SlideShare a Scribd company logo

5 Cool Things About SQL

Connor McDonald, profile picture
Connor McDonald

The document discusses several topics related to SQL: 1) SQLNet compression - How ordering data in a query can significantly reduce the amount of data sent over the network by compressing repeated values. Ordering by additional columns further improves compression. 2) NULLs and indexes - There is a misconception that indexes cannot be used with queries involving NULL values, but indexes can support queries searching for NULL values. 3) Subquery caching - Repeated scalar subqueries are cached and evaluated only once to improve performance of queries containing subqueries.

Technology
1 of 96
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
1 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8
2 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Five things you probably didn’t know about SQL Thomas Kyte http://asktom.oracle.com/
3 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8
4 The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle’s products remains at the sole discretion of Oracle.
5 Who am I • Been with Oracle since 1993 • User of Oracle since 1987 • The “Tom” behind AskTom in Oracle Magazine www.oracle.com/oramag • Expert Oracle Database Architecture • Effective Oracle by Design • Expert One on One Oracle • Beginning Oracle
6 <Insert Picture Here> Five things you probably didn’t know about SQL • SQLNet Compression • NULLS and Indexes and Cardinality • You are being watched! • Scalar Subquery Caching • Calling statement level non-deterministic functions
7 SQLNet Compression
8 SQLNet Compression • How you retrieve the data matters • Not all result sets are the same – even if they have the same data
9 SQLNet Compression ops$tkyte%ORA11GR2> create table t 2 as 3 select * 4 from all_objects; Table created. ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
10 SQLNet Compression ops$tkyte%ORA11GR2> set arraysize 15 ops$tkyte%ORA11GR2> set autotrace traceonly statistics
11 SQLNet Compression ops$tkyte%ORA11GR2> select * from t; 72228 rows selected. Statistics -------------------------------------------------------- 5794 consistent gets 8015033 bytes sent via SQL*Net to client 53385 bytes received via SQL*Net from client 4817 SQL*Net roundtrips to/from client 72228 rows processed
12 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp; 72228 rows selected. Statistics -------------------------------------------------------- 1031 consistent gets 3427630 bytes sent via SQL*Net to client 53385 bytes received via SQL*Net from client 4817 SQL*Net roundtrips to/from client 72228 rows processed
13 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp, object_type, owner; 72228 rows selected. Statistics ---------------------------------------------------------- 1031 consistent gets 3280011 bytes sent via SQL*Net to client 53385 bytes received via SQL*Net from client 4817 SQL*Net roundtrips to/from client 72228 rows processed
14 SQLNet Compression ops$tkyte%ORA11GR2> set arraysize 100 ops$tkyte%ORA11GR2> set autotrace traceonly statistics
15 SQLNet Compression ops$tkyte%ORA11GR2> select * from t; 72228 rows selected. Statistics -------------------------------------------------------- 1842 consistent gets 7482943 bytes sent via SQL*Net to client 8362 bytes received via SQL*Net from client 724 SQL*Net roundtrips to/from client 72228 rows processed
16 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp; 72228 rows selected. Statistics ------------------------------------------------------- 1031 consistent gets 2907819 bytes sent via SQL*Net to client 8362 bytes received via SQL*Net from client 724 SQL*Net roundtrips to/from client 72228 rows processed
17 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp, object_type, owner; 72228 rows selected. Statistics ---------------------------------------------------------- 1031 consistent gets 2760200 bytes sent via SQL*Net to client 8362 bytes received via SQL*Net from client 724 SQL*Net roundtrips to/from client 72228 rows processed
18 SQLNet Compression No Order 15 Some Order 15 Very Ordered 15 No Order 100 Some Order 100 Very Ordered 100 Bytes Sent 8.01 m 3.42 m 3.28 m 7.48 m 2.90 m 2.76 m % of original 100% 43% 41% 93% 36% 34% Consistent Gets 5832 1033 1033 1741 1033 1033 ops$tkyte%ORA11GR2> select round(1033*8/1024,2) from dual; ROUND(1033*8/1024,2) -------------------- 8.07
19 SQLNet Compression No Order 1000 Some Order 1000 Very Ordered 1000 No Order 100 Some Order 100 Very Ordered 100 Bytes Sent 7.39 m 2.82 m 2.67 m 7.48 m 2.90 m 2.76 m % of original 92% 35% 33% 93% 36% 34% Consistent Gets 1105 1033 1033 1741 1033 1033 ops$tkyte%ORA11GR2> select round(1033*8/1024,2) from dual; ROUND(1033*8/1024,2) -------------------- 8.07
20 NULLS and Indexes and Cardinality
21 “Wrong cardinality = Wrong Plan”
22 NULLs and Cardinality ops$tkyte%ORA11GR2> create table t 2 pctfree 20 3 as 4 select a.*, 5 case when mod(rownum,100) <= 50 6 then last_ddl_time 7 end end_date 8 from all_objects a; Table created.
