ORACL

E
SQL TUNING
 Agend
a
• Introduction
• Single table access
• Joins
• Useful tip & techniques
• Test case/Q&A
INTRODUCTION
SQL
SELECT F. * , D.*
FROM
SB_DTM. FCT_ACCOUNT
F, SB_DTM.DIM_ACCOUNT D
WHERE DAYI D =
TRUNC( SYSDATE) - 1 AND
F.ACCOUNT_ID = D.ACCOUNT
AND D.EXP_DATE IS NOT NULL
ORDER BY 1 , 2 ;

• How to know the SQL is running

effective or not?

5
 Execution Plan
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |

| 0 | SELECT STATEMENT | | 2928K| 1279M| 208K ( 1 ) | 00:41:43 |

| 1 | MERGE JOIN | | 2928K| 1279M| 208K ( 1 ) | 00:41:43 |
|* 2 | TABLE ACCESS BY INDEX ROWID| DIM_ACCOUNT | 6609K| 1393M| 113K ( 1 ) | 00:22:42 |
| 3 | INDEX FULL SCAN | | 11M| | 481 ( 1 ) | 00:00:06 |
DIM_ACCOUNT_IDX_AC
|* 4 | SORT JOIN | | 1685K| 380M| 95109 ( 1 ) | 00:19:02 |
| 5 | PARTITION RANGE SINGLE | | 1685K| 380M| 8529 ( 6 ) | 00:01:43 |
|* 6 | TABLE ACCESS FULL | FCT_ACCOUNT | 1685K| 380M| 8529 ( 6 ) | 00:01:43 |

Predicate Information ( i d e n t i f i e d by operation i d ) :

2 - filter("D"."EXP_DATE" IS NOT NULL)

4 - access("F"."ACCOUNT_ID"="D"."ACCOUNT")
f i l t er ( " F" . " ACCOUNT_I D" =" D" . " ACCOU
NT" )
6 - filter("DAYID"=TRUNC(SYSDATE@!)-1)

6
 Execution Plan

• Generated by Oracle SQL engine

• Extremely complicated algorithms - CBO based
• The way Oracle process/retrieves data
• The more blocks accessed, the higher cost
(esp. OLTP)
Generate execution plan process

Query

Query
Transformations

Costing and Access

Paths

Execution Plan

8
 Query transformation

• SUBQUERY into JOIN

• Predicate pushing
• View merging
• …
• Behind the scene

9
 Costing and Access Paths

• Table/index/column statistics
• Number of rows, distinct values, number of nulls
• Min/Max
• Column histogram
• Index: Blevel; distinct keys, clustering factor
• …
• System statistics
• Multi-block read count
• Single-block/Multi-block read time
• CPU time
• …
• Optimizer parameters
• V$SES_OPTIMIZER_ENV

10
 Statistic – Where does it come from?

• Automatic Scheduler Job

• Start at 10PM every day – by default
• Gather for tables with change >10%
• Package: DBMS_STATS
• Customize for complex & large database

11
 View Execution Plan
• Oracle SQL Developer (F10)
• PL/SQL Developer
• Toad (Ctrl+E)
• Oracle supplied package DBMS_XPLAN (*)
• DISPLAY
• DISPLAY_CURSOR
• DISPLAY_AWR
• SQL Plus (SET AUTOTRACE ON|TRACEONLY) (*)
• SQL Monitoring (DBMS_SQLTUNE)
 Reading Execution Plan
SELECT * FROM SCOTT.EMP
WHERE DEPTNO IN DEPTNO FROM SCOTT.DEPT WHERE DNAME='RESEARCH')
(SELECT AND ENAME LIKE
'C%';
Plan hash value: 351108634

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |

| 0 | SELECT STATEMENT | | 1 | 109 | 4 (0)| 00:00:01 |

| 1 | NESTED LOOPS | | 1 | 109 | 4 (0)| 00:00:01 |
|* 2 | TABLE ACCESS FULL | EMP | 1 | 87 | 3 (0)| 00:00:01 |
|* 3 | TABLE ACCESS BY INDEX ROWID| DEPT | 1 | 22 | 1 (0)| 00:00:01 |
|* 4 | INDEX UNIQUE SCAN | PK_DEPT | 1 | | 0 (0)| 00:00:01 |

Predicate Information (identified by operation id):

2 - filter("ENAME" LIKE 'C%')

