Commit 05f1e100 authored by monty@mysql.com's avatar monty@mysql.com

merge with 4.0

parents 7c725b24 5443cba3
......@@ -61,10 +61,11 @@ typedef struct st_ftb_expr FTB_EXPR;
struct st_ftb_expr
{
FTB_EXPR *up;
byte *quot, *qend;
my_off_t docid[2]; /* for index search and for scan */
my_off_t docid[2];
/* ^^^^^^^^^^^^^^^^^^ FTB_{EXPR,WORD} common section */
float weight;
float cur_weight;
byte *quot, *qend;
uint flags;
uint yesses; /* number of "yes" words matched */
uint nos; /* number of "no" words matched */
......@@ -77,6 +78,7 @@ typedef struct st_ftb_word
FTB_EXPR *up;
MI_KEYDEF *keyinfo;
my_off_t docid[2]; /* for index search and for scan */
/* ^^^^^^^^^^^^^^^^^^ FTB_{EXPR,WORD} common section */
my_off_t key_root;
float weight;
uint ndepth;
......@@ -299,13 +301,15 @@ static void _ftb_init_index_search(FT_INFO *ftb)
if (ftbw->flags & FTB_FLAG_TRUNC)
{
/*
special treatment for truncation operator :((
1. +trunc* and there're other (not +trunc*) words
special treatment for truncation operator
1. there are some (besides this) +words
| no need to search in the index, it can never ADD new rows
| to the result, and to remove half-matched rows we do scan anyway
2. -trunc*
| same as 1.
3. trunc*
3. in 1 and 2, +/- need not be on the same expr. level,
but can be on any upper level, as in +word +(trunc1* trunc2*)
4. otherwise
| We have to index-search for this prefix.
| It may cause duplicates, as in the index (sorted by <word,docid>)
| <aaaa,row1>
......@@ -313,23 +317,32 @@ static void _ftb_init_index_search(FT_INFO *ftb)
| <aacc,row1>
| Searching for "aa*" will find row1 twice...
*/
if ( test(ftbw->flags&FTB_FLAG_NO) || /* 2 */
(test(ftbw->flags&FTB_FLAG_YES) && /* 1 */
ftbw->up->ythresh - ftbw->up->yweaks >1)) /* 1 */
FTB_EXPR *ftbe;
for (ftbe=(FTB_EXPR*)ftbw;
ftbe->up && !(ftbe->up->flags & FTB_FLAG_TRUNC);
ftbe->up->flags|= FTB_FLAG_TRUNC, ftbe=ftbe->up)
{
if (ftbe->flags & FTB_FLAG_NO || /* 2 */
ftbe->up->ythresh - ftbe->up->yweaks >1) /* 1 */
{
FTB_EXPR *top_ftbe=ftbe->up->up;
ftbw->docid[0]=HA_POS_ERROR;
ftbw->up->yweaks++;
continue;
for (ftbe=ftbw->up; ftbe != top_ftbe; ftbe=ftbe->up)
if (ftbe->flags & FTB_FLAG_YES)
ftbe->yweaks++;
ftbe=0;
break;
}
else /* 3 */
{
}
if (!ftbe)
continue;
/* 3 */
if (!is_tree_inited(& ftb->no_dupes))
init_tree(&ftb->no_dupes,0,0,sizeof(my_off_t),
_ftb_no_dupes_cmp, 0, NULL, NULL);
init_tree(& ftb->no_dupes,0,0,sizeof(my_off_t),
_ftb_no_dupes_cmp,0,0,0);
else
reset_tree(& ftb->no_dupes);
}
}
if (_ft2_search(ftb, ftbw, 1))
return;
......
......@@ -161,6 +161,23 @@ Only MyISAM tables support collections
Function MATCH ... AGAINST() is used to do a search
some test foobar implements vector space model
drop table t1;
create table t1 (a varchar(200) not null, fulltext (a));
insert t1 values ("aaa10 bbb20"), ("aaa20 bbb15"), ("aaa30 bbb10");
select * from t1 where match a against ("+aaa* +bbb*" in boolean mode);
a
aaa10 bbb20
aaa20 bbb15
aaa30 bbb10
select * from t1 where match a against ("+aaa* +bbb1*" in boolean mode);
a
aaa20 bbb15
aaa30 bbb10
select * from t1 where match a against ("+aaa* +ccc*" in boolean mode);
a
select * from t1 where match a against ("+aaa10 +(bbb*)" in boolean mode);
a
aaa10 bbb20
drop table t1;
CREATE TABLE t1 (
id int(11),
ticket int(11),
......
......@@ -41,4 +41,12 @@ venue_id venue_text dt name entity_id
select * from t1 left join t2 on venue_id = entity_id where match(name) against('aberdeen') and dt = '2003-05-23 19:30:00';
venue_id venue_text dt name entity_id
1 a1 2003-05-23 19:30:00 aberdeen town hall 1
select * from t1 left join t2 on (venue_id = entity_id and match(name) against('aberdeen' in boolean mode)) where dt = '2003-05-23 19:30:00';
venue_id venue_text dt name entity_id
1 a1 2003-05-23 19:30:00 aberdeen town hall 1
NULL a2 2003-05-23 19:30:00 NULL NULL
select * from t1 left join t2 on (venue_id = entity_id and match(name) against('aberdeen')) where dt = '2003-05-23 19:30:00';
venue_id venue_text dt name entity_id
1 a1 2003-05-23 19:30:00 aberdeen town hall 1
NULL a2 2003-05-23 19:30:00 NULL NULL
drop table t1,t2;
......@@ -80,6 +80,17 @@ delete from t1 where MATCH(a,b) AGAINST ("indexes");
select * from t1;
drop table t1;
#
# why to scan strings for trunc*
#
create table t1 (a varchar(200) not null, fulltext (a));
insert t1 values ("aaa10 bbb20"), ("aaa20 bbb15"), ("aaa30 bbb10");
select * from t1 where match a against ("+aaa* +bbb*" in boolean mode);
select * from t1 where match a against ("+aaa* +bbb1*" in boolean mode);
select * from t1 where match a against ("+aaa* +ccc*" in boolean mode);
select * from t1 where match a against ("+aaa10 +(bbb*)" in boolean mode);
drop table t1;
#
# Check bug reported by Matthias Urlichs
#
......
......@@ -41,5 +41,7 @@ create table t2 (name varchar(255) not null default '', entity_id int(11) not nu
insert into t2 (name, entity_id) values ('aberdeen town hall', 1), ('glasgow royal concert hall', 2), ('queen\'s hall, edinburgh', 3);
select * from t1 left join t2 on venue_id = entity_id where match(name) against('aberdeen' in boolean mode) and dt = '2003-05-23 19:30:00';
select * from t1 left join t2 on venue_id = entity_id where match(name) against('aberdeen') and dt = '2003-05-23 19:30:00';
select * from t1 left join t2 on (venue_id = entity_id and match(name) against('aberdeen' in boolean mode)) where dt = '2003-05-23 19:30:00';
select * from t1 left join t2 on (venue_id = entity_id and match(name) against('aberdeen')) where dt = '2003-05-23 19:30:00';
drop table t1,t2;
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment