ft_read(byte *buf) { return -1; } virtual int rnd_init(bool scan=1)=0; diff --git a/sql/item.cc b/sql/item.cc index 0462a78376f..7c7cad7c290 100644 --- a/sql/item.cc +++ b/sql/item.cc @@ -801,11 +801,11 @@ bool Item_field::fix_fields(THD *thd, TABLE_LIST *tables, Item **ref) not_found_field) { /* - We can't find table field in table list of current select, + We can't find table field in table list of current select, consequently we have to find it in outer subselect(s). - We can't join lists of outer & current select, because of scope - of view rules. For example if both tables (outer & current) have - field 'field' it is not mistake to refer to this field without + We can't join lists of outer & current select, because of scope + of view rules. For example if both tables (outer & current) have + field 'field' it is not mistake to refer to this field without mention of table name, but if we join tables in one list it will cause error ER_NON_UNIQ_ERROR in find_field_in_tables. */ @@ -842,8 +842,8 @@ bool Item_field::fix_fields(THD *thd, TABLE_LIST *tables, Item **ref) break; } if (sl->resolve_mode == SELECT_LEX::SELECT_MODE && - (refer= find_item_in_list(this, sl->item_list, &counter, - REPORT_EXCEPT_NOT_FOUND)) != + (refer= find_item_in_list(this, sl->item_list, &counter, + REPORT_EXCEPT_NOT_FOUND)) != (Item **) not_found_item) { if (*refer && (*refer)->fixed) // Avoid crash in case of error @@ -920,7 +920,7 @@ bool Item_field::fix_fields(THD *thd, TABLE_LIST *tables, Item **ref) TABLE *table=field->table; field->query_id=thd->query_id; table->used_fields++; - table->used_keys&=field->part_of_key; + table->used_keys.intersect(field->part_of_key); } fixed= 1; return 0; @@ -929,14 +929,14 @@ bool Item_field::fix_fields(THD *thd, TABLE_LIST *tables, Item **ref) void Item::init_make_field(Send_field *tmp_field, enum enum_field_types field_type) -{ +{ char *empty_name= (char*) ""; - tmp_field->db_name= empty_name; + tmp_field->db_name= empty_name; tmp_field->org_table_name= empty_name; tmp_field->org_col_name= empty_name; tmp_field->table_name= empty_name; tmp_field->col_name= name; - tmp_field->charsetnr= collation.collation->number; + tmp_field->charsetnr= collation.collation->number; tmp_field->flags=maybe_null ? 0 : NOT_NULL_FLAG; tmp_field->type=field_type; tmp_field->length=max_length; diff --git a/sql/item_cmpfunc.h b/sql/item_cmpfunc.h index 0522c1bcde2..28ba6199bee 100644 --- a/sql/item_cmpfunc.h +++ b/sql/item_cmpfunc.h @@ -647,7 +647,6 @@ class Item_func_in :public Item_int_func ~Item_func_in() { delete array; delete in_item; } optimize_type select_optimize() const { return array ? OPTIMIZE_KEY : OPTIMIZE_NONE; } - Item *key_item() const { return args[0]; } void print(String *str); enum Functype functype() const { return IN_FUNC; } const char *func_name() const { return " IN "; } diff --git a/sql/item_func.cc b/sql/item_func.cc index ce0614f1e7c..b47c833d404 100644 --- a/sql/item_func.cc +++ b/sql/item_func.cc @@ -81,12 +81,12 @@ bool Item_func::agg_arg_collations(DTCollation &c, Item **av, uint count) } -bool Item_func::agg_arg_collations_for_comparison(DTCollation &c, +bool Item_func::agg_arg_collations_for_comparison(DTCollation &c, Item **av, uint count) { if (agg_arg_collations(c, av, count)) return TRUE; - + if (c.derivation == DERIVATION_NONE) { my_coll_agg_error(av, count, func_name()); @@ -211,7 +211,7 @@ Item_func::fix_fields(THD *thd, TABLE_LIST *tables, Item **ref) item= *arg; if (item->maybe_null) maybe_null=1; - + with_sum_func= with_sum_func || item->with_sum_func; used_tables_cache|= item->used_tables(); not_null_tables_cache|= item->not_null_tables(); @@ -2603,9 +2603,13 @@ void Item_func_match::init_search(bool no_order) DBUG_VOID_RETURN; if (key == NO_SUCH_KEY) + { + List<Item> fields; + for (uint i=1; i < arg_count; i++) + fields.push_back(args[i]); concat=new Item_func_concat_ws(new Item_string(" ",1, - default_charset_info), - fields); + cmp_collation.collation), fields); + } if (master) { @@ -2617,20 +2621,25 @@ void Item_func_match::init_search(bool no_order) } String *ft_tmp= 0; - char tmp1[FT_QUERY_MAXLEN]; - String tmp2(tmp1,sizeof(tmp1),default_charset_info); // MATCH ... AGAINST (NULL) is meaningless, but possible - if (!(ft_tmp=key_item()->val_str(&tmp2))) + if (!(ft_tmp=key_item()->val_str(&value))) { - ft_tmp= &tmp2; - tmp2.set("",0,default_charset_info); + ft_tmp= &value; + value.set("",0,cmp_collation.collation); } - ft_handler=table->file->ft_init_ext(mode, key, + if (ft_tmp->charset() != cmp_collation.collation) + { + search_value.copy(ft_tmp->ptr(), ft_tmp->length(), ft_tmp->charset(), + cmp_collation.collation); + ft_tmp= &search_value; + } + + if (join_key && !no_order) flags|=FT_SORTED; + ft_handler=table->file->ft_init_ext(flags, key, (byte*) ft_tmp->ptr(), - ft_tmp->length(), - join_key && !no_order); + ft_tmp->length()); if (join_key) table->file->ft_handler=ft_handler; @@ -2641,7 +2650,6 @@ void Item_func_match::init_search(bool no_order) bool Item_func_match::fix_fields(THD *thd, TABLE_LIST *tlist, Item **ref) { - List_iterator<Item> li(fields); Item *item; maybe_null=1; @@ -2653,51 +2661,37 @@ bool Item_func_match::fix_fields(THD *thd, TABLE_LIST *tlist, Item **ref) modifications to find_best and auto_close as complement to auto_init code above. */ - if (Item_func::fix_fields(thd, tlist, ref) || !const_item()) + if (Item_func::fix_fields(thd, tlist, ref) || !args[0]->const_item()) { my_error(ER_WRONG_ARGUMENTS,MYF(0),"AGAINST"); return 1; } - while ((item=li++)) + const_item_cache=0; + for (uint i=1 ; i < arg_count ; i++) { - if (item->fix_fields(thd, tlist, li.ref()) || item->check_cols(1)) - return 1; + item=args[i]; if (item->type() == Item::REF_ITEM) - li.replace(item= *((Item_ref *)item)->ref); - if (item->type() != Item::FIELD_ITEM || !item->used_tables()) + args[i]= item= *((Item_ref *)item)->ref; + if (item->type() != Item::FIELD_ITEM) key=NO_SUCH_KEY; used_tables_cache|=item->used_tables(); } /* check that all columns come from the same table */ if (my_count_bits(used_tables_cache) != 1) key=NO_SUCH_KEY; - const_item_cache=0; - table=((Item_field *)fields.head())->field->table; - table->fulltext_searched=1; - record=table->record[0]; - if (key == NO_SUCH_KEY && mode != FT_BOOL) + if (key == NO_SUCH_KEY && !(flags & FT_BOOL)) { my_error(ER_WRONG_ARGUMENTS,MYF(0),"MATCH"); return 1; } - - return 0; -} - -bool Item_func_match::walk(Item_processor processor, byte *arg) -{ - List_iterator_fast<Item> li(fields); - Item *item; - while ((item= li++)) - if (item->walk(processor, arg)) - return 1; - return Item_func::walk(processor, arg); + table=((Item_field *)item)->field->table; + table->fulltext_searched=1; + return agg_arg_collations_for_comparison(cmp_collation, args+1, arg_count-1); } bool Item_func_match::fix_index() { - List_iterator_fast<Item> li(fields); Item_field *item; uint ft_to_key[MAX_KEY], ft_cnt[MAX_KEY], fts=0, keynr; uint max_cnt=0, mkeys=0; @@ -2708,7 +2702,7 @@ bool Item_func_match::fix_index() for (keynr=0 ; keynr < table->keys ; keynr++) { if ((table->key_info[keynr].flags & HA_FULLTEXT) && - (table->keys_in_use_for_query & (((key_map)1) << keynr))) + (table->keys_in_use_for_query.is_set(keynr))) { ft_to_key[fts]=keynr; ft_cnt[fts]=0; @@ -2719,8 +2713,9 @@ bool Item_func_match::fix_index() if (!fts) goto err; - while ((item=(Item_field*)(li++))) + for (uint i=1; i < arg_count; i++) { + item=(Item_field*)args[i]; for (keynr=0 ; keynr < fts ; keynr++) { KEY *ft_key=&table->key_info[ft_to_key[keynr]]; @@ -2754,8 +2749,8 @@ bool Item_func_match::fix_index() for (keynr=0 ; keynr <= mkeys ; keynr++) { - // for now, partial keys won't work. SerG - if (max_cnt < fields.elements || + // partial keys doesn't work + if (max_cnt < arg_count-1 || max_cnt < table->key_info[ft_to_key[keynr]].key_parts) continue; @@ -2765,21 +2760,20 @@ bool Item_func_match::fix_index() } err: - if (mode == FT_BOOL) + if (flags & FT_BOOL) { key=NO_SUCH_KEY; return 0; } - my_printf_error(ER_FT_MATCHING_KEY_NOT_FOUND, - ER(ER_FT_MATCHING_KEY_NOT_FOUND),MYF(0)); + my_error(ER_FT_MATCHING_KEY_NOT_FOUND,MYF(0)); return 1; } bool Item_func_match::eq(const Item *item, bool binary_cmp) const { - if (item->type() != FUNC_ITEM || - func_name() != ((Item_func*)item)->func_name()) + if (item->type() != FUNC_ITEM || ((Item_func*)item)->functype() != FT_FUNC || + flags != ((Item_func_match*)item)->flags) return 0; Item_func_match *ifm=(Item_func_match*) item; @@ -2817,7 +2811,8 @@ double Item_func_match::val() (byte *)a->ptr(), a->length())); } else - DBUG_RETURN(ft_handler->please->find_relevance(ft_handler, record, 0)); + DBUG_RETURN(ft_handler->please->find_relevance(ft_handler, + table->record[0], 0)); } diff --git a/sql/item_func.h b/sql/item_func.h index fe9c2645216..9a6e2a71f64 100644 --- a/sql/item_func.h +++ b/sql/item_func.h @@ -971,20 +971,18 @@ public: class Item_func_match :public Item_real_func { public: - List<Item> fields; - String value; - TABLE *table; - Item_func_match *master; - FT_INFO * ft_handler; - Item *concat; - byte *record; - uint key, mode; + uint key, flags; bool join_key; + DTCollation cmp_collation; + FT_INFO *ft_handler; + TABLE *table; + Item_func_match *master; // for master-slave optimization + Item *concat; // Item_func_concat_ws + String value; // value of concat + String search_value; // key_item()'s value converted to cmp_collation - Item_func_match(List<Item> &a, Item *b): Item_real_func(b), - fields(a), table(0), master(0), ft_handler(0), - concat(0), key(0), join_key(0) - {} + Item_func_match(List<Item> &a, uint b): Item_real_func(a), key(0), flags(b), + join_key(0), ft_handler(0), table(0), master(0), concat(0) { } ~Item_func_match() { if (!master && ft_handler) @@ -999,6 +997,7 @@ public: delete concat; } enum Functype functype() const { return FT_FUNC; } + const char *func_name() const { return "match"; } void update_used_tables() {} table_map not_null_tables() const { return 0; } bool fix_fields(THD *thd, struct st_table_list *tlist, Item **ref); @@ -1008,28 +1007,6 @@ public: bool fix_index(); void init_search(bool no_order); - - bool walk(Item_processor processor, byte *arg); -}; - - -class Item_func_match_nl :public Item_func_match -{ -public: - Item_func_match_nl(List<Item> &a, Item *b) - :Item_func_match(a,b) - { mode=FT_NL; } - const char *func_name() const { return "match_nl"; } -}; - - -class Item_func_match_bool :public Item_func_match -{ -public: - Item_func_match_bool(List<Item> &a, Item *b) - :Item_func_match(a,b) - { mode=FT_BOOL; } - const char *func_name() const { return "match_bool"; } }; diff --git a/sql/lex.h b/sql/lex.h index 5d79e378d4f..7f3328fa7cb 100644 --- a/sql/lex.h +++ b/sql/lex.h @@ -153,6 +153,7 @@ static SYMBOL symbols[] = { { "EXECUTE", SYM(EXECUTE_SYM),0,0}, { "EXPLAIN", SYM(DESCRIBE),0,0}, { "EXISTS", SYM(EXISTS),0,0}, + { "EXPANSION", SYM(EXPANSION_SYM),0,0}, { "EXTENDED", SYM(EXTENDED_SYM),0,0}, { "FAST", SYM(FAST_SYM),0,0}, { "FIELDS", SYM(COLUMNS),0,0}, diff --git a/sql/mysql_priv.h b/sql/mysql_priv.h index ce5c9ff0924..f7b12bd959c 100644 --- a/sql/mysql_priv.h +++ b/sql/mysql_priv.h @@ -23,16 +23,21 @@ #include <signal.h> #include <thr_lock.h> #include <my_base.h> /* Needed by field.h */ -#include <my_bitmap.h> +#include <sql_bitmap.h> #include <assert.h> #ifdef __EMX__ #undef write /* remove pthread.h macro definition for EMX */ #endif -typedef ulonglong table_map; /* Used for table bits in join */ -typedef ulong key_map; /* Used for finding keys */ -typedef ulong key_part_map; /* Used for finding key parts */ +/* TODO convert all these three maps to Bitmap classes */ +typedef