3 - filter("DNAME"='RESEARCH') Inside out – Top to bottom
4 - access("DEPTNO"="DEPTNO")
1st Table/idx without a child
Plan changes

14
SINGLE TABLE ACCESS
Heap table structure

16
 Table Access Methods

• Tablescan (Full scan)

• Read every block
• Up to high water mark
• Multi-block read
• Consider parallel if SQL is called infrequently
• Index scan
• Tree traversal
• Single block read
• Cache
Oracle B*Tree index

18
Index scan

• Range scan
• WHERE NAME = :1
• WHERE LASTUPDATE BETWEEN :1 AND :2
• WHERE NAME <> :1 ?
• Unique scan – Only for Unique index
• WHERE ID= :1
• Skip scan
• Index on (col1, col2); WHERE COL2 = :1
• Full scan (should be avoided)
• Output row is sorted
• Fast full scan
• Output is not sorted
Table scan vs Index scan

• Table scan
• “most” table data. 80/20 90/10
• How many rows really need to query?
• Report/aggregate SQL
• DSS/Data warehouse
• Index scan
• Selective query
• OLTP
• Histogram (data skew)
 Histogram – Data distribution

Hà Nội TP.Hồ Chí Minh Quảng Ninh Hà Giang … Total

100.000 150.000 30.000 500 500.000

SELECT * FROM TBL_ACCOUNT WHERE PROVINCE=‘HANOI’;

SELECT * FROM TBL_ACCOUNT WHERE PROVINCE=?;

con.prepareStatement()
 Selective index
• SELECT ID FROM MT_LIST WHERE PROCESS=‘N’;
• MT_LIST: 100.000.000 rows, PROCESS=‘N’: 100
• CREATE INDEX MT_LIST_PROCESS_N_I ON
MT_LIST(CASE WHEN PROCESS=‘N’ THEN ‘N’
ELSE NULL END);
• SELECT ID FROM MT_LIST WHERE (CASE WHEN
PROCESS=‘N’ THEN ‘N’ ELSE NULL END)=‘N’;
• Oracle does not index NULL value in B*Tree index
(except composite index)
• Index small
• Small overhead when update PROCESS=‘N’ to ‘Y’

22
 Forcing Tablescan or Index - Hint
• Hint
• Looks like comment: /* */
• Placement: directly after SELECT, INSERT, UPDATE,
DELETE
• Format: /*+ */
• Can be appeared in various places in a SQL
• Combine multiple keyword
• Example
• SELECT /*+ FULL(A) INDEX(B(COL1)) PARALLEL(8) */
COL1 FROM A, B WHERE …
• Full list: SQL Reference
• Only use Hint when Oracle failed to produce good
plan

