2014-03-27 10:08:00 +01:00
#
# Engine-agnostic tests for statistics-based selectivity calculations.
# - selectivity tests that depend on the engine should go into
# t/selectivity.test. That test is run with myisam/innodb/xtradb.
# - this file is for tests that don't depend on the engine.
#
drop table if exists t0,t1,t2,t3;
select @@global.use_stat_tables;
@@global.use_stat_tables
COMPLEMENTARY
select @@session.use_stat_tables;
@@session.use_stat_tables
COMPLEMENTARY
set @save_use_stat_tables=@@use_stat_tables;
set use_stat_tables='preferably';
set @save_optimizer_use_condition_selectivity=@@optimizer_use_condition_selectivity;
set @save_histogram_size=@@histogram_size;
set @save_histogram_type=@@histogram_type;
#
# MDEV-5917: EITS: different order of predicates in IN (...) causes different estimates
#
create table t1(a int);
insert into t1 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t2 (col1 int);
# one value in 1..100 range
2014-06-11 19:08:06 +02:00
insert into t2 select ta.a + tb.a*10 from t1 ta, t1 tb;
2014-03-27 10:08:00 +01:00
# ten values in 100...200 range
2014-06-11 19:08:06 +02:00
insert into t2 select 100 + ta.a + tb.a*10 from t1 ta, t1 tb, t1 tc;
2014-03-27 10:08:00 +01:00
set histogram_type='SINGLE_PREC_HB';
set histogram_size=100;
set optimizer_use_condition_selectivity=4;
analyze table t2 persistent for all;
Table Op Msg_type Msg_text
2014-03-27 21:32:53 +01:00
test.t2 analyze status Engine-independent statistics collected
2014-03-27 10:08:00 +01:00
test.t2 analyze status OK
# The following two must have the same in 'Extra' column:
explain extended select * from t2 where col1 IN (20, 180);
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t2 ALL NULL NULL NULL NULL 1100 1.35 Using where
Warnings:
Note 1003 select `test`.`t2`.`col1` AS `col1` from `test`.`t2` where (`test`.`t2`.`col1` in (20,180))
explain extended select * from t2 where col1 IN (180, 20);
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t2 ALL NULL NULL NULL NULL 1100 1.35 Using where
Warnings:
Note 1003 select `test`.`t2`.`col1` AS `col1` from `test`.`t2` where (`test`.`t2`.`col1` in (180,20))
drop table t1, t2;
#
# MDEV-5926: EITS: Histogram estimates for column=least_possible_value are wrong
#
create table t0(a int);
insert into t0 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t1(a int);
2014-06-11 19:08:06 +02:00
insert into t1 select ta.a from t0 ta, t0 tb, t0 tc;
2014-03-27 10:08:00 +01:00
set histogram_size=20;
set histogram_type='single_prec_hb';
analyze table t1 persistent for all;
Table Op Msg_type Msg_text
2014-03-27 21:32:53 +01:00
test.t1 analyze status Engine-independent statistics collected
2014-03-27 10:08:00 +01:00
test.t1 analyze status OK
set use_stat_tables='preferably';
set optimizer_use_condition_selectivity=4;
# Should select about 10%:
explain extended select * from t1 where a=2;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 1000 9.52 Using where
Warnings:
Note 1003 select `test`.`t1`.`a` AS `a` from `test`.`t1` where (`test`.`t1`.`a` = 2)
# Should select about 10%:
explain extended select * from t1 where a=1;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 1000 9.52 Using where
Warnings:
Note 1003 select `test`.`t1`.`a` AS `a` from `test`.`t1` where (`test`.`t1`.`a` = 1)