ulonglong table_map; /* Used for table bits in join */ +typedef Bitmap<64> key_map; /* Used for finding keys */ +typedef ulong key_part_map; /* Used for finding key parts */ + +/* useful constants */ +extern const key_map key_map_empty; +extern const key_map key_map_full; #include "mysql_com.h" #include <violite.h> @@ -654,8 +659,8 @@ enum find_item_error_report_type {REPORT_ALL_ERRORS, REPORT_EXCEPT_NOT_FOUND, extern const Item **not_found_item; Item ** find_item_in_list(Item *item, List<Item> &items, uint *counter, find_item_error_report_type report_error); -key_map get_key_map_from_key_list(TABLE *table, - List<String> *index_list); +void get_key_map_from_key_list(key_map *map, TABLE *table, + List<String> *index_list); bool insert_fields(THD *thd,TABLE_LIST *tables, const char *db_name, const char *table_name, List_iterator<Item> *it); @@ -663,7 +668,7 @@ bool setup_tables(TABLE_LIST *tables); int setup_wild(THD *thd, TABLE_LIST *tables, List<Item> &fields, List<Item> *sum_func_list, uint wild_num); int setup_fields(THD *thd, Item** ref_pointer_array, TABLE_LIST *tables, - List<Item> &item, bool set_query_id, + List<Item> &item, bool set_query_id, List<Item> *sum_func_list, bool allow_sum_func); void unfix_item_list(List<Item> item_list); int setup_conds(THD *thd,TABLE_LIST *tables,COND **conds); diff --git a/sql/mysqld.cc b/sql/mysqld.cc index 8b79431b823..fafc67e059d 100644 --- a/sql/mysqld.cc +++ b/sql/mysqld.cc @@ -2164,7 +2164,7 @@ static int init_common_variables(const char *conf_file_name, int argc, init_global_datetime_format(DATETIME_FORMAT_TYPE, 1)) return 1; - if (use_temp_pool && bitmap_init(&temp_pool,1024,1)) + if (use_temp_pool && bitmap_init(&temp_pool,0,1024,1)) return 1; return 0; } @@ -3579,7 +3579,7 @@ enum options_mysqld OPT_CONNECT_TIMEOUT, OPT_DELAYED_INSERT_TIMEOUT, OPT_DELAYED_INSERT_LIMIT, OPT_DELAYED_QUEUE_SIZE, OPT_FLUSH_TIME, OPT_FT_MIN_WORD_LEN, - OPT_FT_MAX_WORD_LEN, OPT_FT_MAX_WORD_LEN_FOR_SORT, OPT_FT_STOPWORD_FILE, + OPT_FT_MAX_WORD_LEN, OPT_FT_QUERY_EXPANSION_LIMIT, OPT_FT_STOPWORD_FILE, OPT_INTERACTIVE_TIMEOUT, OPT_JOIN_BUFF_SIZE, OPT_KEY_BUFFER_SIZE, OPT_KEY_CACHE_BLOCK_SIZE, OPT_KEY_CACHE_DIVISION_LIMIT, OPT_KEY_CACHE_AGE_THRESHOLD, @@ -4223,10 +4223,10 @@ replicating a LOAD DATA INFILE command.", "The maximum length of the word to be included in a FULLTEXT index. Note: FULLTEXT indexes must be rebuilt after changing this variable.", (gptr*) &ft_max_word_len, (gptr*) &ft_max_word_len, 0, GET_ULONG, REQUIRED_ARG, HA_FT_MAXLEN, 10, HA_FT_MAXLEN, 0, 1, 0}, - { "ft_max_word_len_for_sort", OPT_FT_MAX_WORD_LEN_FOR_SORT, - "The maximum length of the word for repair_by_sorting. Longer words are included the slow way. The lower this value, the more words will be put in one sort bucket.", - (gptr*) &ft_max_word_len_for_sort, (gptr*) &ft_max_word_len_for_sort, 0, GET_ULONG, - REQUIRED_ARG, 20, 4, HA_FT_MAXLEN, 0, 1, 0}, + { "ft_query_expansion_limit", OPT_FT_QUERY_EXPANSION_LIMIT, + "Number of best matches to use for query expansion", + (gptr*) &ft_query_expansion_limit, (gptr*) &ft_query_expansion_limit, 0, GET_ULONG, + REQUIRED_ARG, 20, 0, 1000, 0, 1, 0}, { "ft_stopword_file", OPT_FT_STOPWORD_FILE, "Use stopwords from this file instead of built-in list.", (gptr*) &ft_stopword_file, (gptr*) &ft_stopword_file, 0, GET_STR, diff --git a/sql/opt_range.cc b/sql/opt_range.cc index 0ceeda9f5ef..e2761832e65 100644 --- a/sql/opt_range.cc +++ b/sql/opt_range.cc @@ -305,7 +305,7 @@ static ha_rows check_quick_keys(PARAM *param,uint index,SEL_ARG *key_tree, static QUICK_SELECT *get_quick_select(PARAM *param,uint index, SEL_ARG *key_tree); #ifndef DBUG_OFF -static void print_quick(QUICK_SELECT *quick,key_map needed_reg); +static void print_quick(QUICK_SELECT *quick,const key_map* needed_reg); #endif static SEL_TREE *tree_and(PARAM *param,SEL_TREE *tree1,SEL_TREE *tree2); static SEL_TREE *tree_or(PARAM *param,SEL_TREE *tree1,SEL_TREE *tree2); @@ -364,7 +364,7 @@ SQL_SELECT *make_select(TABLE *head, table_map const_tables, SQL_SELECT::SQL_SELECT() :quick(0),cond(0),free_cond(0) { - quick_keys=0; needed_reg=0; + quick_keys.clear_all(); needed_reg.clear_all(); my_b_clear(&file); } @@ -567,19 +567,29 @@ SEL_ARG *SEL_ARG::clone_tree() return root; } -/***************************************************************************** -** Test if a key can be used in different ranges -** Returns: -** -1 if impossible select -** 0 if can't use quick_select -** 1 if found usable range -** Updates the following in the select parameter: -** needed_reg ; Bits for keys with may be used if all prev regs are read -** quick ; Parameter to use when reading records. -** In the table struct the following information is updated: -** quick_keys ; Which keys can be used -** quick_rows ; How many rows the key matches -*****************************************************************************/ +/* + Test if a key can be used in different ranges + + SYNOPSIS + SQL_SELECT::test_quick_select(thd,keys_to_use, prev_tables, + limit, force_quick_range) + + Updates the following in the select parameter: + needed_reg - Bits for keys with may be used if all prev regs are read + quick - Parameter to use when reading records. + In the table struct the following information is updated: + quick_keys - Which keys can be used + quick_rows - How many rows the key matches + + RETURN VALUES + -1 if impossible select + 0 if can't use quick_select + 1 if found usable range + + TODO + check if the function really needs to modify keys_to_use, and change the + code to pass it by reference if not +*/ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, table_map prev_tables, @@ -590,17 +600,17 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, double scan_time; DBUG_ENTER("test_quick_select"); DBUG_PRINT("enter",("keys_to_use: %lu prev_tables: %lu const_tables: %lu", - (ulong) keys_to_use, (ulong) prev_tables, + keys_to_use.to_ulonglong(), (ulong) prev_tables, (ulong) const_tables)); delete quick; quick=0; - needed_reg=0; quick_keys=0; + needed_reg.clear_all(); quick_keys.clear_all(); if (!cond || (specialflag & SPECIAL_SAFE_MODE) && ! force_quick_range || !limit) DBUG_RETURN(0); /* purecov: inspected */ if (!((basflag= head->file->table_flags()) & HA_KEYPOS_TO_RNDPOS) && - keys_to_use == (uint) ~0 || !keys_to_use) + keys_to_use.is_set_all() || keys_to_use.is_clear_all()) DBUG_RETURN(0); /* Not smart database */ records=head->file->records; if (!records) @@ -616,14 +626,14 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, DBUG_PRINT("info",("Time to scan table: %g", read_time)); - keys_to_use&=head->keys_in_use_for_query; - if (keys_to_use) + keys_to_use.intersect(head->keys_in_use_for_query); + if (!keys_to_use.is_clear_all()) { MEM_ROOT *old_root,alloc; SEL_TREE *tree; KEY_PART *key_parts; PARAM param; - + /* set up parameter that is passed to all functions */ param.thd= thd; param.baseflag=basflag; @@ -650,7 +660,7 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, for (idx=0 ; idx < head->keys ; idx++) { - if (!(keys_to_use & ((key_map) 1L << idx))) + if (!keys_to_use.is_set(idx)) continue; KEY *key_info= &head->key_info[idx]; if (key_info->flags & HA_FULLTEXT) @@ -666,7 +676,7 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, key_parts->null_bit= key_info->key_part[part].null_bit; if (key_parts->field->type() == FIELD_TYPE_BLOB) key_parts->part_length+=HA_KEY_BLOB_LENGTH; - key_parts->image_type = + key_parts->image_type = (key_info->flags & HA_SPATIAL) ? Field::itMBR : Field::itRAW; } param.real_keynr[param.keys++]=idx; @@ -697,11 +707,11 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, uint keynr= param.real_keynr[idx]; if ((*key)->type == SEL_ARG::MAYBE_KEY || (*key)->maybe_flag) - needed_reg|= (key_map) 1 << keynr; + needed_reg.set_bit(keynr); found_records=check_quick_select(¶m, idx, *key); if (found_records != HA_POS_ERROR && found_records > 2 && - head->used_keys & ((table_map) 1 << keynr) && + head->used_keys.is_set(keynr) && (head->file->index_flags(keynr) & HA_KEY_READ_ONLY)) { /* @@ -744,7 +754,7 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use, my_pthread_setspecific_ptr(THR_MALLOC,old_root); thd->no_errors=0; } - DBUG_EXECUTE("info",print_quick(quick,needed_reg);); + DBUG_EXECUTE("info",print_quick(quick,&needed_reg);); /* Assume that if the user is using 'limit' we will only need to scan limit rows if we are using a key @@ -2143,7 +2153,7 @@ check_quick_select(PARAM *param,uint idx,SEL_ARG *tree) if (records != HA_POS_ERROR) { uint key=param->real_keynr[idx]; - param->table->quick_keys|= (key_map) 1 << key; + param->table->quick_keys.set_bit(key); param->table->quick_rows[key]=records; param->table->quick_key_parts[key]=param->max_key_part+1; } @@ -2869,7 +2879,7 @@ print_key(KEY_PART *key_part,const char *key,uint used_length) fputc('/',DBUG_FILE); if (field->real_maybe_null()) { - length++; // null byte is not in part_length + length++; // null byte is not in part_length if (*key++) { fwrite("NULL",sizeof(char),4,DBUG_FILE); @@ -2885,17 +2895,18 @@ print_key(KEY_PART *key_part,const char *key,uint used_length) } } -static void print_quick(QUICK_SELECT *quick,key_map needed_reg) +static void print_quick(QUICK_SELECT *quick,const key_map* needed_reg) { QUICK_RANGE *range; + char buf[MAX_KEY/8+1]; DBUG_ENTER("print_param"); if (! _db_on_ || !quick) DBUG_VOID_RETURN; List_iterator<QUICK_RANGE> li(quick->ranges); DBUG_LOCK_FILE; - fprintf(DBUG_FILE,"Used quick_range on key: %d (other_keys: %lu):\n", - quick->index, (ulong) needed_reg); + fprintf(DBUG_FILE,"Used quick_range on key: %d (other_keys: 0x%s):\n", + quick->index, needed_reg->print(buf)); while ((range=li++)) { if (!(range->flag & NO_MIN_RANGE)) diff --git a/sql/opt_range.h b/sql/opt_range.h index 128f6259055..1c209dc7027 100644 --- a/sql/opt_range.h +++ b/sql/opt_range.h @@ -128,7 +128,7 @@ class SQL_SELECT :public Sql_alloc { SQL_SELECT(); ~SQL_SELECT(); bool check_quick(THD *thd, bool force_quick_range, ha_rows limit) - { return test_quick_select(thd, ~0L,0,limit, force_quick_range) < 0; } + { return test_quick_select(thd, key_map(~0), 0, limit, force_quick_range) < 0; } inline bool skipp_record() { return cond ? cond->val_int() == 0 : 0; } int test_quick_select(THD *thd, key_map keys, table_map prev_tables, ha_rows limit, bool force_quick_range=0); diff --git a/sql/opt_sum.cc b/sql/opt_sum.cc index 34cce98d087..932aceebdbb 100644 --- a/sql/opt_sum.cc +++ b/sql/opt_sum.cc @@ -683,7 +683,7 @@ static bool find_key_for_maxmin(bool max_fl, TABLE_REF *ref, The following test is false when the key in the key tree is converted (for example to upper case) */ - if (field->part_of_key & ((key_map) 1 << idx)) + if (field->part_of_key.is_set(idx)) { table->key_read= 1; table->file->extra(HA_EXTRA_KEYREAD); @@ -696,7 +696,7 @@ static bool find_key_for_maxmin(bool max_fl, TABLE_REF *ref, return 0; } - + /* Check whether found key is in range specified by conditions @@ -707,7 +707,7 @@ static bool find_key_for_maxmin(bool max_fl, TABLE_REF *ref, field in: Field used the MIN/MAX expression cond in: WHERE condition range_fl in: Says whether there is a condition to to be checked - prefix_len in: Length of the constant part of the key + prefix_len in: Length of the constant part of the key RETURN 0 ok diff --git a/sql/set_var.cc b/sql/set_var.cc index 5bebbce8532..57076bec603 100644 --- a/sql/set_var.cc +++ b/sql/set_var.cc @@ -569,15 +569,15 @@ struct show_var_st