23
 Forcing Tablescan or Index - Hint

• /*+ FULL(table alias) */

• /*+ INDEX(table alias(column)) */
• /*+ INDEX(table alias, index) */
• /*+ INDEX_FFS(table alias, index) */
Partition Pruning
CREATE TABLE quarterly _regional_s ales
(deptno number, item_no varchar2(20),
txn_date d a t e , txn_amount number, s t a t e v archar2(2))
PARTITION BY RANGE ( t x n _ d a t e )
SUBPARTITION BY LIST ( s t a t e )
(PARTITION q1_1999 VALUES LESS THAN (TO_DATE('1-APR-1999','DD-MON-YYYY'))
(SUBPARTITION q1_1999_northwest VALUES ( ' O R ' , ' WA ' ) ,
SUBPARTITION q1_1999_southwest VALUES ( ' A Z ' , ' U T ' , ' N M ' ) ,
SUBPARTITION q1_1999_northeast VALUES ( ' N Y ' , ' V M ' , ' N J ' ) ,
SUBPARTITION q1_1999_southeast VALUES ( ' F L ' , ' G A ' ) ,
SUBPARTITION q1_1999_northcentral VALUES ( ' S D ' , ' W I ' ) ,
SUBPARTITION q1_1999_southcentral VALUES ( ' O K ' , ' T X ' ) ) ,
PARTITION q2_1999 VALUES LESS THAN ( TO_DATE('1-JUL-1999','DD-MON-YYYY'))
(SUBPARTITION q2_1999_northwest VALUES ( ' O R ' , ' WA ' ) ,
SUBPARTITION q2_1999_southwest VALUES ( ' A Z ' , ' U T ' , ' N M ' ) ,
SUBPARTITION q2_1999_northeast VALUES ( ' N Y ' , ' V M ' , ' N J ' ) ,
SUBPARTITION q2_1999_southeast VALUES ( ' F L ' , ' G A ' ) ,
SUBPARTITION q2_1999_northcentral VALUES ( ' S D ' , ' W I ' ) ,
SUBPARTITION q2_1999_southcentral VALUES ( ' O K ' , ' T X ' ) ) ,
PARTITION q3_1999 VALUES LESS THAN (TO_DATE('1-OCT-1999','DD-MON-
YYYY'))
(SUBPARTITION q3_1999_northwest VALUES ( ' O R ' , ' WA ' ) ,
SUBPARTITION q3_1999_southwest VALUES ( ' A Z ' , ' U T ' , ' N M ' ) ,
SUBPARTITION q3_1999_northeast VALUES ( ' N Y ' , ' V M ' , ' N J ' ) ,
SUBPARTITION q3_1999_southeast VALUES ( ' F L ' , ' G A ' ) ,
SUBPARTITION q3_1999_northcentral VALUES ( ' S D ' , ' W I ' ) ,
SUBPARTITION q3_1999_southcentral VALUES ( ' O K ' , ' T X ' ) ) ,
PARTITION q4_1999 VALUES LESS THAN ( TO_DATE('1-JAN-2000','DD-MON-YYYY'))
(SUBPARTITION q4_1999_northwest VALUES ( ' O R ' , ' WA ' ) ,
SUBPARTITION q4_1999_southwest VALUES ( ' A Z ' , ' U T ' , ' N M ' ) ,
SUBPARTITION q4_1999_northeast VALUES ( ' N Y ' , ' V M ' , ' N J ' ) ,
SUBPARTITION q4_1999_southeast VALUES ( ' F L ' , ' G A ' ) ,
SUBPARTITION q4_1999_northcentral VALUES ( ' S D ' , ' W I ' ) ,
SUBPARTITION q4_1999_southcentral VALUES ( ' O K ' , ' T X ' ) ) ) ;

25
 Partition Pruning

• Partition Range/List/Hash All

• SELECT * FROM QUARTERLY_REGIONAL_SALES WHERE txn_amount>100000;

• Partition Range/List/Hash Iterator

• SELECT * FROM QUARTERLY_REGIONAL_SALES WHERE TXN_DATE Between :B1 and :B2;

• SELECT * FROM QUARTERLY_REGIONAL_SALES WHERE TXN_DATE >:B1 AND STATE='NY';
• Avoid TXN_DATE <= (>=) TO_DATE(’01/03/2018’,’DD/MM/YYYY’) – uncertain
partitions to be scanned
But TXN_DATE >= TO_DATE(’01/03/2014’,’DD/MM/YYYY’) and TXN_DATE <
TO_DATE(’01/04/2014’,’DD/MM/YYYY’)

• Partition Range/List/Hash Single

• SELECT * FROM QUARTERLY_REGIONAL_SALES WHERE TXN_DATE =:B1 AND STATE =
'NY';
 Partition Pruning not used

• Do not apply function/expression on column (same as

index)
• SELECT * FROM QUARTERLY_REGIONAL_SALES WHERE
TXN_DATE + 30 = :B1 > ?
• TXN_DATE = :B1 - 30
• SELECT * FROM QUARTERLY_REGIONAL_SALES
WHERE TO_CHAR(TXN_DATE,’MM/YYYY’) = :B1

27
 Summary

• When to use Full table scan/Index scan?

• Find FT with values > 100M
• Sum total loan of a company

29
 Workshop – Info

IP: 10.0.1.107
Port: 1521
DB: oracle
SID: orcl
User/pass: hr/ admin#123
 Workshop – Histogram – Data
distribution

Hà Nội TP.Hồ Chí Minh Quảng Ninh Hà Giang … Total

100.000 150.000 30.000 500 500.000

SELECT * FROM TBL_ACCOUNT WHERE PROVINCE=‘HANOI’;

