The display modes of execution plans vary with the database system. Execution plans in OceanBase Database are displayed in the form of trees. The key to understanding an execution plan is to understand its operators. This topic describes details about common execution plans that contain operators for table access by normal index primary key (TABLE SCAN operator), table access by global index primary key (TABLE SCAN operator), or join algorithms (JOIN operator).
Table access by index primary key
In OceanBase Database V4.1, the logic of table access by normal index primary key and the logic of table access by global index primary key are encapsulated in the TABLE SCAN operator. When an execution plan is displayed, the is_index_back field is used to identify whether the operator needs to perform table access by index primary key, and the is_global_index field is used to identify whether the operator scans a global index.
The following sample code shows an execution plan that contains the TABLE SCAN operator:
obclient> CREATE TABLE t1(c1 INT PRIMARY KEY, c2 INT, c3 INT, c4 INT, INDEX k1(c2,c3));
Query OK, 0 rows affected
Q1:
obclient> EXPLAIN EXTENDED SELECT * FROM t1 WHERE c1 = 1\G
*************************** 1. row ***************************
Query Plan:
| ==================================
|ID|OPERATOR |NAME|EST. ROWS|COST|
----------------------------------
|0 |TABLE GET|t1 |1 |53 |
==================================
Outputs & filters:
-------------------------------------
0 - output([t1.c1(0x7f22fbe69340)], [t1.c2(0x7f22fbe695c0)], [t1.c3(0x7f22fbe69840)], [t1.c4(0x7f22fbe69ac0)]), filter(nil),
access([t1.c1(0x7f22fbe69340)], [t1.c2(0x7f22fbe695c0)], [t1.c3(0x7f22fbe69840)], [t1.c4(0x7f22fbe69ac0)]), partitions(p0),
is_index_back=false,
range_key([t1.c1(0x7f22fbe69340)]), range[1 ; 1],
range_cond([t1.c1(0x7f22fbe69340) = 1(0x7f22fbe68cf0)])
Q2:
obclient> EXPLAIN EXTENDED SELECT * FROM t1 WHERE c2 < 1 AND c3 < 1 AND c4 < 1\G
*************************** 1. row ***************************
Query Plan:
| ======================================
|ID|OPERATOR |NAME |EST. ROWS|COST |
--------------------------------------
|0 |TABLE SCAN|t1(k1)|100 |12422|
======================================
Outputs & filters:
-------------------------------------
0 - output([t1.c1(0x7f22fbd1e220)], [t1.c2(0x7f227decec40)], [t1.c3(0x7f227decf9b0)], [t1.c4(0x7f22fbd1dfa0)]), filter([t1.c3(0x7f227decf9b0) < 1(0x7f227decf360)], [t1.c4(0x7f22fbd1dfa0) < 1(0x7f22fbd1d950)]),
access([t1.c2(0x7f227decec40)], [t1.c3(0x7f227decf9b0)], [t1.c4(0x7f22fbd1dfa0)], [t1.c1(0x7f22fbd1e220)]), partitions(p0),
is_index_back=true, filter_before_indexback[true,false],
range_key([t1.c2(0x7f227decec40)], [t1.c3(0x7f227decf9b0)], [t1.c1(0x7f22fbd1e220)]),
range(NULL,MAX,MAX ; 1,MIN,MIN),
range_cond([t1.c2(0x7f227decec40) < 1(0x7f227dece5f0)])
Q3:
obclient> CREATE TABLE t2(c1 INT PRIMARY KEY, c2 INT, c3 INT) PARTITION BY
HASH(c1) PARTITIONS 4;
Query OK, 0 rows affected
obclient> CREATE INDEX i2 ON t2(c2) GLOBAL;
Query OK, 0 rows affected
obclient> EXPLAIN SELECT * FROM t2 WHERE c2 = 1;
+------------------------------------------------------------------+
| Query Plan |
+------------------------------------------------------------------+
| ======================================================== |
| |ID|OPERATOR |NAME |EST.ROWS|EST.TIME(us)| |
| -------------------------------------------------------- |
| |0 |DISTRIBUTED TABLE SCAN|t2(i2)|1 |28 | |
| ======================================================== |
| Outputs & filters: |
| ------------------------------------- |
| 0 - output([t2.c1], [t2.c2], [t2.c3]), filter(nil), rowset=256 |
| access([t2.c1], [t2.c2], [t2.c3]), partitions(p0) |
| is_index_back=true, is_global_index=true, |
| range_key([t2.c2], [t2.c1]), range(1,MIN ; 1,MAX), |
| range_cond([t2.c2 = 1]) |
+------------------------------------------------------------------+
In the preceding sample code, the outputs & filters section in the execution plan shows in detail the output information of the TABLE SCAN operator. The following table describes the parameters in the output information.