init_vars[]= { {"ft_boolean_syntax", (char*) ft_boolean_syntax, SHOW_CHAR}, {"ft_min_word_len", (char*) &ft_min_word_len, SHOW_LONG}, {"ft_max_word_len", (char*) &ft_max_word_len, SHOW_LONG}, - {"ft_max_word_len_for_sort",(char*) &ft_max_word_len_for_sort, SHOW_LONG}, + {"ft_query_expansion_limit",(char*) &ft_query_expansion_limit, SHOW_LONG}, {"ft_stopword_file", (char*) &ft_stopword_file, SHOW_CHAR_PTR}, {"have_bdb", (char*) &have_berkeley_db, SHOW_HAVE}, {"have_crypt", (char*) &have_crypt, SHOW_HAVE}, {"have_compress", (char*) &have_compress, SHOW_HAVE}, {"have_innodb", (char*) &have_innodb, SHOW_HAVE}, - {"have_isam", (char*) &have_isam, SHOW_HAVE}, + {"have_isam", (char*) &have_isam, SHOW_HAVE}, {"have_raid", (char*) &have_raid, SHOW_HAVE}, - {"have_symlink", (char*) &have_symlink, SHOW_HAVE}, + {"have_symlink", (char*) &have_symlink, SHOW_HAVE}, {"have_openssl", (char*) &have_openssl, SHOW_HAVE}, {"have_query_cache", (char*) &have_query_cache, SHOW_HAVE}, {"init_file", (char*) &opt_init_file, SHOW_CHAR_PTR}, diff --git a/sql/share/czech/errmsg.txt b/sql/share/czech/errmsg.txt index 89cf199c80e..2176fcbd441 100644 --- a/sql/share/czech/errmsg.txt +++ b/sql/share/czech/errmsg.txt @@ -293,3 +293,4 @@ character-set=latin2 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/danish/errmsg.txt b/sql/share/danish/errmsg.txt index 6df62f93a84..791060ac744 100644 --- a/sql/share/danish/errmsg.txt +++ b/sql/share/danish/errmsg.txt @@ -287,3 +287,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/dutch/errmsg.txt b/sql/share/dutch/errmsg.txt index 9862e1d5026..62e210f1016 100644 --- a/sql/share/dutch/errmsg.txt +++ b/sql/share/dutch/errmsg.txt @@ -295,3 +295,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/english/errmsg.txt b/sql/share/english/errmsg.txt index aa073231c00..a0d746ce0ca 100644 --- a/sql/share/english/errmsg.txt +++ b/sql/share/english/errmsg.txt @@ -277,10 +277,11 @@ character-set=latin1 "Variable '%-.64s' is not a variable component (Can't be used as XXXX.variable_name)", "Unknown collation: '%-.64s'", "SSL parameters in CHANGE MASTER are ignored because this MySQL slave was compiled without SSL support. They can be used later when MySQL slave with SSL will be started." -"Server is running in --secure-auth mode, but '%s@%s' has a password in the old format; please change the password to the new format", +"Server is running in --secure-auth mode, but '%s'@'%s' has a password in the old format; please change the password to the new format", "Field or reference '%-.64s%s%-.64s%s%-.64s' of SELECT #%d was resolved in SELECT #%d", "Wrong parameter or combination of parameters for START SLAVE UNTIL" "It is recommended to run with --skip-slave-start when doing step-by-step replication with START SLAVE UNTIL, otherwise you are not safe in case of unexpected slave's mysqld restart" "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/estonian/errmsg.txt b/sql/share/estonian/errmsg.txt index 37ab8d97bfb..0510e3fc76b 100644 --- a/sql/share/estonian/errmsg.txt +++ b/sql/share/estonian/errmsg.txt @@ -289,3 +289,4 @@ character-set=latin7 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/french/errmsg.txt b/sql/share/french/errmsg.txt index 0fbf0bd3dd8..485cbe45724 100644 --- a/sql/share/french/errmsg.txt +++ b/sql/share/french/errmsg.txt @@ -284,3 +284,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/german/errmsg.txt b/sql/share/german/errmsg.txt index 2961b27529d..c9760879f3c 100644 --- a/sql/share/german/errmsg.txt +++ b/sql/share/german/errmsg.txt @@ -296,3 +296,4 @@ character-set=latin1 "SQL-Thread soll nicht gestartet werden. Daher werden UNTIL-Optionen ignoriert" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/greek/errmsg.txt b/sql/share/greek/errmsg.txt index 328fc20acb8..90ab24610ff 100644 --- a/sql/share/greek/errmsg.txt +++ b/sql/share/greek/errmsg.txt @@ -284,3 +284,4 @@ character-set=greek "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/hungarian/errmsg.txt b/sql/share/hungarian/errmsg.txt index 2e64b986790..c456e9580b3 100644 --- a/sql/share/hungarian/errmsg.txt +++ b/sql/share/hungarian/errmsg.txt @@ -286,3 +286,4 @@ character-set=latin2 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/italian/errmsg.txt b/sql/share/italian/errmsg.txt index 4e6b963e516..09b533c0fb7 100644 --- a/sql/share/italian/errmsg.txt +++ b/sql/share/italian/errmsg.txt @@ -284,3 +284,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/japanese/errmsg.txt b/sql/share/japanese/errmsg.txt index e4d31cb5d8f..15b8640cb29 100644 --- a/sql/share/japanese/errmsg.txt +++ b/sql/share/japanese/errmsg.txt @@ -286,3 +286,4 @@ character-set=ujis "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/korean/errmsg.txt b/sql/share/korean/errmsg.txt index 973985734b9..d31efc94d46 100644 --- a/sql/share/korean/errmsg.txt +++ b/sql/share/korean/errmsg.txt @@ -284,3 +284,4 @@ character-set=euckr "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/norwegian-ny/errmsg.txt b/sql/share/norwegian-ny/errmsg.txt index 29484d52916..515a322f136 100644 --- a/sql/share/norwegian-ny/errmsg.txt +++ b/sql/share/norwegian-ny/errmsg.txt @@ -286,3 +286,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/norwegian/errmsg.txt b/sql/share/norwegian/errmsg.txt index b13a0fdec5b..faea5402fc6 100644 --- a/sql/share/norwegian/errmsg.txt +++ b/sql/share/norwegian/errmsg.txt @@ -286,3 +286,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/polish/errmsg.txt b/sql/share/polish/errmsg.txt index 87a6b92e942..2a7ea3accb8 100644 --- a/sql/share/polish/errmsg.txt +++ b/sql/share/polish/errmsg.txt @@ -288,3 +288,4 @@ character-set=latin2 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/portuguese/errmsg.txt b/sql/share/portuguese/errmsg.txt index 128bbfb057f..1f3bbb5c02f 100644 --- a/sql/share/portuguese/errmsg.txt +++ b/sql/share/portuguese/errmsg.txt @@ -285,3 +285,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/romanian/errmsg.txt b/sql/share/romanian/errmsg.txt index ee81629d9db..78daafe04d1 100644 --- a/sql/share/romanian/errmsg.txt +++ b/sql/share/romanian/errmsg.txt @@ -288,3 +288,4 @@ character-set=latin2 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/russian/errmsg.txt b/sql/share/russian/errmsg.txt index 3f272ee870a..462afdfa782 100644 --- a/sql/share/russian/errmsg.txt +++ b/sql/share/russian/errmsg.txt @@ -286,3 +286,4 @@ character-set=koi8r "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "ëÅÛ ÚÁÐÒÏÓÏ× ÎÅ ÍÏÖÅÔ ÕÓÔÁÎÏ×ÉÔØ ÒÁÚÍÅÒ %lu, ÎÏ×ÙÊ ÒÁÚÍÅÒ ËÅÛÁ ÚÐÒÏÓÏ× - %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/serbian/errmsg.txt b/sql/share/serbian/errmsg.txt index 2801d47739f..9d12654f150 100644 --- a/sql/share/serbian/errmsg.txt +++ b/sql/share/serbian/errmsg.txt @@ -279,3 +279,4 @@ character-set=cp1250 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/slovak/errmsg.txt b/sql/share/slovak/errmsg.txt index b72dc664aa0..dd222c2ef30 100644 --- a/sql/share/slovak/errmsg.txt +++ b/sql/share/slovak/errmsg.txt @@ -292,3 +292,4 @@ character-set=latin2 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/spanish/errmsg.txt b/sql/share/spanish/errmsg.txt index ae19b2b14a1..a21d306a3db 100644 --- a/sql/share/spanish/errmsg.txt +++ b/sql/share/spanish/errmsg.txt @@ -286,3 +286,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/swedish/errmsg.txt b/sql/share/swedish/errmsg.txt index 0f866322f87..4d3e69a5b0f 100644 --- a/sql/share/swedish/errmsg.txt +++ b/sql/share/swedish/errmsg.txt @@ -284,3 +284,4 @@ character-set=latin1 "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "Query cache failed to set size %lu, new query cache size is %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/share/ukrainian/errmsg.txt b/sql/share/ukrainian/errmsg.txt index 399af72cafa..283700b7891 100644 --- a/sql/share/ukrainian/errmsg.txt +++ b/sql/share/ukrainian/errmsg.txt @@ -289,3 +289,4 @@ character-set=koi8u "SQL thread is not to be started so UNTIL options are ignored" "Incorrect index name '%-.100s'", "ëÅÛ ÚÁÐÉÔ¦× ÎÅÓÐÒÏÍÏÖÅÎ ×ÓÔÁÎÏ×ÉÔÉ ÒÏÚÍ¦Ò %lu, ÎÏ×ÉÊ ÒÏÚÍ¦Ò ËÅÛÁ ÚÁÐÉÔ¦× - %lu", +"Column '%-.64s' cannot be part of FULLTEXT index" diff --git a/sql/slave.cc b/sql/slave.cc index bb190762855..415969a2955 100644 --- a/sql/slave.cc +++ b/sql/slave.cc @@ -330,7 +330,7 @@ err: void init_slave_skip_errors(const char* arg) { const char *p; - if (bitmap_init(&slave_error_mask,MAX_SLAVE_ERROR,0)) + if (bitmap_init(&slave_error_mask,0,MAX_SLAVE_ERROR,0)) { fprintf(stderr, "Badly out of memory, please check your system status\n"); exit(1); diff --git a/sql/sql_acl.cc b/sql/sql_acl.cc index b8423fc6b7a..b54e6a95bc5 100644 --- a/sql/sql_acl.cc +++ b/sql/sql_acl.cc @@ -3545,8 +3545,16 @@ int mysql_revoke_all(THD *thd, List <LEX_USER> &list) VOID(pthread_mutex_unlock(&acl_cache->lock)); rw_unlock(&LOCK_grant); close_thread_tables(thd); + + /* XXX this should not be necessary. The error message is already printed + by replace_xxx_table. my_error() should be use above instead of + sql_print_error(), and print ER_NONEXISTING_GRANT - as other grant + commands do */ + /* when this code is deleted, the error slot (error 1268) can be reused, + as this error code was not present in any MySQL release */ if (result) my_error(ER_REVOKE_GRANTS, MYF(0)); + DBUG_RETURN(result); } diff --git a/sql/sql_base.cc b/sql/sql_base.cc index 53f430924d7..cc068451ecf 100644 --- a/sql/sql_base.cc +++ b/sql/sql_base.cc @@ -1802,7 +1802,7 @@ Field *find_field_in_table(THD *thd,TABLE *table,const char *name,uint length, goto found; } } - if (allow_rowid && + if (allow_rowid && !my_strcasecmp(system_charset_info, name, "_rowid") && (field=table->rowid_field)) goto found; @@ -1815,7 +1815,7 @@ Field *find_field_in_table(THD *thd,TABLE *table,const char *name,uint length, { field->query_id=thd->query_id; table->used_fields++; - table->used_keys&= field->part_of_key; + table->used_keys.intersect(field->part_of_key); } else thd->dupp_field=field; @@ -1844,7 +1844,7 @@ Field *find_field_in_table(THD *thd,TABLE *table,const char *name,uint length, RETURN VALUES 0 Field is not found or field is not unique- error message is reported - not_found_field Function was called with report_error == FALSE and + not_found_field Function was called with report_error == FALSE and field was not found. no error message reported. found field */ @@ -2168,21 +2168,21 @@ bool setup_tables(TABLE_LIST *tables) table->used_keys= table->keys_for_keyread; if (table_list->use_index) { - key_map map= get_key_map_from_key_list(table, - table_list->use_index); - if (map == ~(key_map) 0) + key_map map; + get_key_map_from_key_list(&map, table, table_list->use_index); + if (map.is_set_all()) DBUG_RETURN(1); table->keys_in_use_for_query=map; } if (table_list->ignore_index) { - key_map map= get_key_map_from_key_list(table, - table_list->ignore_index); - if (map == ~(key_map) 0) + key_map map; + get_key_map_from_key_list(&map, table, table_list->ignore_index); + if (map.is_set_all()) DBUG_RETURN(1); - table->keys_in_use_for_query &= ~map; + table->keys_in_use_for_query.subtract(map); } - table->used_keys &= table->keys_in_use_for_query; + table->used_keys.intersect(table->keys_in_use_for_query); if (table_list->shared || table->clear_query_id) { table->clear_query_id= 0; @@ -2200,24 +2200,26 @@ bool setup_tables(TABLE_LIST *tables) } -key_map get_key_map_from_key_list(TABLE *table, - List<String> *index_list) +void get_key_map_from_key_list(key_map *map, TABLE *table, + List<String> *index_list) { - key_map map=0; List_iterator_fast<String> it(*index_list); String *name; uint pos; + + map->clear_all(); while ((name=it++)) { if ((pos=find_type(name->c_ptr(), &table->keynames, 1+2)) <= 0) { my_error(ER_KEY_COLUMN_DOES_NOT_EXITS, MYF(0), name->c_ptr(), table->real_name); - return (~ (key_map) 0); + map->set_all(); + return; } - map|= ((key_map) 