SELECT * FROM TBL_ACCOUNT WHERE PROVINCE=?;

con.prepareStatement()
Workshop – Skip Scan

• Skip scan
• Index on (col1, col2); WHERE COL2 = :1
 Workshop – Index Function
• SELECT ID FROM MT_LIST WHERE PROCESS=‘N’;
• MT_LIST: 100.000.000 rows, PROCESS=‘N’: 100
• CREATE INDEX MT_LIST_PROCESS_N_I ON MT_LIST(CASE WHEN
PROCESS=‘N’ THEN ‘N’ ELSE NULL END);
• SELECT ID FROM MT_LIST WHERE (CASE WHEN PROCESS=‘N’ THEN
‘N’ ELSE NULL END)=‘N’;
• Oracle does not index NULL value in B*Tree index (except composite
index)
• Index small
• Small overhead when update PROCESS=‘N’ to ‘Y’

29
JOINS
 Join

• Retrieve data from more than 1

table
• Only join 2 tables at a time
• Join types
• Nested loop
• Hash join
• Merge join
• Merge join cartesian (*)
 Nested loop
SELECT A1.*, A2.* FROM SB_DTM.DIM_ACCOUNT A1, SB_DTM.FCT_ACCOUNT A2
WHERE A1.ACCOUNT_SK=A2.ACCOUNT_SK AND A1.OPENING_DATE BETWEEN :B1 AND
:B2
-
| Id | Operation | Name | Rows | Bytes | Cost
-
| 0 | SELECT STATEMENT | | 8852K| 3866M| 907K
|* 1 | FILTER | | | |
| 2 | TABLE ACCESS BY LOCAL INDEX ROWID| FCT_ACCOUNT | 327 | 77499 | 30
| 3 | NESTED LOOPS | | 8857K| 3866M| 907K
|* 4 | TABLE ACCESS FULL | DIM_ACCOUNT | 27086 | 5845K| 100K
| 5 | PARTITION RANGE ALL | | 327 | | 27
|* 6 | INDEX RANGE SCAN | IDX_FCT_ACCOUNT_SK_I | 327 | | 27
-

Predicate Information (identified by operation id):

1 - filter(TO_DATE(:B2)>=TO_DATE(:B1))
4 - filter("A1"."OPENING_DATE">=:B1
AND "A1"."OPENING_DATE"<=:B2)
6 - access("A1"."ACCOUNT_SK"="A2"."ACCOUNT_SK")
 Merge join
SELECT /*+ USE_MERGE(A1, A2) LEADING(A1) */ A1.*, A2.*
FROM SB_DTM.DIM_ACCOUNT A1, SB_DTM.FCT_ACCOUNT A2
WHERE A1.ACCOUNT_SK=A2.ACCOUNT_SK
AND A1.OPENING_DATE BETWEEN :B1 AND :B2

- - - - - - - - - - - - - - - - - - - - - - - - - -
-

| Id | Operation | Name | Rows | Bytes |TempSpc| Cost

| 0 | SELECT STATEMENT | | 8852K| 3866M| | 126M (1)|
(%CPU)|
|* 1 | FILTER | | | | | |
- - - - - - - - - - - - - - - - - - - - - - - - - -
|
- 2 | MERGE JOIN | | 8852K| 3866M| | 126M (1)|
| 3 | SORT JOIN | | 27086 | 5845K| 16M| 102K (2)|
|* 4 | TABLE ACCESS FULL | DIM_ACCOUNT | 27086 | 5845K| | 100K (2)|
|* 5 | SORT JOIN | | 2249M| 496G| 1634G| 126M (1)|
| 6 | PARTITION RANGE ALL| | TABLE 2249M| 496G| | 11M (5)|
| 7 | ACCESS FULL | FCT_ACCOUNT | 2249M| 496G| | 11M (5)|
- - - - - - - - - - - - - - - - - - - - - - - - - - -

Predicate Information (identified by operation id):

1 - filter(TO_DATE(:B2)>=TO_DATE(:B1))
4 - filter("A1"."OPENING_DATE">=:B1 AND "A1"."OPENING_DATE"<=:B2)
5 - access("A1"."ACCOUNT_SK"="A2"."ACCOUNT_SK")
filter("A1"."ACCOUNT_SK"="A2"."ACCOUNT_SK")
Merge join (continue)
 Hash join
SELECT /*+ USE_HASH(A1, A2) LEADING(A1) */ A1.*, A2.*
FROM SB_DTM.DIM_ACCOUNT A1, SB_DTM.FCT_ACCOUNT A2
WHERE A1.ACCOUNT_SK=A2.ACCOUNT_SK
AND A1.OPENING_DATE BETWEEN :B1 AND :B2
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost

| 0 | SELECT STATEMENT | | 8852K| 3866M| | 37M

|* 1 | FILTER | | | | |
|* 2 | HASH JOIN | | 8852K| 3866M| 6168K| 37M
|* 3 | TABLE ACCESS FULL | DIM_ACCOUNT | 27086 | 5845K| | 100K
| 4 | PARTITION RANGE ALL| | 2249M| 496G| | 11M
| 5 | TABLE ACCESS FULL | FCT_ACCOUNT | 2249M| 496G| | 11M