# Must not have filtered=100%:
explain extended select * from t1 where a=0;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 1000 9.52 Using where
Warnings:
Note 1003 select `test`.`t1`.`a` AS `a` from `test`.`t1` where (`test`.`t1`.`a` = 0)
# Again, must not have filtered=100%:
explain extended select * from t1 where a=-1;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 1000 9.52 Using where
Warnings:
Note 1003 select `test`.`t1`.`a` AS `a` from `test`.`t1` where (`test`.`t1`.`a` = <cache>(-(1)))
drop table t0, t1;
#
# MDEV-4362: Selectivity estimates for IN (...) do not depend on whether the values are in range
#
create table t1 (col1 int);
set @a=-1;
2014-06-11 19:08:06 +02:00
create table t2 (a int) select (@a:=@a+1) as a from information_schema.session_variables ta limit 100;
insert into t1 select ta.a from t2 ta, t2 tb where ta.a < 100 and tb.a < 100;
2014-03-27 10:08:00 +01:00
select min(col1), max(col1), count(*) from t1;
min(col1) max(col1) count(*)
0 99 10000
set histogram_size=100;
analyze table t1 persistent for all;
Table Op Msg_type Msg_text
2014-03-27 21:32:53 +01:00
test.t1 analyze status Engine-independent statistics collected
2014-03-27 10:08:00 +01:00
test.t1 analyze status OK
explain extended select * from t1 where col1 in (1,2,3);
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 10000 3.37 Using where
Warnings:
Note 1003 select `test`.`t1`.`col1` AS `col1` from `test`.`t1` where (`test`.`t1`.`col1` in (1,2,3))
# Must not cause fp division by zero, or produce nonsense numbers:
explain extended select * from t1 where col1 in (-1,-2,-3);
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 10000 5.94 Using where
Warnings:
Note 1003 select `test`.`t1`.`col1` AS `col1` from `test`.`t1` where (`test`.`t1`.`col1` in (<cache>(-(1)),<cache>(-(2)),<cache>(-(3))))
explain extended select * from t1 where col1<=-1;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 10000 1.00 Using where
Warnings:
Note 1003 select `test`.`t1`.`col1` AS `col1` from `test`.`t1` where (`test`.`t1`.`col1` <= <cache>(-(1)))
drop table t1, t2;
#
2014-04-01 18:59:51 +02:00
# MDEV-5984: EITS: Incorrect filtered% value for single-table select with range access
#
create table t1(a int);
insert into t1 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t2 (a int, b int, col1 varchar(64), col2 varchar(64), key(a,b));
2014-06-11 19:08:06 +02:00
insert into t2 select ta.a+10*tb.a, tc.a+10*td.a, 'filler-data1', 'filler-data2' from t1 ta, t1 tb, t1 tc, t1 td;
2014-04-01 18:59:51 +02:00
set histogram_size=100;
set optimizer_use_condition_selectivity=4;
set use_stat_tables='preferably';
analyze table t2 persistent for all;
Table Op Msg_type Msg_text
test.t2 analyze status Engine-independent statistics collected
test.t2 analyze status Table is already up to date
# This must show filtered=100%:
explain extended select * from t2 where a in (1,2,3) and b in (1,2,3);
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t2 range a a 10 NULL 9 100.00 Using index condition
Warnings:
Note 1003 select `test`.`t2`.`a` AS `a`,`test`.`t2`.`b` AS `b`,`test`.`t2`.`col1` AS `col1`,`test`.`t2`.`col2` AS `col2` from `test`.`t2` where ((`test`.`t2`.`a` in (1,2,3)) and (`test`.`t2`.`b` in (1,2,3)))
drop table t2, t1;
#
2014-04-25 17:12:06 +02:00
# MDEV-5980: EITS: if condition is used for REF access, its selectivity is still in filtered%
#
create table t0(a int);