1) << (pos-1); + map->set_bit(pos-1); } - return map; + return; } /**************************************************************************** @@ -2262,7 +2264,7 @@ insert_fields(THD *thd,TABLE_LIST *tables, const char *db_name, if (field->query_id == thd->query_id) thd->dupp_field=field; field->query_id=thd->query_id; - table->used_keys&= field->part_of_key; + table->used_keys.intersect(field->part_of_key); } /* All fields are used */ table->used_fields=table->fields; @@ -2353,8 +2355,8 @@ int setup_conds(THD *thd,TABLE_LIST *tables,COND **conds) /* Mark field used for table cache */ t1->field[i]->query_id=t2->field[j]->query_id=thd->query_id; cond_and->list.push_back(tmp); - t1->used_keys&= t1->field[i]->part_of_key; - t2->used_keys&= t2->field[j]->part_of_key; + t1->used_keys.intersect(t1->field[i]->part_of_key); + t2->used_keys.intersect(t2->field[j]->part_of_key); break; } } diff --git a/sql/sql_bitmap.h b/sql/sql_bitmap.h new file mode 100644 index 00000000000..53293e39591 --- /dev/null +++ b/sql/sql_bitmap.h @@ -0,0 +1,110 @@ +#include <my_global.h> +//#include <mysql_version.h> +//#include <mysql_embed.h> +//#include <my_sys.h> +//#include <m_string.h> +//#include <hash.h> +//#include <signal.h> +//#include <thr_lock.h> +//#include <my_base.h> +#include <my_bitmap.h> +#include <assert.h> + +template <uint default_width> class Bitmap +{ + MY_BITMAP map; + uchar buffer[(default_width+7)/8]; +public: + Bitmap() { init(); } + Bitmap(Bitmap& from) { *this=from; } + Bitmap(uint prefix_to_set) { init(prefix_to_set); } + void init() { bitmap_init(&map, buffer, default_width, 0); } + void init(uint prefix_to_set) { init(); set_prefix(prefix_to_set); } + uint length() const { return default_width; } + Bitmap& operator=(const Bitmap& map2) + { + init(); + memcpy(buffer, map2.buffer, sizeof(buffer)); + return *this; + } + void set_bit(uint n) { bitmap_set_bit(&map, n); } + void clear_bit(uint n) { bitmap_clear_bit(&map, n); } + void set_prefix(uint n) { bitmap_set_prefix(&map, n); } + void set_all() { bitmap_set_all(&map); } + void clear_all() { bitmap_clear_all(&map); } + void intersect(Bitmap& map2) { bitmap_intersect(&map, &map2.map); } + void intersect(ulonglong map2buff) + { + MY_BITMAP map2; + bitmap_init(&map2, (uchar *)&map2buff, sizeof(ulonglong)*8, 0); + bitmap_intersect(&map, &map2); + } + void subtract(Bitmap& map2) { bitmap_subtract(&map, &map2.map); } + void merge(Bitmap& map2) { bitmap_union(&map, &map2.map); } + my_bool is_set(uint n) const { return bitmap_is_set(&map, n); } + my_bool is_prefix(uint n) const { return bitmap_is_prefix(&map, n); } + my_bool is_clear_all() const { return bitmap_is_clear_all(&map); } + my_bool is_set_all() const { return bitmap_is_set_all(&map); } + my_bool is_subset(const Bitmap& map2) const { return bitmap_is_subset(&map, &map2.map); } + my_bool operator==(const Bitmap& map2) const { return bitmap_cmp(&map, &map2.map); } + char *print(char *buf) const + { + char *s=buf; int i; + for (i=sizeof(buffer)-1; i>=0 ; i--) + { + if ((*s=_dig_vec[buffer[i] >> 4]) != '0') + break; + if ((*s=_dig_vec[buffer[i] & 15]) != '0') + break; + } + for (s++, i-- ; i>=0 ; i--) + { + *s++=_dig_vec[buffer[i] >> 4]; + *s++=_dig_vec[buffer[i] & 15]; + } + *s=0; + return buf; + } + ulonglong to_ulonglong() const + { + if (sizeof(buffer) >= 8) + return uint8korr(buffer); + DBUG_ASSERT(sizeof(buffer) >= 4); + uint4korr(buffer); + } +}; + +template <> class Bitmap<64> +{ + ulonglong map; +public: + Bitmap<64>() { } + Bitmap<64>(uint prefix_to_set) { set_prefix(prefix_to_set); } + void init() { } + void init(uint prefix_to_set) { set_prefix(prefix_to_set); } + uint length() const { return 64; } + void set_bit(uint n) { map|= ((ulonglong)1) << n; } + void clear_bit(uint n) { map&= ~(((ulonglong)1) << n); } + void set_prefix(uint n) + { + if (n >= length()) + set_all(); + else + map= (((ulonglong)1) << n)-1; + } + void set_all() { map=~(ulonglong)0; } + void clear_all() { map=(ulonglong)0; } + void intersect(Bitmap<64>& map2) { map&= map2.map; } + void intersect(ulonglong map2) { map&= map2; } + void subtract(Bitmap<64>& map2) { map&= ~map2.map; } + void merge(Bitmap<64>& map2) { map|= map2.map; } + my_bool is_set(uint n) const { return test(map & (((ulonglong)1) << n)); } + my_bool is_prefix(uint n) const { return map == (((ulonglong)1) << n)-1; } + my_bool is_clear_all() const { return map == (ulonglong)0; } + my_bool is_set_all() const { return map == ~(ulonglong)0; } + my_bool is_subset(const Bitmap<64>& map2) const { return !(map & ~map2.map); } + my_bool operator==(const Bitmap<64>& map2) const { return map == map2.map; } + char *print(char *buf) const { longlong2str(map,buf,16); return buf; } + ulonglong to_ulonglong() const { return map; } +}; + diff --git a/sql/sql_delete.cc b/sql/sql_delete.cc index 070d4cbbce9..2985f7e04a0 100644 --- a/sql/sql_delete.cc +++ b/sql/sql_delete.cc @@ -85,7 +85,8 @@ int mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds, ORDER *order, /* Handler didn't support fast delete; Delete rows one by one */ } - table->used_keys=table->quick_keys=0; // Can't use 'only index' + table->used_keys.clear_all(); + table->quick_keys.clear_all(); // Can't use 'only index' select=make_select(table,0,0,conds,&error); if (error) DBUG_RETURN(-1); @@ -98,7 +99,7 @@ int mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds, ORDER *order, } /* If running in safe sql mode, don't allow updates without keys */ - if (!table->quick_keys) + if (table->quick_keys.is_clear_all()) { thd->lex.select_lex.options|=QUERY_NO_INDEX_USED; if (safe_update && !using_limit) @@ -292,7 +293,7 @@ multi_delete::initialize_tables(JOIN *join) walk=walk->next; /* Don't use KEYREAD optimization on this table */ tbl->no_keyread=1; - tbl->used_keys= 0; + tbl->used_keys.clear_all(); if (tbl->file->has_transactions()) log_delayed= transactional_tables= 1; else if (tbl->tmp_table != NO_TMP_TABLE) diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc index 596c24541fb..a6c6080dd98 100644 --- a/sql/sql_parse.cc +++ b/sql/sql_parse.cc @@ -124,16 +124,6 @@ static bool end_active_trans(THD *thd) static HASH hash_user_connections; -/* - get_or_create_user_conn() - - RETURN VALUE - 0 OK; thd->user_connect is set, thd->user_connect->connections is - incremented by one (so that the user_connect object cannot be dropped - while we are using it) - 1 error -*/ - static int get_or_create_user_conn(THD *thd, const char *user, const char *host, USER_RESOURCES *mqh) @@ -167,7 +157,7 @@ static int get_or_create_user_conn(THD *thd, const char *user, uc->user_len= user_len; uc->host=uc->user + uc->user_len + 1; uc->len = temp_len; - uc->connections = 1; /* Another BUG FIX by Heikki Oct 26, 2003 */ + uc->connections = 1; uc->questions=uc->updates=uc->conn_per_hour=0; uc->user_resources=*mqh; if (max_user_connections && mqh->connections > max_user_connections) @@ -180,14 +170,12 @@ static int get_or_create_user_conn(THD *thd, const char *user, return_val=1; goto end; } - } else { - uc->connections++; /* BUG FIX by Heikki Oct 26, 2003 */ } - thd->user_connect=uc; end: (void) pthread_mutex_unlock(&LOCK_user_conn); return return_val; + } #ifndef NO_EMBEDDED_ACCESS_CHECKS @@ -211,9 +199,8 @@ end: are 'IN'. RETURN VALUE - 0 OK; thd->user, thd->master_access, thd->priv_user, thd->db, - thd->db_access, and thd->user_connect->connections are updated; OK is - sent to client; + 0 OK; thd->user, thd->master_access, thd->priv_user, thd->db and + thd->db_access are updated; OK is sent to client; -1 access denied or handshake error; error is sent to client; >0 error, not sent to client */ @@ -327,26 +314,13 @@ int check_user(THD *thd, enum enum_server_command command, thd->db_access=0; /* Don't allow user to connect if he has done too many queries */ - - /* - BUG FIXes by Heikki Oct 26, 2003 - - NOTE that the call get_or_create_user_conn() increments the connection - count by 1 if it succeeds! This is to prevent the dropping of the - thd->user_connect object before we check the connection limits in - check_for_max_user_connections. - */ - - if (get_or_create_user_conn(thd,thd->user,thd->host_or_ip,&ur)) - DBUG_RETURN(1); - - if (thd->user_connect && + if ((ur.questions || ur.updates || + ur.connections || max_user_connections) && + get_or_create_user_conn(thd,thd->user,thd->host_or_ip,&ur)) + DBUG_RETURN(-1); + if (thd->user_connect && thd->user_connect->user_resources.connections && check_for_max_user_connections(thd, thd->user_connect)) - { - decrease_user_connections(thd->user_connect); - - DBUG_RETURN(1); - } + DBUG_RETURN(-1); /* Change database if necessary: OK or FAIL is sent in mysql_change_db */ if (db && db[0]) @@ -416,21 +390,14 @@ static int check_for_max_user_connections(THD *thd, USER_CONN *uc) int error=0; DBUG_ENTER("check_for_max_user_connections"); - /* - FIX by Heikki Oct 25, 2003: protect the operation with a mutex: is there - a race condition here otherwise? - */ - - (void) pthread_mutex_lock(&LOCK_user_conn); - if (max_user_connections && - (max_user_connections < (uint) uc->connections)) + (max_user_connections < (uint) uc->connections)) { net_printf(thd,ER_TOO_MANY_USER_CONNECTIONS, uc->user); error=1; goto end; } - + uc->connections++; if (uc->user_resources.connections && uc->conn_per_hour++ >= uc->user_resources.connections) { @@ -439,11 +406,7 @@ static int check_for_max_user_connections(THD *thd, USER_CONN *uc) (long) uc->user_resources.connections); error=1; } - end: - - (void) pthread_mutex_unlock(&LOCK_user_conn); - DBUG_RETURN(error); } @@ -451,27 +414,13 @@ end: static void decrease_user_connections(USER_CONN *uc) { DBUG_ENTER("decrease_user_connections"); - - /* - FIX by Heikki Oct 25, 2003: protect the operation with a mutex: is there - a race condition here otherwise? - */ - - (void) pthread_mutex_lock(&LOCK_user_conn); - - if (uc->connections == 0) - printf("Error: trying to decrease user %s connections below zero!\n", - uc->user); - if ((!--uc->connections) && !mqh_used) + if ((uc->connections && !