| Parameter | Description |
|---|---|
| operator | TABLE SCAN and TABLE GET are the two forms of the TABLE SCAN operator.
|
| name | The index selected for accessing data. The name of the selected index follows the table name. The absence of the index name means that the primary table is scanned. In OceanBase Database, the primary table has the same structure as the index, and the primary table is an index. |
| output | The output columns of the operator. |
| filter | The filter predicates of the operator. In this example, filter is set to nil because no filter condition is configured for the TABLE GET operator in Q1. |
| partitions | The partitions to be scanned in the query. |
| is_index_back | Indicates whether table access is required by the operator. In Q1, the primary table is selected. Therefore, table access is not required. In query Q2, where the indexed columns are c2, c3, and c1, table access by index primary key is required because the query needs to return column c4. |
| is_global_index | Indicates whether the operator performs a global index scan, including scan of the global index table and table access by global index primary key. For example, in query Q2, a normal index is used. Therefore, the value of is_global_index is false. In query Q3, the i2 index, which is a global index, of the t2 table is used. Therefore, the value of is_global_index is true. |
| filter_before_indexback | Corresponds to each filter and indicates whether the filter directly applies to the index or after the TABLE ACCESS BY INDEX PRIMARY KEY operation. For example, in query Q2, filter c3 < 1 can be directly applied to the index, which reduces the number of TABLE ACCESS BY INDEX PRIMARY KEY operations. However, filter c4 < 1 can be applied only after column c4 is fetched through table access. |
| range_key/range/range_cond |
|
Join order
JOIN operators join data in two tables based on specified conditions. Join operations are classified into three types: inner join, outer join, and semi/anti join.
OceanBase Database supports the following JOIN operators: NESTED LOOP JOIN (NLJ), MERGE JOIN (MJ), and HASH JOIN (HJ).
NLJ
In the following example, queries Q1 and Q2 use NLJ based on the hint. Operator 0 is an NLJ operator and has two subnodes: operators 1 and 2. Its execution logic is shown in the following list:
Run operator 1 to read a row.
Run operator 2 to read all rows.
Join the result sets of operators 1 and 2 and apply the filter condition to export the results.
Repeat step 1 until operator 1 stops iteration.
obclient> CREATE TABLE t1 (c1 INT, c2 INT);
Query OK, 0 rows affected
obclient> CREATE TABLE t2 (d1 INT, d2 INT, PRIMARY KEY (d1));
Query OK, 0 rows affected
Q1:
obclient> EXPLAIN SELECT /*+USE_NL(t1, t2)*/ t1.c2 + t2.d2 FROM t1, t2 WHERE c2 = d2\G
*************************** 1. row ***************************
Query Plan:
===========================================
|ID|OPERATOR |NAME|EST. ROWS|COST |
-------------------------------------------
|0 |NESTED-LOOP JOIN| |9782 |411238|
|1 | TABLE SCAN |T1 |999 |647 |
|2 | MATERIAL | |999 |1519 |
|3 | TABLE SCAN |T2 |999 |647 |
===========================================
Outputs & filters:
-------------------------------------
0 - output([T1.C2 + T2.D2]), filter(nil),
conds([T1.C2 = T2.D2]), nl_params_(nil)
1 - output([T1.C2]), filter(nil),
access([T1.C2]), partitions(p0)
2 - output([T2.D2]), filter(nil)
3 - output([T2.D2]), filter(nil),
access([T2.D2]), partitions(p0)
The MATERIAL operator materializes the data output of subsequent operators.