Predicate Information

1 - filter(TO_DATE(:B2)>=TO_DATE(:B1))
2 - access("A1"."ACCOUNT_SK“
="A2"."ACCOUNT_SK")
3 - filter("A1"."OPENING_DATE">=:B1
AND "A1"."OPENING_DATE"<=:B2)
Choosing the Join Method
• Nested-loop
• Joins involving a relatively small subset
of table data and where the join is supported
by an index
• No memory overhead
• Hash/Merge join
• More suitable when a large proportion of the tables
are being joined or if there is no suitable index.
• Hash join consume less memory
• Merge join is suitable when result needed to be
sorted
• Merge join Cartesian *
 Join order (Driving table)
SELECT A1.*, A2.* FROM
SB_DTM.DIM_ACCOUNT
A1,
SB_DTM.FCT_ACCOUNT A2
WHERE A1.ACCOUNT_SK=A2.ACCOUNT_SK AND
A1.OPENING_DATE BETWEEN :B1 AND :B2
AND A2.CO_CODE = :B3

Start with DIM_ACCOUNT or

FCT_ACCOUNT?
 Join order (Driving table)
- - - - - - - - - - - - - - - -- - - - - - - - -
| I d | Operation | Name | Rows |
- - - - - - - - - - - - - - - -- - - - - - - - -
| 0 | SELECT STATEMENT | | 56698 |
| * 1 | FILTER | | |
|* 2 | TABLE ACCESS BY LOCAL INDEX ROWID| FCT_ACCOUNT | 2 |
| 3 | NESTED LOOPS | | 8857K|
|* 4 | TABLE ACCESS FULL | DIM_ACCOUNT | 27086 |
| 5 | PARTITION RANGE ALL | | 327 |
|* 6 | INDEX RANGE SCAN | IDX_FCT_ACCOUNT_SK_I | 327 |
- - - - - - - - - - - - - - - -- - - - - - - - -

P redicate I n f o r m a t i o n ( i d e n t i f i e d by o p e r a t i o n i d ) :
- - - - - - - - - - - - - - - --
1 - filter(TO_DATE(:B2)>=TO_DATE(:B1))
2 - filter("A2"."CO_CODE"=:B3)
4 - filter("A1"."OPENING_DATE">=:B1 AND "A1"."OPENING_DATE"<=:B2)
6 - access("A1"."ACCOUNT_SK"="A2"."ACCOUNT_SK")

39
 Join order (Driving table)
- - - - - - - - - - - - - - - -- --
| I d | Oper at i on | Name | Rows |
- - - - - - - - - - - - - - - -- --
| 0 | SELECT STATEMENT | | 56698 |
| * 1 | FILTER | | |
|* 2 | HASH JOIN | | 56698 |
| 3 | PARTITION RANGE ALL| | 11M|
|* 4 | TABLE ACCESS FULL | FCT_ACCOUNT | 11M|
|* 5 | TABLE ACCESS FULL DIM_ACCOUNT | 27086 |
- - - - - - - - - - - - - - - -- --
|

P redicate I n f o r m a t i o n ( i d e n t i f i e d by o p e r a t i o n i d ) :
- - - - - - - - - - - - - - - --

1 - filter(TO_DATE(:B2)>=TO_DATE(:B1))
2 - access("A1"."ACCOUNT_SK"="A2"."ACCOUNT_SK")
4 - filter("A2"."CO_CODE"=:B3)
5 - filter("A1"."OPENING_DATE">=:B1 AND "A1"."OPENING_DATE"<=:B2)

40
 Join order (Driving table)
• The table that returns the fewest number of rows
should be the driving table
• Filter as many rows as possible early
• nPr(n,n) possible combinations
• THE MORE TABLE JOIN, THE MORE CHANCE
ORACLE WILL PRODUCE SUB-OPTIMAL PLAN