--uc->connections) && !mqh_used) { - /* Last connection for user; delete it if we do not need the object - for keeping maximum-queries-per-hour statistics (mqh) */ - + /* Last connection for user; Delete it */ + (void) pthread_mutex_lock(&LOCK_user_conn); (void) hash_delete(&hash_user_connections,(byte*) uc); + (void) pthread_mutex_unlock(&LOCK_user_conn); } - - (void) pthread_mutex_unlock(&LOCK_user_conn); - DBUG_VOID_RETURN; } @@ -1304,7 +1253,7 @@ bool dispatch_command(enum enum_server_command command, THD *thd, else { /* we've authenticated new user */ - if (save_uc) /* BUG FIX by Heikki Oct 26, 2003: always keep the count */ + if (max_connections && save_uc) decrease_user_connections(save_uc); x_free((gptr) save_db); x_free((gptr) save_user); @@ -3729,6 +3678,7 @@ mysql_new_select(LEX *lex, bool move_down) unit->link_prev= 0; unit->return_to= lex->current_select; select_lex->include_down(unit); + // TODO: assign resolve_mode for fake subquery after merging with new tree } else { @@ -3748,7 +3698,6 @@ mysql_new_select(LEX *lex, bool move_down) fake->make_empty_select(); fake->linkage= GLOBAL_OPTIONS_TYPE; fake->select_limit= lex->thd->variables.select_limit; - fake->resolve_mode= SELECT_LEX::SELECT_MODE; } } @@ -4737,8 +4686,8 @@ Item * all_any_subquery_creator(Item *left_expr, { if ((cmp == &comp_eq_creator) && !all) // = ANY <=> IN return new Item_in_subselect(left_expr, select_lex); - - if ((cmp == &comp_ne_creator) && all) // <> ALL <=> NOT IN + + if ((cmp == &comp_ne_creator) && all) // <> ALL <=> NOT IN return new Item_func_not(new Item_in_subselect(left_expr, select_lex)); Item_allany_subselect *it= diff --git a/sql/sql_select.cc b/sql/sql_select.cc index 9bdb989b6a1..59d3242ff37 100644 --- a/sql/sql_select.cc +++ b/sql/sql_select.cc @@ -35,6 +35,9 @@ const char *join_type_str[]={ "UNKNOWN","system","const","eq_ref","ref", "ref_or_null","unique_subquery","index_subquery" }; +const key_map key_map_empty(0); +const key_map key_map_full(~0); + static void optimize_keyuse(JOIN *join, DYNAMIC_ARRAY *keyuse_array); static bool make_join_statistics(JOIN *join,TABLE_LIST *tables,COND *conds, DYNAMIC_ARRAY *keyuse); @@ -114,7 +117,7 @@ static int join_read_next_same_or_null(READ_RECORD *info); static COND *make_cond_for_table(COND *cond,table_map table, table_map used_table); static Item* part_of_refkey(TABLE *form,Field *field); -static uint find_shortest_key(TABLE *table, key_map usable_keys); +static uint find_shortest_key(TABLE *table, const key_map *usable_keys); static bool test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order, ha_rows select_limit, bool no_changes); static int create_sort_index(THD *thd, JOIN *join, ORDER *order, @@ -526,7 +529,7 @@ JOIN::optimize() conds= optimize_cond(conds,&cond_value); if (thd->net.report_error) { - error= 1; + error= 1; DBUG_PRINT("error",("Error from optimize_cond")); DBUG_RETURN(1); } @@ -556,7 +559,7 @@ JOIN::optimize() if (res < 0) { zero_result_cause= "No matching min/max row"; - error=0; + error=0; DBUG_RETURN(0); } zero_result_cause= "Select tables optimized away"; @@ -703,7 +706,7 @@ JOIN::optimize() /* Force MySQL to read the table in sorted order to get result in ORDER BY order. - */ + */ tmp_table_param.quick_group=0; } order=0; @@ -765,7 +768,7 @@ JOIN::optimize() make_join_readinfo(this, (select_options & (SELECT_DESCRIBE | SELECT_NO_JOIN_CACHE)) | - (select_lex->ftfunc_list->elements ? + (select_lex->ftfunc_list->elements ? SELECT_NO_JOIN_CACHE : 0)); /* @@ -831,7 +834,7 @@ JOIN::optimize() conds, 1))); } - + } /* Need to tell Innobase that to play it safe, it should fetch all @@ -1494,7 +1497,7 @@ JOIN::cleanup() int mysql_select(THD *thd, Item ***rref_pointer_array, - TABLE_LIST *tables, uint wild_num, List<Item> &fields, + TABLE_LIST *tables, uint wild_num, List<Item> &fields, COND *conds, uint og_num, ORDER *order, ORDER *group, Item *having, ORDER *proc_param, ulong select_options, select_result *result, SELECT_LEX_UNIT *unit, @@ -1558,7 +1561,7 @@ err: JOIN *curr_join= (join->need_tmp&&join->tmp_join? (join->tmp_join->error=join->error,join->tmp_join): join); - + thd->proc_info="end"; err= join->cleanup(); if (thd->net.report_error) @@ -1576,7 +1579,7 @@ err: static ha_rows get_quick_record_count(THD *thd, SQL_SELECT *select, TABLE *table, - key_map keys,ha_rows limit) + const key_map *keys,ha_rows limit) { int error; DBUG_ENTER("get_quick_record_count"); @@ -1584,7 +1587,7 @@ static ha_rows get_quick_record_count(THD *thd, SQL_SELECT *select, { select->head=table; table->reginfo.impossible_range=0; - if ((error=select->test_quick_select(thd, keys,(table_map) 0,limit)) + if ((error=select->test_quick_select(thd, *(key_map *)keys,(table_map) 0,limit)) == 1) DBUG_RETURN(select->quick->records); if (error == -1) @@ -1638,9 +1641,13 @@ make_join_statistics(JOIN *join,TABLE_LIST *tables,COND *conds, { TABLE *table; stat_vector[i]=s; + s->keys.init(); + s->const_keys.init(); + s->checked_keys.init(); + s->needed_reg.init(); table_vector[i]=s->table=table=tables->table; table->file->info(HA_STATUS_VARIABLE | HA_STATUS_NO_LOCK);// record count - table->quick_keys=0; + table->quick_keys.clear_all(); table->reginfo.join_tab=s; table->reginfo.not_exists_optimize=0; bzero((char*) table->const_key_parts, sizeof(key_part_map)*table->keys); @@ -1786,24 +1793,25 @@ make_join_statistics(JOIN *join,TABLE_LIST *tables,COND *conds, { start_keyuse=keyuse; key=keyuse->key; - s->keys|= (key_map) 1 << key; // QQ: remove this ? + s->keys.set_bit(key); // QQ: remove this ? - refs=const_ref=0; - eq_part=0; + refs=0; + const_ref.clear_all(); + eq_part.clear_all(); do { if (keyuse->val->type() != Item::NULL_ITEM && !keyuse->optimize) { if (!((~found_const_table_map) & keyuse->used_tables)) - const_ref|= (key_map) 1 << keyuse->keypart; + const_ref.set_bit(keyuse->keypart); else refs|=keyuse->used_tables; - eq_part|= (key_map) 1 << keyuse->keypart; + eq_part.set_bit(keyuse->keypart); } keyuse++; } while (keyuse->table == table && keyuse->key == key); - if (eq_part == PREV_BITS(uint,table->key_info[key].key_parts) && + if (eq_part.is_prefix(table->key_info[key].key_parts) && (table->key_info[key].flags & HA_NOSAME) && !table->fulltext_searched) { @@ -1859,7 +1867,7 @@ make_join_statistics(JOIN *join,TABLE_LIST *tables,COND *conds, if (s->worst_seeks < 2.0) // Fix for small tables s->worst_seeks=2.0; - if (s->const_keys) + if (! s->const_keys.is_clear_all()) { ha_rows records; SQL_SELECT *select; @@ -1868,7 +1876,7 @@ make_join_statistics(JOIN *join,TABLE_LIST *tables,COND *conds, s->on_expr ? s->on_expr : conds, &error); records= get_quick_record_count(join->thd, select, s->table, - s->const_keys, join->row_limit); + &s->const_keys, join->row_limit); s->quick=select->quick; s->needed_reg=select->needed_reg; select->quick=0; @@ -2096,9 +2104,9 @@ add_key_field(KEY_FIELD **key_fields,uint and_level, else { JOIN_TAB *stat=field->table->reginfo.join_tab; - key_map possible_keys= (field->key_start & - field->table->keys_in_use_for_query); - stat[0].keys|= possible_keys; // Add possible keys + key_map possible_keys=field->key_start; + possible_keys.intersect(field->table->keys_in_use_for_query); + stat[0].keys.merge(possible_keys); // Add possible keys /* Save the following cases: @@ -2117,7 +2125,7 @@ add_key_field(KEY_FIELD **key_fields,uint and_level, for (uint i=0; i<num_values; i++) is_const&= (*value)->const_item(); if (is_const) - stat[0].const_keys |= possible_keys; + stat[0].const_keys.merge(possible_keys); /* We can't always use indexes when comparing a string index to a number. cmp_type() is checked to allow compare of dates to numbers. @@ -2248,14 +2256,13 @@ add_key_fields(JOIN_TAB *stat,KEY_FIELD **key_fields,uint *and_level, */ static uint -max_part_bit(key_map bits) +max_part_bit(key_part_map bits) { uint found; for (found=0; bits & 1 ; found++,bits>>=1) ; return found; } - static void add_key_part(DYNAMIC_ARRAY *keyuse_array,KEY_FIELD *key_field) { @@ -2267,7 +2274,7 @@ add_key_part(DYNAMIC_ARRAY *keyuse_array,KEY_FIELD *key_field) { for (uint key=0 ; key < form->keys ; key++) { - if (!(form->keys_in_use_for_query & (((key_map) 1) << key))) + if (!(form->keys_in_use_for_query.is_set(key))) continue; if (form->key_info[key].flags & HA_FULLTEXT) continue; // ToDo: ft-keys in non-ft queries. SerG @@ -2458,7 +2465,7 @@ update_ref_and_keys(THD *thd, DYNAMIC_ARRAY *keyuse,JOIN_TAB *join_tab, /* Save ptr to first use */ if (!use->table->reginfo.join_tab->keyuse) use->table->reginfo.join_tab->keyuse=save_pos; - use->table->reginfo.join_tab->checked_keys|= (key_map) 1 << use->key; + use->table->reginfo.join_tab->checked_keys.set_bit(use->key); save_pos++; } i=(uint) (save_pos-(KEYUSE*) keyuse->buffer); @@ -2599,7 +2606,7 @@ find_best(JOIN *join,table_map rest_tables,uint idx,double record_count, rec= s->records/MATCHING_ROWS_IN_OTHER_TABLE; // Assumed records/key for (keyuse=s->keyuse ; keyuse->table == table ;) { - key_map found_part=0; + key_part_map found_part=0; table_map found_ref=0; uint key=keyuse->key; KEY *keyinfo=table->key_info+key; @@ -2668,7 +2675,7 @@ find_best(JOIN *join,table_map rest_tables,uint idx,double record_count, { if (!found_ref) { // We found a const key - if (table->quick_keys & ((key_map) 1 << key)) + if (table->quick_keys.is_set(key)) records= (double) table->quick_rows[key]; else { @@ -2692,7 +2699,7 @@ find_best(JOIN *join,table_map rest_tables,uint idx,double record_count, /* Limit the number of matched rows */ tmp= records; set_if_smaller(tmp, (double) thd->variables.max_seeks_for_key); - if (table->used_keys & ((key_map) 1 << key)) + if (table->used_keys.is_set(key)) { /* we can use only index tree */ uint keys_per_block= table->file->block_size/2/ @@ -2719,7 +2726,7 @@ find_best(JOIN *join,table_map rest_tables,uint idx,double record_count, Check if quick_range could determinate how many rows we will match */ - if (table->quick_keys & ((key_map) 1 << key) && + if (table->quick_keys.is_set(key) && table->quick_key_parts[key] <= max_key_part) tmp=records= (double) table->quick_rows[key]; else @@ -2771,7 +2778,7 @@ find_best(JOIN *join,table_map rest_tables,uint idx,double record_count, } /* Limit the number of matched rows */ set_if_smaller(tmp, (double) thd->variables.max_seeks_for_key); - if (table->used_keys & ((key_map) 1 << key)) + if (table->used_keys.is_set(key)) { /* we can use only index tree */ uint keys_per_block= table->file->block_size/2/ @@ -2810,7 +2817,7 @@ find_best(JOIN *join,table_map rest_tables,uint idx,double record_count, !(s->quick && best_key && s->quick->index == best_key->key && best_max_key_part >= s->table->quick_key_parts[best_key->key]) && !((s->table->file->table_flags() & HA_TABLE_SCAN_ON_INDEX) && - s->table->used_keys && best_key) && + ! s->table->used_keys.is_clear_all() && best_key) && !(s->table->force_index && best_key)) { // Check full join ha_rows rnd_records= s->found_records; @@ -3024,7 +3031,7 @@ get_best_combination(JOIN *join) if (j->type == JT_SYSTEM) continue; - if (!j->keys || !(keyuse= join->best_positions[tablenr].key)) + if (j->keys.is_clear_all() || !(keyuse= join->best_positions[tablenr].key)) { j->type=JT_ALL; if (tablenr != join->const_tables) @@ -3249,13 +3256,13 @@ make_simple_join(JOIN *join,TABLE *tmp_table) join->row_limit=join->unit->select_limit_cnt; join->do_send_rows = (join->row_limit) ? 1 : 0; - join_tab->cache.buff=0; /* No cacheing */ + join_tab->cache.buff=0; /* No caching */ join_tab->table=tmp_table; join_tab->select=0; join_tab->select_cond=0; join_tab->quick=0; join_tab->type= JT_ALL; /* Map through all records */ - join_tab->keys= (uint) ~0; /* test everything in quick */ + join_tab->keys.init(~0); /* test everything in quick */ join_tab->info=0; join_tab->on_expr=0; join_tab->ref.key = -1; @@ -3337,13 +3344,13 @@ make_join_select(JOIN *join,SQL_SELECT *select,COND *cond) { /* Use quick key read if it's a constant and it's not used with key reading */ - if (tab->needed_reg == 0 && tab->type != JT_EQ_REF + if (tab->needed_reg.is_clear_all() && tab->type != JT_EQ_REF && tab->type != JT_FT && (tab->type != JT_REF || (uint) tab->ref.key == tab->quick->index)) { sel->quick=tab->quick; // Use value from get_quick_... - sel->quick_keys=0; - sel->needed_reg=0; + sel->quick_keys.clear_all(); + sel->needed_reg.clear_all(); } else { @@ -3354,12 +3361,13 @@ make_join_select(JOIN *join,SQL_SELECT *select,COND *cond) uint ref_key=(uint) sel->head->reginfo.join_tab->ref.key+1; if (i == join->const_tables && ref_key) { - if (tab->const_keys && tab->table->reginfo.impossible_range) + if (!tab->const_keys.is_clear_all() && + tab->table->reginfo.impossible_range) DBUG_RETURN(1); } else if (tab->type == JT_ALL && ! use_quick_range) { - if (tab->const_keys && + if (!tab->const_keys.is_clear_all() && tab->table->reginfo.impossible_range) DBUG_RETURN(1); // Impossible range /* @@ -3369,9 +3377,9 @@ make_join_select(JOIN *join,SQL_SELECT *select,COND *cond) the index if we are using limit and this is the first table */ - if ((tab->keys & ~ tab->const_keys && i > 0) || - (tab->const_keys && i == join->const_tables && - join->unit->select_limit_cnt < + if ((!tab->keys.is_subset(tab->const_keys) && i > 0) || + (!tab->const_keys.is_clear_all() && i == join->const_tables && + join->unit->select_limit_cnt < join->best_positions[i].records_read && !