insert into t0 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t1(key1 int, col1 int, key(key1));
2014-06-11 19:08:06 +02:00
insert into t1 select ta.a, ta.a from t0 ta, t0 tb, t0 tc;
2014-04-25 17:12:06 +02:00
set histogram_size=100;
set use_stat_tables='preferably';
set optimizer_use_condition_selectivity=4;
analyze table t1 persistent for all;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status Table is already up to date
# 10% is ok
explain extended select * from t1 where col1=2;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 1000 9.90 Using where
Warnings:
Note 1003 select `test`.`t1`.`key1` AS `key1`,`test`.`t1`.`col1` AS `col1` from `test`.`t1` where (`test`.`t1`.`col1` = 2)
# Must show 100%, not 10%
explain extended select * from t1 where key1=2;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t1 ref key1 key1 5 const 98 100.00
Warnings:
Note 1003 select `test`.`t1`.`key1` AS `key1`,`test`.`t1`.`col1` AS `col1` from `test`.`t1` where (`test`.`t1`.`key1` = 2)
drop table t0, t1;
2014-04-25 17:04:54 +02:00
# MDEV-6003: EITS: ref access, keypart2=const vs keypart2=expr - inconsistent filtered% value
#
create table t0(a int);
insert into t0 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t1 (
kp1 int, kp2 int,
filler1 char(100),
filler2 char(100),
key(kp1, kp2)
);
insert into t1
select
2014-06-11 19:08:06 +02:00
ta.a,
tb.a,
2014-04-25 17:04:54 +02:00
'filler-data-1',
'filler-data-2'
2014-06-11 19:08:06 +02:00
from t0 ta, t0 tb, t0 tc;
2014-04-25 17:04:54 +02:00
set histogram_size=100;
set use_stat_tables='preferably';
set optimizer_use_condition_selectivity=4;
analyze table t1 persistent for all;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status Table is already up to date
# NOTE: 10*100%, 10*100% rows is ok
explain extended select * from t0, t1 where t1.kp1=t0.a and t1.kp2=t0.a+1;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t0 ALL NULL NULL NULL NULL 10 100.00 Using where
1 SIMPLE t1 ref kp1 kp1 10 test.t0.a,func 10 100.00 Using index condition
Warnings:
Note 1003 select `test`.`t0`.`a` AS `a`,`test`.`t1`.`kp1` AS `kp1`,`test`.`t1`.`kp2` AS `kp2`,`test`.`t1`.`filler1` AS `filler1`,`test`.`t1`.`filler2` AS `filler2` from `test`.`t0` join `test`.`t1` where ((`test`.`t1`.`kp1` = `test`.`t0`.`a`) and (`test`.`t1`.`kp2` = (`test`.`t0`.`a` + 1)))
# NOTE: t0: 10*100% is ok, t1: 10*9.90% is bad. t1 should have 10*100%.
explain extended select * from t0, t1 where t1.kp1=t0.a and t1.kp2=4;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 SIMPLE t0 ALL NULL NULL NULL NULL 10 100.00 Using where
1 SIMPLE t1 ref kp1 kp1 10 test.t0.a,const 10 100.00
Warnings:
Note 1003 select `test`.`t0`.`a` AS `a`,`test`.`t1`.`kp1` AS `kp1`,`test`.`t1`.`kp2` AS `kp2`,`test`.`t1`.`filler1` AS `filler1`,`test`.`t1`.`filler2` AS `filler2` from `test`.`t0` join `test`.`t1` where ((`test`.`t1`.`kp1` = `test`.`t0`.`a`) and (`test`.`t1`.`kp2` = 4))
drop table t0, t1;
#
2014-05-05 12:24:54 +02:00
# MDEV-6209: Assertion `join->best_read < double(1.79769313486231570815e+308L)'
# failed in bool greedy_search with optimizer_use_condition_selectivity>1
#
SET optimizer_use_condition_selectivity = 2;
CREATE TABLE t1 (a CHAR(6), b INT, PRIMARY KEY (a,b)) ENGINE=MyISAM;
INSERT INTO t1 VALUES ('foo',1),('bar',2);
SELECT * FROM t1 AS t1_1, t1 AS t1_2 WHERE NOT ( t1_1.a <> 'baz');
a b a b
DROP TABLE t1;
