1
0
mirror of https://github.com/MariaDB/server.git synced 2025-07-29 05:21:33 +03:00

In best_access_path() change record_count to 1.0 if its less than 1.0.

In essence this means that we expect the user query to have at least
one matching row in the end.
This change will not affect the estimated rows for the plan, but will
ensure that the cost for adding a table is not neglected because of
record count being too low.

The reasons for this is that if we have table combination that
together has a very high selectivity then join record_count could
become very low (close to 0)

This would cause costs for all future tables to be so small that they
are irrelevant for the rest of the plan.
This has been shown to be the case in some performance benchmarks and
in a few mtr tests.

There is also still a problem in selectivity calculations as joining two
tables in different order causes a different estimation of total rows.
This can be seen in selectivity_innodb.test, test 'Q20' where joining
nation,supplier is expecting 1.111 rows_out while joining supplier,nation
is expecting 0.04 rows_out.

The reason for 0.04 is that the optimizer estimates 'supplier' to have
10 matching rows, and joining with nation (eq_ref) has 1 row. However
selectivity of n_name = 'UNITED STATES' makes the optimizer things
that there will be only 0.04 matching rows.

This patch avoids this "too low row count" to affect cost
caclulations.
This commit is contained in:
Monty
2022-12-02 17:18:50 +02:00
committed by Sergei Petrunia
parent 02f6ba571e
commit 8b7c0d69d2
10 changed files with 51 additions and 45 deletions