(join->select_options & OPTION_FOUND_ROWS))) { @@ -3409,13 +3417,15 @@ make_join_select(JOIN *join,SQL_SELECT *select,COND *cond) else { sel->needed_reg=tab->needed_reg; - sel->quick_keys=0; + sel->quick_keys.clear_all(); } - if ((sel->quick_keys | sel->needed_reg) & ~tab->checked_keys) + if (!sel->quick_keys.is_subset(tab->checked_keys) || + !sel->needed_reg.is_subset(tab->checked_keys)) { - tab->keys=sel->quick_keys | sel->needed_reg; - tab->use_quick= (sel->needed_reg && - (!select->quick_keys || + tab->keys=sel->quick_keys; + tab->keys.merge(sel->needed_reg); + tab->use_quick= (!sel->needed_reg.is_clear_all() && + (select->quick_keys.is_clear_all() || (select->quick && (select->quick->records >= 100L)))) ? 2 : 1; @@ -3480,7 +3490,7 @@ make_join_readinfo(JOIN *join, uint options) table->file->index_init(tab->ref.key); tab->read_first_record= join_read_key; tab->read_record.read_record= join_no_more_records; - if (table->used_keys & ((key_map) 1 << tab->ref.key) && + if (table->used_keys.is_set(tab->ref.key) && !table->no_keyread) { table->key_read=1; @@ -3498,7 +3508,7 @@ make_join_readinfo(JOIN *join, uint options) delete tab->quick; tab->quick=0; table->file->index_init(tab->ref.key); - if (table->used_keys & ((key_map) 1 << tab->ref.key) && + if (table->used_keys.is_set(tab->ref.key) && !table->no_keyread) { table->key_read=1; @@ -3573,14 +3583,14 @@ make_join_readinfo(JOIN *join, uint options) if (!table->no_keyread) { if (tab->select && tab->select->quick && - table->used_keys & ((key_map) 1 << tab->select->quick->index)) + table->used_keys.is_set(tab->select->quick->index)) { table->key_read=1; table->file->extra(HA_EXTRA_KEYREAD); } - else if (table->used_keys && ! (tab->select && tab->select->quick)) + else if (!table->used_keys.is_clear_all() && ! (tab->select && tab->select->quick)) { // Only read index tree - tab->index=find_shortest_key(table, table->used_keys); + tab->index=find_shortest_key(table, & table->used_keys); tab->table->file->index_init(tab->index); tab->read_first_record= join_read_first; tab->type=JT_NEXT; // Read with index_first / index_next @@ -3924,7 +3934,7 @@ return_zero_rows(JOIN *join, select_result *result,TABLE_LIST *tables, DBUG_ENTER("return_zero_rows"); if (select_options & SELECT_DESCRIBE) - { + { select_describe(join, false, false, false, info); DBUG_RETURN(0); } @@ -4654,6 +4664,12 @@ create_tmp_table(THD *thd,TMP_TABLE_PARAM *param,List<Item> &fields, table->db_low_byte_first=1; // True for HEAP and MyISAM table->temp_pool_slot = temp_pool_slot; table->copy_blobs= 1; + table->keys_for_keyread.init(); + table->keys_in_use.init(); + table->read_only_keys.init(); + table->quick_keys.init(); + table->used_keys.init(); + table->keys_in_use_for_query.init(); /* Calculate which type of fields we will store in the temporary table */ @@ -5853,7 +5869,7 @@ join_read_first(JOIN_TAB *tab) { int error; TABLE *table=tab->table; - if (!table->key_read && (table->used_keys & ((key_map) 1 << tab->index)) && + if (!table->key_read && table->used_keys.is_set(tab->index) && !table->no_keyread) { table->key_read=1; @@ -5890,7 +5906,7 @@ join_read_last(JOIN_TAB *tab) { TABLE *table=tab->table; int error; - if (!table->key_read && (table->used_keys & ((key_map) 1 << tab->index)) && + if (!table->key_read && table->used_keys.is_set(tab->index) && !table->no_keyread) { table->key_read=1; @@ -6588,18 +6604,21 @@ static int test_if_order_by_key(ORDER *order, TABLE *table, uint idx, return reverse; } -static uint find_shortest_key(TABLE *table, key_map usable_keys) +static uint find_shortest_key(TABLE *table, const key_map *usable_keys) { uint min_length= (uint) ~0; uint best= MAX_KEY; - for (uint nr=0; usable_keys ; usable_keys>>=1, nr++) + if (!usable_keys->is_clear_all()) { - if (usable_keys & 1) + for (uint nr=0; nr < table->keys ; nr++) { - if (table->key_info[nr].key_length < min_length) + if (usable_keys->is_set(nr)) { - min_length=table->key_info[nr].key_length; - best=nr; + if (table->key_info[nr].key_length < min_length) + { + min_length=table->key_info[nr].key_length; + best=nr; + } } } } @@ -6645,7 +6664,7 @@ is_subkey(KEY_PART_INFO *key_part, KEY_PART_INFO *ref_key_part, static uint test_if_subkey(ORDER *order, TABLE *table, uint ref, uint ref_key_parts, - key_map usable_keys) + const key_map& usable_keys) { uint nr; uint min_length= (uint) ~0; @@ -6653,10 +6672,10 @@ test_if_subkey(ORDER *order, TABLE *table, uint ref, uint ref_key_parts, uint not_used; KEY_PART_INFO *ref_key_part= table->key_info[ref].key_part; KEY_PART_INFO *ref_key_part_end= ref_key_part + ref_key_parts; - - for (nr= 0; usable_keys; usable_keys>>= 1, nr++) + + for (nr= 0 ; nr < table->keys ; nr++) { - if ((usable_keys & 1) && + if (usable_keys.is_set(nr) && table->key_info[nr].key_length < min_length && table->key_info[nr].key_parts >= ref_key_parts && is_subkey(table->key_info[nr].key_part, ref_key_part, @@ -6694,16 +6713,17 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, LINT_INIT(ref_key_parts); /* Check which keys can be used to resolve ORDER BY */ - usable_keys= ~(key_map) 0; + usable_keys.set_all(); for (ORDER *tmp_order=order; tmp_order ; tmp_order=tmp_order->next) { if ((*tmp_order->item)->type() != Item::FIELD_ITEM) { - usable_keys=0; + usable_keys.clear_all(); break; } - if (!(usable_keys&= (((Item_field*) (*tmp_order->item))->field-> - part_of_sortkey))) + usable_keys.intersect( + ((Item_field*) (*tmp_order->item))->field->part_of_sortkey); + if (usable_keys.is_clear_all()) break; // No usable keys } @@ -6729,7 +6749,7 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, */ int order_direction; uint used_key_parts; - if (!(usable_keys & ((key_map) 1 << ref_key))) + if (!usable_keys.is_set(ref_key)) { /* We come here when ref_key is not among usable_keys @@ -6739,8 +6759,8 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, If using index only read, only consider other possible index only keys */ - if (table->used_keys & (((key_map) 1 << ref_key))) - usable_keys|= table->used_keys; + if (table->used_keys.is_set(ref_key)) + usable_keys.merge(table->used_keys); if ((new_ref_key= test_if_subkey(order, table, ref_key, ref_key_parts, usable_keys)) < MAX_KEY) { @@ -6756,10 +6776,10 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, select->quick->init(); } ref_key= new_ref_key; - } + } } /* Check if we get the rows in requested sorted order by using the key */ - if ((usable_keys & ((key_map) 1 << ref_key)) && + if (usable_keys.is_set(ref_key) && (order_direction = test_if_order_by_key(order,table,ref_key, &used_key_parts))) { @@ -6810,7 +6830,7 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, /* check if we can use a key to resolve the group */ /* Tables using JT_NEXT are handled here */ uint nr; - key_map keys=usable_keys; + key_map keys; /* If not used with LIMIT, only use keys if the whole query can be @@ -6818,12 +6838,19 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, retrieving all rows through an index. */ if (select_limit >= table->file->records) - keys&= (table->used_keys | table->file->keys_to_use_for_scanning()); + { + keys=*table->file->keys_to_use_for_scanning(); + keys.merge(table->used_keys); + } + else + keys.set_all(); + + keys.intersect(usable_keys); - for (nr=0; keys ; keys>>=1, nr++) + for (nr=0; nr < table->keys ; nr++) { uint not_used; - if (keys & 1) + if (keys.is_set(nr)) { int flag; if ((flag=test_if_order_by_key(order, table, nr, ¬_used))) @@ -6835,7 +6862,7 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, join_read_last); table->file->index_init(nr); tab->type=JT_NEXT; // Read with index_first(), index_next() - if (table->used_keys & ((key_map) 1 << nr)) + if (table->used_keys.is_set(nr)) { table->key_read=1; table->file->extra(HA_EXTRA_KEYREAD); @@ -6860,7 +6887,7 @@ test_if_skip_sort_order(JOIN_TAB *tab,ORDER *order,ha_rows select_limit, order How table should be sorted filesort_limit Max number of rows that needs to be sorted select_limit Max number of rows in final output - Used to decide if we should use index or not + Used to decide if we should use index or not IMPLEMENTATION @@ -8690,7 +8717,7 @@ static void select_describe(JOIN *join, bool need_tmp_table, bool need_order, Item *item_null= new Item_null(); CHARSET_INFO *cs= &my_charset_latin1; DBUG_ENTER("select_describe"); - DBUG_PRINT("info", ("Select 0x%lx, type %s, message %s", + DBUG_PRINT("info", ("Select 0x%lx, type %s, message %s", (ulong)join->select_lex, join->select_lex->type, message)); /* Don't log this into the slow query log */ @@ -8724,7 +8751,7 @@ static void select_describe(JOIN *join, bool need_tmp_table, bool need_order, tmp2.length(0); item_list.empty(); - item_list.push_back(new Item_int((int32) + item_list.push_back(new Item_int((int32) join->select_lex->select_number)); item_list.push_back(new Item_string(join->select_lex->type, strlen(join->select_lex->type), @@ -8746,21 +8773,23 @@ static void select_describe(JOIN *join, bool need_tmp_table, bool need_order, item_list.push_back(new Item_string(join_type_str[tab->type], strlen(join_type_str[tab->type]), cs)); - key_map bits; uint j; - for (j=0,bits=tab->keys ; bits ; j++,bits>>=1) + if (!tab->keys.is_clear_all()) { - if (bits & 1) - { - if (tmp1.length()) - tmp1.append(','); - tmp1.append(table->key_info[j].name); - } + for (j=0 ; j < table->keys ; j++) + { + if (tab->keys.is_set(j)) + { + if (tmp1.length()) + tmp1.append(','); + tmp1.append(table->key_info[j].name); + } + } } if (tmp1.length()) item_list.push_back(new Item_string(tmp1.ptr(),tmp1.length(),cs)); else - item_list.push_back(item_null); + item_list.push_back(item_null); if (tab->ref.key_parts) { KEY *key_info=table->key_info+ tab->ref.key; @@ -8803,10 +8832,9 @@ static void select_describe(JOIN *join, bool need_tmp_table, bool need_order, join->best_positions[i]. records_read, 21)); my_bool key_read=table->key_read; - if (tab->type == JT_NEXT && - ((table->used_keys & ((key_map) 1 << tab->index)))) + if (tab->type == JT_NEXT && table->used_keys.is_set(tab->index)) key_read=1; - + if (tab->info) item_list.push_back(new Item_string(tab->info,strlen(tab->info),cs)); else @@ -8815,8 +8843,9 @@ static void select_describe(JOIN *join, bool need_tmp_table, bool need_order, { if (tab->use_quick == 2) { - sprintf(buff_ptr,"; Range checked for each record (index map: %u)", - tab->keys); + char buf[MAX_KEY/8+1]; + sprintf(buff_ptr,"; Range checked for each record (index map: 0x%s)", + tab->keys.print(buf)); buff_ptr=strend(buff_ptr); } else diff --git a/sql/sql_select.h b/sql/sql_select.h index 6c17a646ee6..aa77722546d 100644 --- a/sql/sql_select.h +++ b/sql/sql_select.h @@ -29,8 +29,8 @@ typedef struct keyuse_t { Item *val; /* or value if no field */ table_map used_tables; uint key, keypart, optimize; - key_map keypart_map; - ha_rows ref_table_rows; + key_part_map keypart_map; + ha_rows ref_table_rows; } KEYUSE; class store_key; @@ -96,9 +96,9 @@ typedef struct st_join_table { key_map const_keys; /* Keys with constant part */ key_map checked_keys; /* Keys checked in find_best */ key_map needed_reg; + key_map keys; /* all keys with can be used */ ha_rows records,found_records,read_time; table_map dependent,key_dependent; - uint keys; /* all keys with can be used */ uint use_quick,index; uint status; // Save status for cache uint used_fields,used_fieldlength,used_blobs; diff --git a/sql/sql_show.cc b/sql/sql_show.cc index c667824224e..4f453bce903 100644 --- a/sql/sql_show.cc +++ b/sql/sql_show.cc @@ -938,7 +938,7 @@ mysqld_show_keys(THD *thd, TABLE_LIST *table_list) protocol->store((const char*) pos, system_charset_info); protocol->store(table->file->index_type(i), system_charset_info); /* Comment */ - if (!