• KISS
Forcing Join types, Order
• /*+ USE_NL(tbl1 tbl2) */
• /*+ USE_HASH(tbl1 tbl2) */
• /*+ USE_MERGE(tbl1 tbl2) */
• /*+ LEADING(tbl1 tbl2) */
Help Oracle determine best plan

• Gather statistics (much like AV defs)

• Use SQL as simple as possible
• Einstein: “Everything should be as simple as it can be, but not
simpler”
• Use hint only when Oracle failed
USEFUL TIP & TECHNIQUES
 Use index effectively
• The more indexes – the slower INSERT / UPDATE / DELETE
• Create index that really speedup queries
• More indexes > more suboptimal plans
• Do not apply function/calculation on column which has index
• TRUNC(LAST_UPDATE)=TRUNC(SYSDATE)-1
• (LAST_UPDATE >= TRUNC(SYSDATE)-1) AND (LAST_UPDATE <
TRUNC(SYSDATE))
• SAL+1000>:X => SAL >: X-1000
• Avoid using NOT, <>
• STATUS NOT IN (1,2,3)=> STATUS in (5)
• COL1 <> VALUE1
• Do not use LIKE ‘%STRING’ with % at first (esp bind
variable)

46
Index with multiple column

• INDEX SKIP SCAN

• Terrible if leading column has high distinct values (eg.
ID)

47
 Bind variable

• Statement with string literals:

• SELECT NAME FROM USERS
WHERE ID = 100;
• SELECT NAME FROM USERS
WHERE ID = 110;
• Statement with bind
variables:
• SELECT NAME FROM USERS
WHERE ID = ?;
• Effects
• Parsing
• Oracle Memory fragment

48
 Bind variable

• Prevent SQL Injection

• Faster to execute
• May produce sub-optimal plan
• SELECT * FROM PROCESS_QUEUE WHERE STATUS=:1
• SELECT * FROM PROCESS_QUEUE WHERE STATUS=0

49
 SELECT * vs. SELECT COLUMN

• SELECT * FROM ACCT WHERE EFFDATE=:1;

• SELECT EFFDATE FROM ACCT WHERE
EFFDATE=:1;
• Reduce network traffic
• Oracle stores data by column order, the rest are skip
when not needed
• Coding best practice: explicitly, do not rely on implicit
• SQL is clear for later review
• INSERT INTO (COLUMN LIST) VALUES ()
• Not INSERT INTO VALUES()
• Add column => Insert Failed – Not enough values

50
Real life test case

• Take top SQL from a database & review

 NO RUNABLE SQL

BUT COST EFFECTIVE SQL

References

• Oracle® Database Performance Tuning Guide

• Troubleshooting Oracle Performance 2008 - Christian
Antognini (2nd 2014)
• Cost-Based Oracle Fundamentals 2006 - Jonathan
Lewis

• Group chat
 h.ma_so_thue mst,
Practice
_char(dk.ngay_dangky,'dd/mm/yyyy hh24:mi:ss') ngay,
01-CA06',
ase when csh.so_cmnd is not null and csh.ma_so_thue is null then 1 -- ca nhan
when ( csh.so_cmnd is null and csh.ma_so_thue is not null )then 5 -- to chuc
when(csh.so_cmnd is not null and csh.ma_so_thue is not null) then 5
end) cn
om BCA_prod.diem_dang_ki ddk , BCA_prod.oto_portal@oto p, BCA_prod.oto oto,
CA_prod.oto_dang_ky dk , BCA_prod.oto_chu_so_huu csh,
CA_prod.dia_danh_hanh_chinh@oto hc
here ddk.id = dk.DIEM_DK_ID
nd ddk.dia_danh_cap_cuoi = hc.id
nd ddk.dia_danh_cap_cuoi in (select id from bca_prod.dia_danh_hanh_chinh )
nd dk.chu_so_huu_id = csh.id
nd oto.so_khung not in (select p.so_khung from BCA_prod.oto_portal@oto
here p.ngay_tao between to_date('10/11/2022', 'dd/MM/yyyy') and to_date('10/12/2022',
d/MM/yyyy'))
nd oto.id = dk.oto_id
nd dk.trangthai_dangky in('01','03','04','06','07','10','12')
nd dk.ngay_dangky between to_date('10/11/2022', 'dd/MM/yyyy') and to_date('10/12/2022',
d/MM/yyyy')
nd (ddk.ten not like '%Test%' or ddk.ten not like '%test%' ) ORDER BY madd, caphc, loaidk
Q&A
Thank you!