View File

@ -599,15 +599,15 @@ and n_name = 'UNITED STATES'
order by s_name
limit 10;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 PRIMARY supplier ALL PRIMARY,i_s_nationkey NULL NULL NULL 10 100.00 Using where; Using filesort
1 PRIMARY nation eq_ref PRIMARY PRIMARY 4 dbt3_s001.supplier.s_nationkey 1 4.00 Using where
1 PRIMARY partsupp ref PRIMARY,i_ps_partkey,i_ps_suppkey i_ps_suppkey 4 dbt3_s001.supplier.s_suppkey 70 100.00 Using where
1 PRIMARY part eq_ref PRIMARY PRIMARY 4 dbt3_s001.partsupp.ps_partkey 1 35.71 Using where; FirstMatch(nation)
1 PRIMARY nation ALL PRIMARY NULL NULL NULL 25 4.00 Using where; Using temporary; Using filesort
1 PRIMARY supplier ref PRIMARY,i_s_nationkey i_s_nationkey 5 dbt3_s001.nation.n_nationkey 1 100.00
1 PRIMARY part ALL PRIMARY NULL NULL NULL 200 7.03 Using where
1 PRIMARY partsupp eq_ref PRIMARY,i_ps_partkey,i_ps_suppkey PRIMARY 8 dbt3_s001.part.p_partkey,dbt3_s001.supplier.s_suppkey 1 7.11 Using where; FirstMatch(supplier)
4 DEPENDENT SUBQUERY lineitem ref i_l_shipdate,i_l_suppkey_partkey,i_l_partkey,i_l_suppkey i_l_suppkey_partkey 10 dbt3_s001.partsupp.ps_partkey,dbt3_s001.partsupp.ps_suppkey 8 14.40 Using where
Warnings:
Note 1276 Field or reference 'dbt3_s001.partsupp.ps_partkey' of SELECT #4 was resolved in SELECT #2
Note 1276 Field or reference 'dbt3_s001.partsupp.ps_suppkey' of SELECT #4 was resolved in SELECT #2
Note 1003 /* select#1 */ select sql_calc_found_rows `dbt3_s001`.`supplier`.`s_name` AS `s_name`,`dbt3_s001`.`supplier`.`s_address` AS `s_address` from `dbt3_s001`.`supplier` semi join (`dbt3_s001`.`part` join `dbt3_s001`.`partsupp`) join `dbt3_s001`.`nation` where `dbt3_s001`.`nation`.`n_nationkey` = `dbt3_s001`.`supplier`.`s_nationkey` and `dbt3_s001`.`nation`.`n_name` = 'UNITED STATES' and `dbt3_s001`.`part`.`p_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`partsupp`.`ps_suppkey` = `dbt3_s001`.`supplier`.`s_suppkey` and `dbt3_s001`.`partsupp`.`ps_availqty` > <expr_cache><`dbt3_s001`.`partsupp`.`ps_partkey`,`dbt3_s001`.`partsupp`.`ps_suppkey`>((/* select#4 */ select 0.5 * sum(`dbt3_s001`.`lineitem`.`l_quantity`) from `dbt3_s001`.`lineitem` where `dbt3_s001`.`lineitem`.`l_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`lineitem`.`l_suppkey` = `dbt3_s001`.`partsupp`.`ps_suppkey` and `dbt3_s001`.`lineitem`.`l_shipDATE` >= <cache>(cast('1993-01-01' as date)) and `dbt3_s001`.`lineitem`.`l_shipDATE` < <cache>(cast('1993-01-01' as date) + interval '1' year))) and `dbt3_s001`.`part`.`p_name` like 'g%' order by `dbt3_s001`.`supplier`.`s_name` limit 10
Note 1003 /* select#1 */ select sql_calc_found_rows `dbt3_s001`.`supplier`.`s_name` AS `s_name`,`dbt3_s001`.`supplier`.`s_address` AS `s_address` from `dbt3_s001`.`supplier` semi join (`dbt3_s001`.`part` join `dbt3_s001`.`partsupp`) join `dbt3_s001`.`nation` where `dbt3_s001`.`supplier`.`s_nationkey` = `dbt3_s001`.`nation`.`n_nationkey` and `dbt3_s001`.`nation`.`n_name` = 'UNITED STATES' and `dbt3_s001`.`partsupp`.`ps_partkey` = `dbt3_s001`.`part`.`p_partkey` and `dbt3_s001`.`partsupp`.`ps_suppkey` = `dbt3_s001`.`supplier`.`s_suppkey` and `dbt3_s001`.`partsupp`.`ps_availqty` > <expr_cache><`dbt3_s001`.`partsupp`.`ps_partkey`,`dbt3_s001`.`partsupp`.`ps_suppkey`>((/* select#4 */ select 0.5 * sum(`dbt3_s001`.`lineitem`.`l_quantity`) from `dbt3_s001`.`lineitem` where `dbt3_s001`.`lineitem`.`l_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`lineitem`.`l_suppkey` = `dbt3_s001`.`partsupp`.`ps_suppkey` and `dbt3_s001`.`lineitem`.`l_shipDATE` >= <cache>(cast('1993-01-01' as date)) and `dbt3_s001`.`lineitem`.`l_shipDATE` < <cache>(cast('1993-01-01' as date) + interval '1' year))) and `dbt3_s001`.`part`.`p_name` like 'g%' order by `dbt3_s001`.`supplier`.`s_name` limit 10
select sql_calc_found_rows
s_name, s_address
from supplier, nation