(table->keys_in_use & ((key_map) 1 << i))) + if (!table->keys_in_use.is_set(i)) protocol->store("disabled",8, system_charset_info); else protocol->store("", 0, system_charset_info); diff --git a/sql/sql_table.cc b/sql/sql_table.cc index ba172aa6bda..1a2021abf90 100644 --- a/sql/sql_table.cc +++ b/sql/sql_table.cc @@ -424,7 +424,7 @@ int mysql_create_table(THD *thd,const char *db, const char *table_name, if (!sql_field->charset) sql_field->charset= create_info->table_charset; sql_field->create_length_to_internal_length(); - + /* Don't pack keys in old tables if the user has requested this */ if ((sql_field->flags & BLOB_FLAG) || sql_field->sql_type == FIELD_TYPE_VAR_STRING && @@ -663,12 +663,11 @@ int mysql_create_table(THD *thd,const char *db, const char *table_name, key_info->usable_key_parts= key_number; key_info->algorithm=key->algorithm; - /* TODO: Add proper checks if handler supports key_type and algorithm */ if (key->type == Key::FULLTEXT) { if (!(file->table_flags() & HA_CAN_FULLTEXT)) { - my_error(ER_TABLE_CANT_HANDLE_FULLTEXT, MYF(0)); + my_error(ER_TABLE_CANT_HANDLE_FT, MYF(0)); DBUG_RETURN(-1); } } @@ -680,6 +679,7 @@ int mysql_create_table(THD *thd,const char *db, const char *table_name, checking for proper key parts number: */ + /* TODO: Add proper checks if handler supports key_type and algorithm */ if (key_info->flags == HA_SPATIAL) { if (key_info->key_parts != 1) @@ -704,6 +704,7 @@ int mysql_create_table(THD *thd,const char *db, const char *table_name, } List_iterator<key_part_spec> cols(key->columns); + CHARSET_INFO *ft_key_charset=0; // for FULLTEXT for (uint column_nr=0 ; (column=cols++) ; column_nr++) { it.rewind(); @@ -727,64 +728,88 @@ int mysql_create_table(THD *thd,const char *db, const char *table_name, from a data prefix, ignoring column->length). */ if (key->type == Key::FULLTEXT) + { + if ((sql_field->sql_type != FIELD_TYPE_STRING && + sql_field->sql_type != FIELD_TYPE_VAR_STRING && + !f_is_blob(sql_field->pack_flag)) || + sql_field->charset == &my_charset_bin || + sql_field->charset->state & MY_CS_NONTEXT || // ucs2 doesn't work yet + (ft_key_charset && sql_field->charset != ft_key_charset)) + { + my_printf_error(ER_BAD_FT_COLUMN,ER(ER_BAD_FT_COLUMN),MYF(0), + column->field_name); + DBUG_RETURN(-1); + } + ft_key_charset=sql_field->charset; + /* + for fulltext keys keyseg length is 1 for blobs (it's ignored in ft + code anyway, and 0 (set to column width later) for char's. it has + to be correct col width for char's, as char data are not prefixed + with length (unlike blobs, where ft code takes data length from a + data prefix, ignoring column->length). + */ column->length=test(f_is_blob(sql_field->pack_flag)); + } else + { column->length*= sql_field->charset->mbmaxlen; - if (f_is_blob(sql_field->pack_flag)) - { - if (!(file->table_flags() & HA_BLOB_KEY)) - { - my_printf_error(ER_BLOB_USED_AS_KEY,ER(ER_BLOB_USED_AS_KEY),MYF(0), - column->field_name); - DBUG_RETURN(-1); - } - if (!column->length) - { - my_printf_error(ER_BLOB_KEY_WITHOUT_LENGTH, - ER(ER_BLOB_KEY_WITHOUT_LENGTH),MYF(0), - column->field_name); - DBUG_RETURN(-1); - } - } - if (key->type == Key::SPATIAL) - { - if (!column->length ) - { - /* - BAR: 4 is: (Xmin,Xmax,Ymin,Ymax), this is for 2D case - Lately we'll extend this code to support more dimensions - */ - column->length=4*sizeof(double); - } - } - if (!(sql_field->flags & NOT_NULL_FLAG)) - { - if (key->type == Key::PRIMARY) - { - /* Implicitly set primary key fields to NOT NULL for ISO conf. */ - sql_field->flags|= NOT_NULL_FLAG; - sql_field->pack_flag&= ~FIELDFLAG_MAYBE_NULL; - } - else - key_info->flags|= HA_NULL_PART_KEY; - if (!(file->table_flags() & HA_NULL_KEY)) - { - my_printf_error(ER_NULL_COLUMN_IN_INDEX,ER(ER_NULL_COLUMN_IN_INDEX), - MYF(0),column->field_name); - DBUG_RETURN(-1); - } - if (key->type == Key::SPATIAL) - { - my_error(ER_SPATIAL_CANT_HAVE_NULL, MYF(0)); - DBUG_RETURN(-1); - } - } - if (MTYP_TYPENR(sql_field->unireg_check) == Field::NEXT_NUMBER) - { - if (column_nr == 0 || (file->table_flags() & HA_AUTO_PART_KEY)) - auto_increment--; // Field is used + if (f_is_blob(sql_field->pack_flag)) + { + if (!(file->table_flags() & HA_BLOB_KEY)) + { + my_printf_error(ER_BLOB_USED_AS_KEY,ER(ER_BLOB_USED_AS_KEY),MYF(0), + column->field_name); + DBUG_RETURN(-1); + } + if (!column->length) + { + my_printf_error(ER_BLOB_KEY_WITHOUT_LENGTH, + ER(ER_BLOB_KEY_WITHOUT_LENGTH),MYF(0), + column->field_name); + DBUG_RETURN(-1); + } + } + if (key->type == Key::SPATIAL) + { + if (!column->length ) + { + /* + BAR: 4 is: (Xmin,Xmax,Ymin,Ymax), this is for 2D case + Lately we'll extend this code to support more dimensions + */ + column->length=4*sizeof(double); + } + } + if (!(sql_field->flags & NOT_NULL_FLAG)) + { + if (key->type == Key::PRIMARY) + { + /* Implicitly set primary key fields to NOT NULL for ISO conf. */ + sql_field->flags|= NOT_NULL_FLAG; + sql_field->pack_flag&= ~FIELDFLAG_MAYBE_NULL; + } + else + key_info->flags|= HA_NULL_PART_KEY; + if (!(file->table_flags() & HA_NULL_KEY)) + { + my_printf_error(ER_NULL_COLUMN_IN_INDEX,ER(ER_NULL_COLUMN_IN_INDEX), + MYF(0),column->field_name); + DBUG_RETURN(-1); + } + if (key->type == Key::SPATIAL) + { + my_error(ER_SPATIAL_CANT_HAVE_NULL, MYF(0)); + DBUG_RETURN(-1); + } + } + if (MTYP_TYPENR(sql_field->unireg_check) == Field::NEXT_NUMBER) + { + if (column_nr == 0 || (file->table_flags() & HA_AUTO_PART_KEY)) + auto_increment--; // Field is used + } } + key_part_info->fieldnr= field; key_part_info->offset= (uint16) sql_field->offset; key_part_info->key_type=sql_field->pack_flag; diff --git a/sql/sql_test.cc b/sql/sql_test.cc index 0ffdd5f1bcf..7e3513367fd 100644 --- a/sql/sql_test.cc +++ b/sql/sql_test.cc @@ -173,10 +173,11 @@ TEST_join(JOIN *join) tab->ref.key_length); if (tab->select) { + char buf[MAX_KEY/8+1]; if (tab->use_quick == 2) fprintf(DBUG_FILE, - " quick select checked for each record (keys: %d)\n", - (int) tab->select->quick_keys); + " quick select checked for each record (keys: %s)\n", + tab->select->quick_keys.print(buf)); else if (tab->select->quick) fprintf(DBUG_FILE," quick select used on key %s, length: %d\n", form->key_info[tab->select->quick->index].name, diff --git a/sql/sql_update.cc b/sql/sql_update.cc index e14e20b62d6..9214894e214 100644 --- a/sql/sql_update.cc +++ b/sql/sql_update.cc @@ -85,7 +85,7 @@ int mysql_update(THD *thd, /* Calculate "table->used_keys" based on the WHERE */ table->used_keys=table->keys_in_use; - table->quick_keys=0; + table->quick_keys.clear_all(); #ifndef NO_EMBEDDED_ACCESS_CHECKS want_privilege=table->grant.want_privilege; table->grant.want_privilege=(SELECT_ACL & ~table->grant.privilege); @@ -149,7 +149,7 @@ int mysql_update(THD *thd, } // Don't count on usage of 'only index' when calculating which key to use - table->used_keys=0; + table->used_keys.clear_all(); select=make_select(table,0,0,conds,&error); if (error || (select && select->check_quick(thd, safe_update, limit)) || !limit) @@ -164,7 +164,7 @@ int mysql_update(THD *thd, DBUG_RETURN(0); } /* If running in safe sql mode, don't allow updates without keys */ - if (!table->quick_keys) + if (table->quick_keys.is_clear_all()) { thd->lex.select_lex.options|=QUERY_NO_INDEX_USED; if (safe_update && !using_limit) @@ -192,7 +192,7 @@ int mysql_update(THD *thd, matching rows before updating the table! */ table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS); - if (old_used_keys & ((key_map) 1 << used_index)) + if (old_used_keys.is_set(used_index)) { table->key_read=1; table->file->extra(HA_EXTRA_KEYREAD); @@ -501,8 +501,7 @@ int multi_update::prepare(List<Item> ¬_used_values, SELECT_LEX_UNIT *unit) if (!tables_to_update) { - my_error(ER_NOT_SUPPORTED_YET, MYF(0), - "You didn't specify any tables to UPDATE"); + my_error(ER_NO_TABLES_USED, MYF(0)); DBUG_RETURN(1); } @@ -533,7 +532,7 @@ int multi_update::prepare(List<Item> ¬_used_values, SELECT_LEX_UNIT *unit) update.link_in_list((byte*) tl, (byte**) &tl->next); tl->shared= table_count++; table->no_keyread=1; - table->used_keys=0; + table->used_keys.clear_all(); table->pos_in_table_list= tl; } } diff --git a/sql/sql_yacc.yy b/sql/sql_yacc.yy index c2a014f915d..00277759763 100644 --- a/sql/sql_yacc.yy +++ b/sql/sql_yacc.yy @@ -135,6 +135,7 @@ bool my_yyoverflow(short **a, YYSTYPE **b,int *yystacksize); %token DROP %token EVENTS_SYM %token EXECUTE_SYM +%token EXPANSION_SYM %token FLUSH_SYM %token HELP_SYM %token INSERT @@ -168,7 +169,7 @@ bool my_yyoverflow(short **a, YYSTYPE **b,int *yystacksize); %token SUPER_SYM %token TRUNCATE_SYM %token UNLOCK_SYM -%token UNTIL_SYM +%token UNTIL_SYM %token UPDATE_SYM %token ACTION @@ -604,8 +605,9 @@ bool my_yyoverflow(short **a, YYSTYPE **b,int *yystacksize); %type <num> type int_type real_type order_dir opt_field_spec lock_option udf_type if_exists opt_local opt_table_options table_options - table_option opt_if_not_exists opt_no_write_to_binlog opt_var_type opt_var_ident_type - delete_option opt_temporary all_or_any opt_distinct opt_ignore_leaves + table_option opt_if_not_exists opt_no_write_to_binlog opt_var_type + opt_var_ident_type delete_option opt_temporary all_or_any opt_distinct + opt_ignore_leaves fulltext_options %type <ulong_num> ULONG_NUM raid_types merge_insert_types @@ -2473,12 +2475,10 @@ simple_expr: | EXISTS exists_subselect { $$= $2; } | singlerow_subselect { $$= $1; } | '{' ident expr '}' { $$= $3; } - | MATCH ident_list_arg AGAINST '(' expr ')' - { Select->add_ftfunc_to_list((Item_func_match *) - ($$=new Item_func_match_nl(*$2,$5))); } - | MATCH ident_list_arg AGAINST '(' expr IN_SYM BOOLEAN_SYM MODE_SYM ')' - { Select->add_ftfunc_to_list((Item_func_match *) - ($$=new Item_func_match_bool(*$2,$5))); } + | MATCH ident_list_arg AGAINST '(' expr fulltext_options ')' + { $2->push_front($5); + Select->add_ftfunc_to_list((Item_func_match*) + ($$=new Item_func_match(*$2,$6))); } | ASCII_SYM '(' expr ')' { $$= new Item_func_ascii($3); } | BINARY expr %prec NEG { @@ -2486,10 +2486,10 @@ simple_expr: 6, &my_charset_latin1)); } | CAST_SYM '(' expr AS cast_type ')' - { - $$= create_func_cast($3, $5, + { + $$= create_func_cast($3, $5, Lex->length ? atoi(Lex->length) : -1, - Lex->charset); + Lex->charset); } | CASE_SYM opt_expr WHEN_SYM when_list opt_else END { $$= new Item_func_case(* $4, $2, $5 ); } @@ -2866,6 +2866,12 @@ simple_expr: | EXTRACT_SYM '(' interval FROM expr ')' { $$=new Item_extract( $3, $5); }; +fulltext_options: + /* nothing */ { $$= FT_NL; } + | WITH QUERY_SYM EXPANSION_SYM { $$= FT_NL | FT_EXPAND; } + | IN_SYM BOOLEAN_SYM MODE_SYM { $$= FT_BOOL; } + ; + udf_expr_list: /* empty */ { $$= NULL; } | expr_list { $$= $1;}; @@ -2992,7 +2998,7 @@ ident_list2: opt_expr: /* empty */ { $$= NULL; } - | expr { $$= $1; }; + | expr { $$= $1; }; opt_else: /* empty */ { $$= NULL; } @@ -4565,6 +4571,7 @@ keyword: | ESCAPE_SYM {} | EVENTS_SYM {} | EXECUTE_SYM {} + | EXPANSION_SYM {} | EXTENDED_SYM {} | FAST_SYM {} | DISABLE_SYM {} diff --git a/sql/table.cc b/sql/table.cc index be5b7edfd93..14853011510 100644 --- a/sql/table.cc +++ b/sql/table.cc @@ -137,10 +137,11 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, outparam->raid_type= head[41]; outparam->raid_chunks= head[42]; outparam->raid_chunksize= uint4korr(head+43); - if (!