23 NULLs and Cardinality ops$tkyte%ORA11GR2> create index t_idx 2 on t(end_date); Index created.
24 NULLs and Cardinality ops$tkyte%ORA11GR2> select count(*) 2 from t 3 where end_date 4 between to_date('01-sep-2010', 'dd-mon-yyyy') 5 and to_date('30-sep-2010', 'dd-mon-yyyy'); COUNT(*) ---------- 36267
25 NULLs and Cardinality ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats(user, 'T'); 3 end; 4 / PL/SQL procedure successfully completed.
26 NULLs and Cardinality ops$tkyte%ORA11GR2> select count(*), 2 count(distinct end_date), 3 count(end_date), 4 min(end_date), 5 max(end_date) 6 from t; CNT CNTD CNT2 MIN MAX ---------- ---------- ---------- --------- --------- 72228 703 36850 01-OCT-02 30-SEP-11
27 NULLs and Cardinality ops$tkyte%ORA11GR2> set autotrace traceonly explain ops$tkyte%ORA11GR2> select * 2 from t 3 where end_date 4 between to_date( '01-sep-2010', 'dd-mon-yyyy' ) 5 and to_date( '30-sep-2010', 'dd-mon-yyyy' ); Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873
28 NULLs and Cardinality -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 36024 | 3588K| 339 (1)| 00:00:05 | |* 1 | TABLE ACCESS FULL| T | 36024 | 3588K| 339 (1)| 00:00:05 | -------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 1 - filter("END_DATE"<=TO_DATE(' 2010-09-30 00:00:00', 'syyyy-mm-dd hh24:mi:ss') AND "END_DATE">=TO_DATE(' 2010-09-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))
29 NULLs and Cardinality ops$tkyte%ORA11GR2> update t 2 set end_date = 3 to_date( '01-jan-9999','dd-mon-yyyy' ) 4 where end_date is null; 35378 rows updated. ops$tkyte%ORA11GR2> commit; Commit complete.
30 NULLs and Cardinality ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats(user, 'T'); 3 end; 4 / PL/SQL procedure successfully completed.
31 NULLs and Cardinality ops$tkyte%ORA11GR2> select * 2 from t 3 where end_date 4 between to_date('01-sep-2010', 'dd-mon-yyyy') 5 and to_date('30-sep-2010', 'dd-mon-yyyy'); Execution Plan ------------------------------------------------- Plan hash value: 470836197
32 NULLs and Cardinality --------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| --------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 175 | 18375 | 10 (0)| | 1 | TABLE ACCESS BY INDEX ROWID| T | 175 | 18375 | 10 (0)| |* 2 | INDEX RANGE SCAN | T_IDX | 175 | | 2 (0)| --------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 1 - filter("END_DATE"<=TO_DATE(' 2010-09-30 00:00:00', 'syyyy-mm-dd hh24:mi:ss') AND "END_DATE">=TO_DATE(' 2010-09-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))
33 “Wrong cardinality = Wrong Plan”
34 <Insert Picture Here> Nulls and Indexes • There is a pervasive myth that indexes and NULLs are like matter and anti-matter • There is the thought that “where column is null” cannot use an index • There is a thought that NULLs are not indexed • None of that is true…
35 NULLs and Indexes ops$tkyte%ORA11GR2> create table t 2 as 3 select a.*, 4 case when mod(rownum,100) > 1 5 then object_type 6 end otype 7 from all_objects a; Table created.