@ -654,15 +654,15 @@ and n_name = 'UNITED STATES'
order by s_name
limit 10;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 PRIMARY supplier ALL PRIMARY,i_s_nationkey NULL NULL NULL 10 100.00 Using where; Using filesort
1 PRIMARY nation eq_ref PRIMARY PRIMARY 4 dbt3_s001.supplier.s_nationkey 1 4.00 Using where
1 PRIMARY partsupp ref PRIMARY,i_ps_partkey,i_ps_suppkey i_ps_suppkey 4 dbt3_s001.supplier.s_suppkey 70 100.00 Using where
1 PRIMARY part eq_ref PRIMARY PRIMARY 4 dbt3_s001.partsupp.ps_partkey 1 35.71 Using where; FirstMatch(nation)
1 PRIMARY nation ALL PRIMARY NULL NULL NULL 25 4.00 Using where; Using temporary; Using filesort
1 PRIMARY supplier ref PRIMARY,i_s_nationkey i_s_nationkey 5 dbt3_s001.nation.n_nationkey 1 100.00
1 PRIMARY part ALL PRIMARY NULL NULL NULL 200 7.81 Using where
1 PRIMARY partsupp eq_ref PRIMARY,i_ps_partkey,i_ps_suppkey PRIMARY 8 dbt3_s001.part.p_partkey,dbt3_s001.supplier.s_suppkey 1 6.40 Using where; FirstMatch(supplier)
4 DEPENDENT SUBQUERY lineitem ref i_l_shipdate,i_l_suppkey_partkey,i_l_partkey,i_l_suppkey i_l_suppkey_partkey 10 dbt3_s001.partsupp.ps_partkey,dbt3_s001.partsupp.ps_suppkey 8 14.40 Using where
Warnings:
Note 1276 Field or reference 'dbt3_s001.partsupp.ps_partkey' of SELECT #4 was resolved in SELECT #2
Note 1276 Field or reference 'dbt3_s001.partsupp.ps_suppkey' of SELECT #4 was resolved in SELECT #2
Note 1003 /* select#1 */ select sql_calc_found_rows `dbt3_s001`.`supplier`.`s_name` AS `s_name`,`dbt3_s001`.`supplier`.`s_address` AS `s_address` from `dbt3_s001`.`supplier` semi join (`dbt3_s001`.`part` join `dbt3_s001`.`partsupp`) join `dbt3_s001`.`nation` where `dbt3_s001`.`nation`.`n_nationkey` = `dbt3_s001`.`supplier`.`s_nationkey` and `dbt3_s001`.`nation`.`n_name` = 'UNITED STATES' and `dbt3_s001`.`part`.`p_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`partsupp`.`ps_suppkey` = `dbt3_s001`.`supplier`.`s_suppkey` and `dbt3_s001`.`partsupp`.`ps_availqty` > <expr_cache><`dbt3_s001`.`partsupp`.`ps_partkey`,`dbt3_s001`.`partsupp`.`ps_suppkey`>((/* select#4 */ select 0.5 * sum(`dbt3_s001`.`lineitem`.`l_quantity`) from `dbt3_s001`.`lineitem` where `dbt3_s001`.`lineitem`.`l_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`lineitem`.`l_suppkey` = `dbt3_s001`.`partsupp`.`ps_suppkey` and `dbt3_s001`.`lineitem`.`l_shipDATE` >= <cache>(cast('1993-01-01' as date)) and `dbt3_s001`.`lineitem`.`l_shipDATE` < <cache>(cast('1993-01-01' as date) + interval '1' year))) and `dbt3_s001`.`part`.`p_name` like 'g%' order by `dbt3_s001`.`supplier`.`s_name` limit 10
Note 1003 /* select#1 */ select sql_calc_found_rows `dbt3_s001`.`supplier`.`s_name` AS `s_name`,`dbt3_s001`.`supplier`.`s_address` AS `s_address` from `dbt3_s001`.`supplier` semi join (`dbt3_s001`.`part` join `dbt3_s001`.`partsupp`) join `dbt3_s001`.`nation` where `dbt3_s001`.`supplier`.`s_nationkey` = `dbt3_s001`.`nation`.`n_nationkey` and `dbt3_s001`.`nation`.`n_name` = 'UNITED STATES' and `dbt3_s001`.`partsupp`.`ps_partkey` = `dbt3_s001`.`part`.`p_partkey` and `dbt3_s001`.`partsupp`.`ps_suppkey` = `dbt3_s001`.`supplier`.`s_suppkey` and `dbt3_s001`.`partsupp`.`ps_availqty` > <expr_cache><`dbt3_s001`.`partsupp`.`ps_partkey`,`dbt3_s001`.`partsupp`.`ps_suppkey`>((/* select#4 */ select 0.5 * sum(`dbt3_s001`.`lineitem`.`l_quantity`) from `dbt3_s001`.`lineitem` where `dbt3_s001`.`lineitem`.`l_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`lineitem`.`l_suppkey` = `dbt3_s001`.`partsupp`.`ps_suppkey` and `dbt3_s001`.`lineitem`.`l_shipDATE` >= <cache>(cast('1993-01-01' as date)) and `dbt3_s001`.`lineitem`.`l_shipDATE` < <cache>(cast('1993-01-01' as date) + interval '1' year))) and `dbt3_s001`.`part`.`p_name` like 'g%' order by `dbt3_s001`.`supplier`.`s_name` limit 10
select sql_calc_found_rows
s_name, s_address
from supplier, nation