Q2:
obclient> EXPLAIN SELECT /*+USE_NL(t1, t2)*/ t1.c2 + t2.d2 FROM t1, t2 WHERE c1 = d1\G
*************************** 1. row ***************************
Query Plan:
| ==========================================
|ID|OPERATOR |NAME|EST. ROWS|COST |
------------------------------------------
|0 |NESTED-LOOP JOIN| |990 |37346|
|1 | TABLE SCAN |T1 |999 |669 |
|2 | TABLE GET |T2 |1 |36 |
==========================================
Outputs & filters:
-------------------------------------
0 - output([T1.C2 + T2.D2]), filter(nil),
conds(nil), nl_params_([T1.C1])
1 - output([T1.C1], [T1.C2]), filter(nil),
access([T1.C1], [T1.C2]), partitions(p0)
2 - output([T2.D2]), filter(nil),
access([T2.D2]), partitions(p0)
In the preceding sample code, the outputs & filters section in the execution plan shows in detail the output information of the NESTED LOOP JOIN operator. The following table describes the parameters in the output information.
| Parameter | Description |
|---|---|
| output | The output expression of the operator. |
| filter | The filter conditions of the operator. In this example, filter is set to nil because no filter condition is configured for the NLJ operator. |
| conds | The join conditions. For example, the join condition in example 1 is t1.c2 = t2.d2. |
| nl_params_ | The pushdown parameters generated based on the data of the table on the left of the NLJ operator. For example, the pushdown parameter in example 2 is t1.c1. For the iteration of each row of the table on the left, NLJ creates a parameter based on nl_params, and creates, based on this parameter and the original join condition c1= d1, a filter condition applicable to the table on the right: d1 = ?. The filter condition is pushed down to the table on the right, extracting the query range of the index. The query range is the range of data to be scanned. In query Q2, the No. 2 operator is a TABLE GET operator because of the pushdown condition d1=?. |
In the following example, no JOIN condition is specified in query Q3. The No. 0 operator is displayed as a NESTED-LOOP JOIN CARTESIAN, which is logically an NLJ operator with no JOIN conditions.
Q3:
obclient> EXPLAIN SELECT t1.c2 + t2.d2 FROM t1, t2\G
*************************** 1. row ***************************
Query Plan:
| =====================================================
|ID|OPERATOR |NAME|EST. ROWS|COST |
-----------------------------------------------------
|0 |NESTED-LOOP JOIN CARTESIAN| |998001 |747480|
|1 | TABLE SCAN |T1 |999 |647 |
|2 | MATERIAL | |999 |1519 |
|3 | TABLE SCAN |T2 |999 |647 |
=====================================================
Outputs & filters:
-------------------------------------
0 - output([T1.C2 + T2.D2]), filter(nil),
conds(nil), nl_params_(nil)
1 - output([T1.C2]), filter(nil),
access([T1.C2]), partitions(p0)
2 - output([T2.D2]), filter(nil)
3 - output([T2.D2]), filter(nil),
access([T2.D2]), partitions(p0)
MERGE JOIN (MJ)
In the following example, query Q4 uses MJ by adding the hint USE_MERGE. Operator 0 is an MJ operator and has two subnodes: operators 1 and 3. The operator merges data of the left and right subnodes, and therefore requires that the data of the two subnodes be ordered in relation to the JOIN column.
Take query Q4 as an example. The JOIN condition is t1.c2 = t2.d2, which means sorting data of t1 by c2, and data of t2 by d2. In query Q4, the output of operator 2 is unordered, and the output of operator 4 is sorted by d1. Both of them do not meet the requirement of MERGE JOIN. Therefore, operators 1 and 3 that are used for sorting are assigned to operator 0.
Q4:
obclient>EXPLAIN SELECT /*+USE_MERGE(t1, t2)*/ t1.c2 + t2.d2 FROM t1, t2
WHERE c2 = d2 AND c1 + d1 > 10\G
*************************** 1. row ***************************
Query Plan:
| ======================================
|ID|OPERATOR |NAME|EST. ROWS|COST |
--------------------------------------
|0 |MERGE JOIN | |3261 |14199|
|1 | SORT | |999 |4505 |
|2 | TABLE SCAN|T1 |999 |669 |
|3 | SORT | |999 |4483 |
|4 | TABLE SCAN|T2 |999 |647 |
======================================
Outputs & filters:
-------------------------------------
0 - output([T1.C2 + T2.D2]), filter(nil),
equal_conds([T1.C2 = T2.D2]), other_conds([T1.C1 + T2.D1 > 10])
1 - output([T1.C2], [T1.C1]), filter(nil), sort_keys([T1.C2, ASC])
2 - output([T1.C2], [T1.C1]), filter(nil),
access([T1.C2], [T1.C1]), partitions(p0)
3 - output([T2.D2], [T2.D1]), filter(nil), sort_keys([T2.D2, ASC])
4 - output([T2.D2], [T2.D1]), filter(nil),
access([T2.D2], [T2.D1]), partitions(p0)
In the following example, the JOIN condition in query Q5 is t1.c1 = t2.d1, which means sorting data of t1 by c1, and data of t2 by d1. In this execution plan, the primary table scan is performed for t2, and the results are sorted by d1. Therefore, it is unnecessary to assign an additional SORT operator. Ideally, proper indexes are selected for tables on the left and right of the join, and the data order specified by the indexes can meet the requirements of MJ. In this case, no SORT operator is needed.