36 NULLs and Indexes ops$tkyte%ORA11GR2> select count(*) from t where otype is null; COUNT(*) ---------- 1445
37 NULLs and Indexes ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
38 NULLs and Indexes ops$tkyte%ORA11GR2> create index t_idx on t(otype,owner); Index created.
39 NULLs and Indexes ops$tkyte%ORA11GR2> set autotrace traceonly explain ops$tkyte%ORA11GR2> select * from t where otype is null; Execution Plan -------------------------------------------------------- Plan hash value: 470836197
40 NULLs and Indexes ------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | ------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1445 | 149K| 96 (0)| 00:00:02 | | 1 | TABLE ACCESS BY INDEX ROWID| T | 1445 | 149K| 96 (0)| 00:00:02 | |* 2 | INDEX RANGE SCAN | T_IDX | 1445 | | 7 (0)| 00:00:01 | ------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - access("OTYPE" IS NULL)
41 NULLs and Indexes ops$tkyte%ORA11GR2> drop index t_idx; Index dropped. ops$tkyte%ORA11GR2> create index t_idx on t(otype,0); Index created.
42 NULLs and Indexes ------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | ------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1445 | 149K| 96 (0)| 00:00:02 | | 1 | TABLE ACCESS BY INDEX ROWID| T | 1445 | 149K| 96 (0)| 00:00:02 | |* 2 | INDEX RANGE SCAN | T_IDX | 1445 | | 7 (0)| 00:00:01 | ------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - access("OTYPE" IS NULL)
43 <Insert Picture Here> Nulls and Indexes • What is true is that entirely NULL key entries are not made in B*Tree indexes • Therefore, an index on just OTYPE cannot be used to find NULLs • But – what about B*Tree cluster indexes and Bitmap indexes?
44 You are being WATCHED!
45 <Insert Picture Here> You are being WATCHED! • 9i and before – V$ tables • 10g – ASH and AWR are obvious • But there is more – We watch what you ask for and change how statistics are gathered based on that.
46 You are being WATCHED! ops$tkyte%ORA11GR2> create table t 2 as 3 select a.*, 4 case when rownum < 500 5 then 1 6 else 99 7 end some_status 8 from all_objects a 9 / Table created.
47 You are being WATCHED! ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats(user,'T'); 3 end; 4 / PL/SQL procedure successfully completed.
48 You are being WATCHED! ops$tkyte%ORA11GR2> select histogram 2 from user_tab_cols 3 where table_name = 'T' 4 and column_name = 'SOME_STATUS'; HISTOGRAM --------------- NONE
49 You are being WATCHED! ops$tkyte%ORA11GR2> create index t_idx on t(some_status); Index created.
50 You are being WATCHED! ops$tkyte%ORA11GR2> set autotrace traceonly explain ops$tkyte%ORA11GR2> select * from t where some_status = 1; Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 36115 | 3526K| 300 (1)| 00:00:04 | |* 1 | TABLE ACCESS FULL| T | 36115 | 3526K| 300 (1)| 00:00:04 | --------------------------------------------------------------------------
51 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status = 99; Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 36115 | 3526K| 300 (1)| 00:00:04 | |* 1 | TABLE ACCESS FULL| T | 36115 | 3526K| 300 (1)| 00:00:04 | --------------------------------------------------------------------------
52 You are being WATCHED! ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
53 You are being WATCHED! ops$tkyte%ORA11GR2> select histogram 2 from user_tab_cols 3 where table_name = 'T' 4 and column_name = 'SOME_STATUS'; HISTOGRAM --------------- FREQUENCY
54 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status = 1; Execution Plan ---------------------------------------------------------- Plan hash value: 470836197 -------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time -------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 539 | 53900 | 10 (0)| 00:00 | 1 | TABLE ACCESS BY INDEX ROWID| T | 539 | 53900 | 10 (0)| 00:00 |* 2 | INDEX RANGE SCAN | T_IDX | 539 | | 2 (0)| 00:00 --------------------------------------------------------------------------------
55 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status = 99; Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 71683 | 7000K| 300 (1)| 00:00:04 | |* 1 | TABLE ACCESS FULL| T | 71683 | 7000K| 300 (1)| 00:00:04 | --------------------------------------------------------------------------
56 You are being WATCHED! ops$tkyte%ORA11GR2> select * 2 from 3 ( 4 select * 5 from sys.col_usage$ 6 where obj# = (select object_id 7 from dba_objects 8 where object_name = 'T' 9 and owner = 'OPS$TKYTE' ) 10 ) 11 unpivot (value for x in 12 ( EQUALITY_PREDS, EQUIJOIN_PREDS, NONEQUIJOIN_PREDS, 13 RANGE_PREDS, LIKE_PREDS, NULL_PREDS ) ) 14 /
57 You are being WATCHED! OBJ# INTCOL# TIMESTAMP X VALUE ---------- ---------- --------- ----------------- ---------- 98040 16 30-SEP-11 EQUALITY_PREDS 1 98040 16 30-SEP-11 EQUIJOIN_PREDS 0 98040 16 30-SEP-11 NONEQUIJOIN_PREDS 0 98040 16 30-SEP-11 RANGE_PREDS 0 98040 16 30-SEP-11 LIKE_PREDS 0 98040 16 30-SEP-11 NULL_PREDS 0 6 rows selected.