@ -709,15 +709,15 @@ and n_name = 'UNITED STATES'
order by s_name
limit 10;
id select_type table type possible_keys key key_len ref rows filtered Extra
1 PRIMARY supplier ALL PRIMARY,i_s_nationkey NULL NULL NULL 10 100.00 Using where; Using filesort
1 PRIMARY nation eq_ref PRIMARY PRIMARY 4 dbt3_s001.supplier.s_nationkey 1 4.00 Using where
1 PRIMARY partsupp ref PRIMARY,i_ps_partkey,i_ps_suppkey i_ps_suppkey 4 dbt3_s001.supplier.s_suppkey 70 100.00 Using where
1 PRIMARY part eq_ref PRIMARY PRIMARY 4 dbt3_s001.partsupp.ps_partkey 1 35.71 Using where; FirstMatch(nation)
1 PRIMARY nation ALL PRIMARY NULL NULL NULL 25 4.00 Using where; Using temporary; Using filesort
1 PRIMARY supplier ref PRIMARY,i_s_nationkey i_s_nationkey 5 dbt3_s001.nation.n_nationkey 1 100.00
1 PRIMARY part ALL PRIMARY NULL NULL NULL 200 7.81 Using where
1 PRIMARY partsupp eq_ref PRIMARY,i_ps_partkey,i_ps_suppkey PRIMARY 8 dbt3_s001.part.p_partkey,dbt3_s001.supplier.s_suppkey 1 6.40 Using where; FirstMatch(supplier)
4 DEPENDENT SUBQUERY lineitem ref i_l_shipdate,i_l_suppkey_partkey,i_l_partkey,i_l_suppkey i_l_suppkey_partkey 10 dbt3_s001.partsupp.ps_partkey,dbt3_s001.partsupp.ps_suppkey 8 14.40 Using where
Warnings:
Note 1276 Field or reference 'dbt3_s001.partsupp.ps_partkey' of SELECT #4 was resolved in SELECT #2
Note 1276 Field or reference 'dbt3_s001.partsupp.ps_suppkey' of SELECT #4 was resolved in SELECT #2
Note 1003 /* select#1 */ select sql_calc_found_rows `dbt3_s001`.`supplier`.`s_name` AS `s_name`,`dbt3_s001`.`supplier`.`s_address` AS `s_address` from `dbt3_s001`.`supplier` semi join (`dbt3_s001`.`part` join `dbt3_s001`.`partsupp`) join `dbt3_s001`.`nation` where `dbt3_s001`.`nation`.`n_nationkey` = `dbt3_s001`.`supplier`.`s_nationkey` and `dbt3_s001`.`nation`.`n_name` = 'UNITED STATES' and `dbt3_s001`.`part`.`p_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`partsupp`.`ps_suppkey` = `dbt3_s001`.`supplier`.`s_suppkey` and `dbt3_s001`.`partsupp`.`ps_availqty` > <expr_cache><`dbt3_s001`.`partsupp`.`ps_partkey`,`dbt3_s001`.`partsupp`.`ps_suppkey`>((/* select#4 */ select 0.5 * sum(`dbt3_s001`.`lineitem`.`l_quantity`) from `dbt3_s001`.`lineitem` where `dbt3_s001`.`lineitem`.`l_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`lineitem`.`l_suppkey` = `dbt3_s001`.`partsupp`.`ps_suppkey` and `dbt3_s001`.`lineitem`.`l_shipDATE` >= <cache>(cast('1993-01-01' as date)) and `dbt3_s001`.`lineitem`.`l_shipDATE` < <cache>(cast('1993-01-01' as date) + interval '1' year))) and `dbt3_s001`.`part`.`p_name` like 'g%' order by `dbt3_s001`.`supplier`.`s_name` limit 10
Note 1003 /* select#1 */ select sql_calc_found_rows `dbt3_s001`.`supplier`.`s_name` AS `s_name`,`dbt3_s001`.`supplier`.`s_address` AS `s_address` from `dbt3_s001`.`supplier` semi join (`dbt3_s001`.`part` join `dbt3_s001`.`partsupp`) join `dbt3_s001`.`nation` where `dbt3_s001`.`supplier`.`s_nationkey` = `dbt3_s001`.`nation`.`n_nationkey` and `dbt3_s001`.`nation`.`n_name` = 'UNITED STATES' and `dbt3_s001`.`partsupp`.`ps_partkey` = `dbt3_s001`.`part`.`p_partkey` and `dbt3_s001`.`partsupp`.`ps_suppkey` = `dbt3_s001`.`supplier`.`s_suppkey` and `dbt3_s001`.`partsupp`.`ps_availqty` > <expr_cache><`dbt3_s001`.`partsupp`.`ps_partkey`,`dbt3_s001`.`partsupp`.`ps_suppkey`>((/* select#4 */ select 0.5 * sum(`dbt3_s001`.`lineitem`.`l_quantity`) from `dbt3_s001`.`lineitem` where `dbt3_s001`.`lineitem`.`l_partkey` = `dbt3_s001`.`partsupp`.`ps_partkey` and `dbt3_s001`.`lineitem`.`l_suppkey` = `dbt3_s001`.`partsupp`.`ps_suppkey` and `dbt3_s001`.`lineitem`.`l_shipDATE` >= <cache>(cast('1993-01-01' as date)) and `dbt3_s001`.`lineitem`.`l_shipDATE` < <cache>(cast('1993-01-01' as date) + interval '1' year))) and `dbt3_s001`.`part`.`p_name` like 'g%' order by `dbt3_s001`.`supplier`.`s_name` limit 10
select sql_calc_found_rows
s_name, s_address
from supplier, nation