Q5:
obclient>EXPLAIN SELECT /*+USE_MERGE(t1, t2)*/ t1.c2 + t2.d2 FROM t1, t2 WHERE c1 = d1\G
*************************** 1. row ***************************
Query Plan:
| =====================================
|ID|OPERATOR |NAME|EST. ROWS|COST|
-------------------------------------
|0 |MERGE JOIN | |990 |6096|
|1 | SORT | |999 |4505|
|2 | TABLE SCAN|T1 |999 |669 |
|3 | TABLE SCAN |T2 |999 |647 |
=====================================
Outputs & filters:
-------------------------------------
0 - output([T1.C2 + T2.D2]), filter(nil),
equal_conds([T1.C1 = T2.D1]), other_conds(nil)
1 - output([T1.C2], [T1.C1]), filter(nil), sort_keys([T1.C1, ASC])
2 - output([T1.C1], [T1.C2]), filter(nil),
access([T1.C1], [T1.C2]), partitions(p0)
3 - output([T2.D1], [T2.D2]), filter(nil),
access([T2.D1], [T2.D2]), partitions(p0)
In the preceding sample code, the outputs & filters section in the execution plan shows in detail the output information of the MERGE JOIN operator. The following table describes the parameters in the output information.
| Parameter | Description |
|---|---|
| output | The output expression of the operator. |
| filter | The filter conditions of the operator. In this example, filter is set to nil because no filter condition is configured for the MJ operator. |
| equal_conds | Equivalent join conditions for MJ. The result sets of the subnodes on the left and right must be ordered in relation to the JOIN column. |
| other_conds | Other JOIN conditions. For example, Q4 has an additional condition: t1.c1 + t2.d1 > 10. |
HASH JOIN (HJ)
In the following example, query Q6 uses HJ based on the USE_HASH hint. Operator 0 is an HJ operator and has two subnodes: operators 1 and 2. Execution logic of this operator:
Read data from the subnode on the left to generate a hash value based on the JOIN column, such as
t1.c1, and then create a hash table.Read data from the subnode on the right to generate a hash value based on the JOIN column, such as
t2.d1, and try to join with the data oft1in the corresponding hash table.
Q6:
obclient> EXPLAIN SELECT /*+USE_HASH(t1, t2)*/ t1.c2 + t2.d2 FROM t1, t2
WHERE c1 = d1 AND c2 + d2 > 1\G
*************************** 1. row ***************************
Query Plan:
| ====================================
|ID|OPERATOR |NAME|EST. ROWS|COST|
------------------------------------
|0 |HASH JOIN | |330 |4850|
|1 | TABLE SCAN|T1 |999 |669 |
|2 | TABLE SCAN|T2 |999 |647 |
====================================
Outputs & filters:
-------------------------------------
0 - output([T1.C2 + T2.D2]), filter(nil),
equal_conds([T1.C1 = T2.D1]), other_conds([T1.C2 + T2.D2 > 1])
1 - output([T1.C1], [T1.C2]), filter(nil),
access([T1.C1], [T1.C2]), partitions(p0)
2 - output([T2.D1], [T2.D2]), filter(nil),
access([T2.D1], [T2.D2]), partitions(p0)
In the preceding sample code, the outputs & filters section in the execution plan shows in detail the output information of the HASH JOIN operator. The following table describes the parameters in the output information.
| Parameter | Description |
|---|---|
| output | The output expression of the operator. |
| filter | The filter conditions of the operator. In this example, filter is set to nil because no filter condition is configured for the HJ operator. |
| equal_conds | The equivalent join. The JOIN columns on the left and right sides are used to calculate the hash value. |
| other_conds | Other JOIN conditions. For example, Q6 has an additional join condition: t1.c2 + t2.d2 > 1. |