(outparam->table_charset=get_charset((uint) head[38],MYF(0)))) - outparam->table_charset=default_charset_info; // QQ display error message? + outparam->table_charset=get_charset((uint) head[38],MYF(0)); null_field_first=1; } + if (!outparam->table_charset) /* unknown charset in head[38] or pre-3.23 frm */ + outparam->table_charset=default_charset_info; outparam->db_record_offset=1; if (db_create_options & HA_OPTION_LONG_BLOB_PTR) outparam->blob_ptr_size=portable_sizeof_char_ptr; @@ -155,10 +156,23 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, VOID(my_seek(file,(ulong) uint2korr(head+6),MY_SEEK_SET,MYF(0))); if (read_string(file,(gptr*) &disk_buff,key_info_length)) goto err_not_open; /* purecov: inspected */ - outparam->keys=keys= disk_buff[0]; - outparam->keys_for_keyread= outparam->keys_in_use= set_bits(key_map, keys); + if (disk_buff[0] & 0x80) + { + outparam->keys= keys= (disk_buff[1] << 7) | (disk_buff[0] & 0x7f); + outparam->key_parts= key_parts= uint2korr(disk_buff+2); + } + else + { + outparam->keys= keys= disk_buff[0]; + outparam->key_parts= key_parts= disk_buff[1]; + } + outparam->keys_for_keyread.init(keys); + outparam->keys_in_use.init(keys); + outparam->read_only_keys.init(0); + outparam->quick_keys.init(); + outparam->used_keys.init(); + outparam->keys_in_use_for_query.init(); - outparam->key_parts=key_parts=disk_buff[1]; n_length=keys*sizeof(KEY)+key_parts*sizeof(KEY_PART_INFO); if (!(keyinfo = (KEY*) alloc_root(&outparam->mem_root, n_length+uint2korr(disk_buff+4)))) @@ -263,7 +277,9 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, record[outparam->reclength]=0; // For purify and ->c_ptr() outparam->rec_buff_length=rec_buff_length; if (my_pread(file,(byte*) record,(uint) outparam->reclength, - (ulong) (uint2korr(head+6)+uint2korr(head+14)), + (ulong) (uint2korr(head+6)+ + ((uint2korr(head+14) == 0xffff ? + uint4korr(head+47) : uint2korr(head+14)))), MYF(MY_NABP))) goto err_not_open; /* purecov: inspected */ /* HACK: table->record[2] is used instead of table->default_values here */ @@ -484,8 +500,8 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, index_flags=outparam->file->index_flags(key); if (!(index_flags & HA_KEY_READ_ONLY)) { - outparam->read_only_keys|= ((key_map) 1 << key); - outparam->keys_for_keyread&= ~((key_map) 1 << key); + outparam->read_only_keys.set_bit(key); + outparam->keys_for_keyread.clear_bit(key); } if (primary_key >= MAX_KEY && (keyinfo->flags & HA_NOSAME)) @@ -545,7 +561,7 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, field->key_length() == keyinfo->key_length ? UNIQUE_KEY_FLAG : MULTIPLE_KEY_FLAG); if (i == 0) - field->key_start|= ((key_map) 1 << key); + field->key_start.set_bit(key); if (field->key_length() == key_part->length && !(field->flags & BLOB_FLAG)) { @@ -553,11 +569,11 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, (field->key_type() != HA_KEYTYPE_TEXT || (!(ha_option & HA_KEY_READ_WRONG_STR) && !(keyinfo->flags & HA_FULLTEXT)))) - field->part_of_key|= ((key_map) 1 << key); + field->part_of_key.set_bit(key); if ((field->key_type() != HA_KEYTYPE_TEXT || !(keyinfo->flags & HA_FULLTEXT)) && !(index_flags & HA_WRONG_ASCII_ORDER)) - field->part_of_sortkey|= ((key_map) 1 << key); + field->part_of_sortkey.set_bit(key); } if (!(key_part->key_part_flag & HA_REVERSE_SORT) && usable_parts == i) @@ -600,7 +616,7 @@ int openfrm(const char *name, const char *alias, uint db_stat, uint prgflag, keyinfo->usable_key_parts=usable_parts; // Filesort } if (primary_key < MAX_KEY && - (outparam->keys_in_use & ((key_map) 1 << primary_key))) + (outparam->keys_in_use.is_set(primary_key))) { outparam->primary_key=primary_key; /* @@ -1108,6 +1124,7 @@ File create_frm(register my_string name, uint reclength, uchar *fileinfo, key_length=keys*(7+NAME_LEN+MAX_REF_PARTS*9)+16; length=(ulong) next_io_size((ulong) (IO_SIZE+key_length+reclength)); int4store(fileinfo+10,length); + if (key_length > 0xffff) key_length=0xffff; int2store(fileinfo+14,key_length); int2store(fileinfo+16,reclength); int4store(fileinfo+18,create_info->max_rows); diff --git a/sql/unireg.cc b/sql/unireg.cc index 3e634f54b4f..8606830e450 100644 --- a/sql/unireg.cc +++ b/sql/unireg.cc @@ -33,7 +33,7 @@ static uchar * pack_screens(List<create_field> &create_fields, uint *info_length, uint *screens, bool small_file); static uint pack_keys(uchar *keybuff,uint key_count, KEY *key_info); -static bool pack_header(uchar *forminfo, enum db_type table_type, +static bool pack_header(uchar *forminfo,enum db_type table_type, List<create_field> &create_fields, uint info_length, uint screens, uint table_options, handler *file); @@ -92,8 +92,8 @@ int rea_create_table(THD *thd, my_string file_name, DBUG_RETURN(1); } - uint key_buff_length=uint2korr(fileinfo+14); - keybuff=(uchar*) my_alloca(key_buff_length); + uint key_buff_length=keys*(7+NAME_LEN+MAX_REF_PARTS*9)+16; + keybuff=(uchar*) my_malloc(key_buff_length, MYF(0)); key_info_length=pack_keys(keybuff,keys,key_info); VOID(get_form_pos(file,fileinfo,&formnames)); if (!(filepos=make_new_entry(file,fileinfo,&formnames,""))) @@ -101,6 +101,7 @@ int rea_create_table(THD *thd, my_string file_name, maxlength=(uint) next_io_size((ulong) (uint2korr(forminfo)+1000)); int2store(forminfo+2,maxlength); int4store(fileinfo+10,(ulong) (filepos+maxlength)); + int4store(fileinfo+47,key_buff_length); fileinfo[26]= (uchar) test((create_info->max_rows == 1) && (create_info->min_rows == 1) && (keys == 0)); int2store(fileinfo+28,key_info_length); @@ -148,7 +149,7 @@ int rea_create_table(THD *thd, my_string file_name, #endif my_free((gptr) screen_buff,MYF(0)); - my_afree((gptr) keybuff); + my_free((gptr) keybuff, MYF(0)); VOID(my_close(file,MYF(MY_WME))); if (ha_create_table(file_name,create_info,0)) goto err2; @@ -156,7 +157,7 @@ int rea_create_table(THD *thd, my_string file_name, err: my_free((gptr) screen_buff,MYF(0)); - my_afree((gptr) keybuff); + my_free((gptr) keybuff, MYF(0)); VOID(my_close(file,MYF(MY_WME))); err2: my_delete(file_name,MYF(0)); @@ -291,10 +292,17 @@ static uint pack_keys(uchar *keybuff,uint key_count,KEY *keyinfo) } *(pos++)=0; - keybuff[0]=(uchar) key_count; - keybuff[1]=(uchar) key_parts; - length=(uint) (keyname_pos-keybuff); - int2store(keybuff+2,length); + if (key_count > 127 || key_parts > 127) + { + keybuff[0]= (key_count & 0x7f) | 0x80; + keybuff[1]= key_count >> 7; + int2store(keybuff+2,key_parts); + } + else + { + keybuff[0]=(uchar) key_count; + keybuff[1]=(uchar) key_parts; + } length=(uint) (pos-keyname_pos); int2store(keybuff+4,length); DBUG_RETURN((uint) (pos-keybuff)); diff --git a/strings/ctype-bin.c b/strings/ctype-bin.c index ebf915398a8..5a785552b45 100644 --- a/strings/ctype-bin.c +++ b/strings/ctype-bin.c @@ -118,6 +118,12 @@ static int my_strcasecmp_bin(CHARSET_INFO * cs __attribute__((unused)), return strcmp(s,t); } +int my_mbcharlen_8bit(CHARSET_INFO *cs __attribute__((unused)), + uint c __attribute__((unused))) +{ + return 1; +} + static int my_mb_wc_bin(CHARSET_INFO *cs __attribute__((unused)), my_wc_t *wc, const unsigned char *str, @@ -264,12 +270,12 @@ static int my_strnxfrm_bin(CHARSET_INFO *cs __attribute__((unused)), static uint my_instr_bin(CHARSET_INFO *cs __attribute__((unused)), - const char *b, uint b_length, - const char *s, uint s_length, - my_match_t *match, uint nmatch) + const char *b, uint b_length, + const char *s, uint s_length, + my_match_t *match, uint nmatch) { register const uchar *str, *search, *end, *search_end; - + if (s_length <= b_length) { if (!s_length) @@ -282,32 +288,32 @@ uint my_instr_bin(CHARSET_INFO *cs __attribute__((unused)), } return 1; /* Empty string is always found */ } - + str= (const uchar*) b; search= (const uchar*) s; end= (const uchar*) b+b_length-s_length+1; search_end= (const uchar*) s + s_length; - + skipp: while (str != end) { if ( (*str++) == (*search)) { register const uchar *i,*j; - - i= str; + + i= str; j= search+1; - + while (j != search_end) if ((*i++) != (*j++)) goto skipp; - + if (nmatch > 0) { match[0].beg= 0; match[0].end= str- (const uchar*)b-1; match[0].mblen= match[0].end; - + if (nmatch > 1) { match[1].beg= match[0].end; @@ -338,7 +344,7 @@ MY_COLLATION_HANDLER my_collation_8bit_bin_handler = static MY_CHARSET_HANDLER my_charset_handler= { NULL, /* ismbchar */ - NULL, /* mbcharlen */ + my_mbcharlen_8bit, /* mbcharlen */ my_numchars_8bit, my_charpos_8bit, my_lengthsp_8bit, diff --git a/strings/ctype-latin1.c b/strings/ctype-latin1.c index a8a5329f844..15798abb85b 100644 --- a/strings/ctype-latin1.c +++ b/strings/ctype-latin1.c @@ -170,14 +170,14 @@ int my_wc_mb_latin1(CHARSET_INFO *cs __attribute__((unused)), { if (str >= end) return MY_CS_TOOSMALL; - + return ((wc < 256) && ((str[0]=uni_latin1[wc]) || (!wc))) ? 1 : MY_CS_ILUNI; } static MY_CHARSET_HANDLER my_charset_handler= { NULL, - NULL, + my_mbcharlen_8bit, my_numchars_8bit, my_charpos_8bit, my_lengthsp_8bit, diff --git a/strings/ctype-simple.c b/strings/ctype-simple.c index aba15f13d4b..6f77d6f3e16 100644 --- a/strings/ctype-simple.c +++ b/strings/ctype-simple.c @@ -1093,7 +1093,7 @@ skipp: MY_CHARSET_HANDLER my_charset_8bit_handler= { NULL, /* ismbchar */ - NULL, /* mbcharlen */ + my_mbcharlen_8bit, /* mbcharlen */ my_numchars_8bit, my_charpos_8bit, my_lengthsp_8bit, diff --git a/strings/ctype-tis620.c b/strings/ctype-tis620.c index a4d8a7d1f79..fd5e58ad8a7 100644 --- a/strings/ctype-tis620.c +++ b/strings/ctype-tis620.c @@ -717,7 +717,7 @@ static MY_COLLATION_HANDLER my_collation_ci_handler = static MY_CHARSET_HANDLER my_charset_handler= { NULL, /* ismbchar */ - NULL, /* mbcharlen */ + my_mbcharlen_8bit, /* mbcharlen */ my_numchars_8bit, my_charpos_8bit, my_lengthsp_8bit, diff --git a/strings/ctype-utf8.c b/strings/ctype-utf8.c index 3ede1aa26f6..b5716c53ea2 100644 --- a/strings/ctype-utf8.c +++ b/strings/ctype-utf8.c @@ -1540,10 +1540,10 @@ static uchar ctype_utf8[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 + 0, 0, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 0 }; static uchar to_lower_utf8[] = { diff --git a/vio/viosocket.c b/vio/viosocket.c index f631dca5268..31941a64102 100644 --- a/vio/viosocket.c +++ b/vio/viosocket.c @@ -192,7 +192,7 @@ vio_should_retry(Vio * vio __attribute__((unused))) int vio_close(Vio * vio) { - int r; + int r=0; DBUG_ENTER("vio_close"); #ifdef __WIN__ if (vio->type == VIO_TYPE_NAMEDPIPE) @@ -206,7 +206,6 @@ int vio_close(Vio * vio) else if (vio->type != VIO_CLOSED) #endif /* __WIN__ */ { - r=0; if (shutdown(vio->sd,2)) r= -1; if (closesocket(vio->sd)) @@ -369,9 +368,9 @@ int vio_read_shared_memory(Vio * vio, gptr buf, int size) remain_local = size; current_postion=buf; - do + do { - if (vio->shared_memory_remain == 0) + if (vio->shared_memory_remain == 0) { if (WaitForSingleObject(vio->event_server_wrote,vio->net->read_timeout*1000) != WAIT_OBJECT_0) { @@ -384,9 +383,9 @@ int vio_read_shared_memory(Vio * vio, gptr buf, int size) length = size; - if (vio->shared_memory_remain < length) + if (vio->shared_memory_remain < length) length = vio->shared_memory_remain; - if (length > remain_local) + if (length > remain_local) length = remain_local; memcpy(current_postion,vio->shared_memory_pos,length); @@ -396,7 +395,7 @@ int vio_read_shared_memory(Vio * vio, gptr buf, int size) current_postion+=length; remain_local-=length; - if (!vio->shared_memory_remain) + if (!vio->shared_memory_remain) if (!SetEvent(vio->event_client_read)) DBUG_RETURN(-1); } while (remain_local); length = size; @@ -419,11 +418,11 @@ int vio_write_shared_memory(Vio * vio, const gptr buf, int size) remain = size; current_postion = buf; - while (remain != 0) + while (remain != 0) { if (WaitForSingleObject(vio->event_server_read,vio->net->write_timeout*1000) != WAIT_OBJECT_0) { - DBUG_RETURN(-1); + DBUG_RETURN(-1); }; sz = remain > shared_memory_buffer_length ? shared_memory_buffer_length: remain; |