mariadb/mysql-test/main/secondary_key_costs.test
Monty 4bf7c966b3 MDEV-34664: Add an option to fix InnoDB's doubling of secondary index cardinalities
(With trivial fixes by sergey@mariadb.com)
Added option fix_innodb_cardinality to optimizer_adjust_secondary_key_costs

Using fix_innodb_cardinality disables the 'divide by 2' of rec_per_key_int
in InnoDB that in effect doubles the Cardinality for secondary keys.
This has the biggest effect for indexes where a few rows has the same key
value. Using this may also cause table scans for very small tables (which
in some cases may be better than an index scan).

The user visible effect is that 'SHOW INDEX FROM table_name' will for
InnoDB show the true Cardinality (and not 2x the real value). It will
also allow the optimizer to chose a better index in some cases as the
division by 2 could have a bad effect for tables with 2-5 identical values
per key.

A few notes about using fix_innodb_cardinality:
- It has direct affect for SHOW INDEX FROM table_name. SHOW INDEX
  will also update the statistics in table share.
- The effect of fix_innodb_cardinality for query plans or EXPLAIN
  is only visible after first open of the table. This is why one must
  do a flush tables or use SHOW INDEX for the option to take effect.
- Using fix_innodb_cardinality can thus affect all user in their query
  plans if they are using the same tables.

Because of this, it is strongly recommended that one uses
optimizer_adjust_secondary_key_costs=fix_innodb_cardinality mainly
in configuration files to not cause issues for other users.
2024-07-29 16:40:53 +03:00

112 lines
3.6 KiB
Text

--source include/have_sequence.inc
--source include/not_embedded.inc
--source include/have_innodb.inc
# Testcase for MDEV-33306 takes ~6 minutes with valgrind:
--source include/not_valgrind.inc
#
# Show the costs for rowid filter
#
create table t1 (
pk int primary key auto_increment,
nm varchar(32),
fl1 tinyint default 0,
fl2 tinyint default 0,
index idx1(nm, fl1),
index idx2(fl2)
) engine=myisam;
create table name (
pk int primary key auto_increment,
nm bigint
) engine=myisam;
create table flag2 (
pk int primary key auto_increment,
fl2 tinyint
) engine=myisam;
insert into name(nm) select seq from seq_1_to_1000 order by rand(17);
insert into flag2(fl2) select seq mod 2 from seq_1_to_1000 order by rand(19);
insert into t1(nm,fl2)
select nm, fl2 from name, flag2 where name.pk = flag2.pk;
analyze table t1 persistent for all;
--disable_ps_protocol
set optimizer_trace="enabled=on";
set optimizer_switch='rowid_filter=on';
set statement optimizer_adjust_secondary_key_costs=0 for
explain select * from t1 where nm like '500%' AND fl2 = 0;
set @trace=(select trace from information_schema.optimizer_trace);
select json_detailed(json_extract(@trace, '$**.considered_access_paths'));
--echo
--echo The following trace should have a different rowid_filter_key cost
--echo
set statement optimizer_adjust_secondary_key_costs=2 for
explain select * from t1 where nm like '500%' AND fl2 = 0;
set @trace=(select trace from information_schema.optimizer_trace);
select json_detailed(json_extract(@trace, '$**.considered_access_paths'));
--enable_ps_protocol
drop table t1, name, flag2;
select @@optimizer_adjust_secondary_key_costs;
set @@optimizer_adjust_secondary_key_costs=7;
select @@optimizer_adjust_secondary_key_costs;
set @@optimizer_adjust_secondary_key_costs=default;
--echo #
--echo # MDEV-33306 Optimizer choosing incorrect index in 10.6, 10.5 but not in 10.4
--echo #
create table t1 (a int primary key, b int, c int, d int, key(b),key(c)) engine=innodb;
insert into t1 select seq, mod(seq,10), mod(seq,2), seq from seq_1_to_50000;
--replace_column 9 #
explain select b, sum(d) from t1 where c=0 group by b;
select b, sum(d) from t1 where c=0 group by b;
set @@optimizer_adjust_secondary_key_costs="disable_forced_index_in_group_by";
--replace_column 9 #
explain select b, sum(d) from t1 where c=0 group by b;
select b, sum(d) from t1 where c=0 group by b;
drop table t1;
--echo #
--echo # MDEV-34664: fix_innodb_cardinality
--echo #
set @save_userstat=@@global.userstat;
set @save_ispsp=@@global.innodb_stats_persistent_sample_pages;
set @@global.innodb_stats_persistent_sample_pages=20;
set @@global.userstat=on;
set use_stat_tables=PREFERABLY_FOR_QUERIES;
create or replace table t1 (a int primary key, b int, c int, d int, key(b,c,d)) engine=innodb;
insert into t1 select seq,seq/100,seq/60,seq/10 from seq_1_to_1000;
create or replace table t2 (a int);
insert into t2 values (1),(2),(3);
analyze table t1;
select count(distinct b),count(distinct b,c), count(distinct b,c,d) from t1;
show index from t1;
explain select * from t1,t2 where t1.b=t2.a;
set @@optimizer_adjust_secondary_key_costs=8;
explain select * from t1,t2 where t1.b=t2.a;
show index from t1;
# Flush tables or show index is needed to refresh the data in table share
flush tables;
explain select * from t1,t2 where t1.b=t2.a;
show index from t1;
# Check that the option does not affect other usage
connect (user2, localhost, root,,);
show index from t1;
connection default;
disconnect user2;
drop table t1,t2;
set global userstat=@save_userstat;
set global innodb_stats_persistent_sample_pages=@save_ispsp;
set @@optimizer_adjust_secondary_key_costs=default;