#
2014-06-07 19:00:26 +02:00
# MDEV-6308: Server crashes in table_multi_eq_cond_selectivity with ...
#
CREATE TABLE t1 (
id varchar(40) COLLATE utf8_bin,
dt datetime,
PRIMARY KEY (id)
);
INSERT INTO t1 VALUES ('foo','2011-04-12 05:18:08'),
('bar','2013-09-19 11:37:03');
CREATE TABLE t2 (
t1_id varchar(40) COLLATE utf8_bin,
f1 varchar(64),
f2 varchar(1024),
KEY (f1,f2(255))
);
INSERT INTO t2 VALUES ('foo','baz','qux'),('bar','baz','qux');
set optimizer_use_condition_selectivity=2;
explain
select * from t1,t2 where t1.id = t2.t1_id and t2.f2='qux' and t2.f1='baz';
id select_type table type possible_keys key key_len ref rows Extra
1 SIMPLE t2 ref f1 f1 325 const,const 1 Using index condition; Using where
1 SIMPLE t1 eq_ref PRIMARY PRIMARY 122 test.t2.t1_id 1
select * from t1,t2 where t1.id = t2.t1_id and t2.f2='qux' and t2.f1='baz';
id dt t1_id f1 f2
foo 2011-04-12 05:18:08 foo baz qux
bar 2013-09-19 11:37:03 bar baz qux
drop table t1,t2;
#
2014-06-10 12:25:16 +02:00
# MDEV-5985: EITS: selectivity estimates look illogical for join and non-key equalities
#
create table t0(a int);
insert into t0 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t1(a int);
2014-06-11 19:08:06 +02:00
insert into t1 select ta.a + tb.a* 10 + tc.a * 100 from t0 ta, t0 tb, t0 tc;
2014-06-10 12:25:16 +02:00
create table t2 as select * from t1;
set histogram_size=100;
set use_stat_tables='preferably';
set optimizer_use_condition_selectivity=4;
analyze table t1 persistent for all;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status OK
analyze table t2 persistent for all;
Table Op Msg_type Msg_text
test.t2 analyze status Engine-independent statistics collected
test.t2 analyze status OK
# Filtered will be 4.95, 9.90
2014-06-11 19:08:06 +02:00
explain extended select * from t1 ta, t2 tb where ta.a < 40 and tb.a < 100;
2014-06-10 12:25:16 +02:00
id select_type table type possible_keys key key_len ref rows filtered Extra
2014-06-11 19:08:06 +02:00
1 SIMPLE ta ALL NULL NULL NULL NULL 1000 4.95 Using where
1 SIMPLE tb ALL NULL NULL NULL NULL 1000 9.90 Using where; Using join buffer (flat, BNL join)
2014-06-10 12:25:16 +02:00
Warnings:
2014-06-11 19:08:06 +02:00
Note 1003 select `test`.`ta`.`a` AS `a`,`test`.`tb`.`a` AS `a` from `test`.`t1` `ta` join `test`.`t2` `tb` where ((`test`.`ta`.`a` < 40) and (`test`.`tb`.`a` < 100))
# Here, tb.filtered should not become 100%:
explain extended select * from t1 ta, t2 tb where ta.a < 40 and tb.a < 100 and tb.a=ta.a;
2014-06-10 12:25:16 +02:00
id select_type table type possible_keys key key_len ref rows filtered Extra
2014-06-11 19:08:06 +02:00
1 SIMPLE ta ALL NULL NULL NULL NULL 1000 4.95 Using where
1 SIMPLE tb ALL NULL NULL NULL NULL 1000 4.95 Using where; Using join buffer (flat, BNL join)
2014-06-10 12:25:16 +02:00
Warnings:
2014-06-11 19:08:06 +02:00
Note 1003 select `test`.`ta`.`a` AS `a`,`test`.`tb`.`a` AS `a` from `test`.`t1` `ta` join `test`.`t2` `tb` where ((`test`.`tb`.`a` = `test`.`ta`.`a`) and (`test`.`ta`.`a` < 40) and (`test`.`ta`.`a` < 100))
2014-06-10 12:25:16 +02:00
drop table t0,t1,t2;
#
2014-03-27 10:08:00 +01:00
# End of the test file
#
set use_stat_tables= @save_use_stat_tables;
set histogram_type=@save_histogram_type;
set histogram_size=@save_histogram_size;
set optimizer_use_condition_selectivity=@save_optimizer_use_condition_selectivity;