58 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status > 100; no rows selected ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
59 You are being WATCHED! OBJ# INTCOL# TIMESTAMP X VALUE ---------- ---------- --------- ----------------- ---------- 98040 16 30-SEP-11 EQUALITY_PREDS 2 98040 16 30-SEP-11 EQUIJOIN_PREDS 0 98040 16 30-SEP-11 NONEQUIJOIN_PREDS 0 98040 16 30-SEP-11 RANGE_PREDS 1 98040 16 30-SEP-11 LIKE_PREDS 0 98040 16 30-SEP-11 NULL_PREDS 0 6 rows selected.
60 <Insert Picture Here> You are being WATCHED! • You can ‘seed’ column stats pre-emptively • Adds more “watching” • Suggests possible extended statistics as well
61 You are being WATCHED! ops$tkyte%ORA11GR2> begin 2 dbms_stats.seed_col_usage( null, null, 10 ); 3 end; 4 / PL/SQL procedure successfully completed.
62 You are being WATCHED! ops$tkyte%ORA11GR2> select * 2 from t 3 where owner = 'SYS' 4 and object_type = 'DIMENSION'; no rows selected
63 You are being WATCHED! ops$tkyte%ORA11GR2> select dbms_stats.report_col_usage( user, 'T' ) 2 from dual; DBMS_STATS.REPORT_COL_USAGE(USER,'T') -------------------------------------------------------------------------------- LEGEND: ....... EQ : Used in single table EQuality predicate RANGE : Used in single table RANGE predicate LIKE : Used in single table LIKE predicate NULL : Used in single table is (not) NULL predicate EQ_JOIN : Used in EQuality JOIN predicate NONEQ_JOIN : Used in NON EQuality JOIN predicate FILTER : Used in single table FILTER predicate JOIN : Used in JOIN predicate GROUP_BY : Used in GROUP BY expression
64 You are being WATCHED! ops$tkyte%ORA11GR2> select dbms_stats.report_col_usage( user, 'T' ) 2 from dual; … COLUMN USAGE REPORT FOR OPS$TKYTE.T ................................... 1. OBJECT_TYPE : EQ 2. OWNER : EQ 3. SOME_STATUS : EQ RANGE 4. (OWNER, OBJECT_TYPE) : FILTER ###############################################################################
65 Scalar Subquery Caching
66 <Insert Picture Here> Scalar Subquery Caching • A scalar subquery is a query that returns zero or one rows and a single column • Can be used anywhere an expression can be used • Is executed conceptually once for each row it is processed against • For example:
67 Scalar Subquery Caching Select dname, (select count(*) from emp where emp.deptno = dept.deptno) from dept; Is a lot like….
68 Scalar Subquery Caching Begin for x in (select dname, deptno from dept) loop select count(*) into cnt from emp where deptno = X.DEPTNO; dbms_output.put_line ( x.dname || ‘ ‘ || x.cnt ); end loop; End;
69 <Insert Picture Here> Scalar Subquery Caching • Conceptually it is like that… • In reality there is caching going on • Up to 255 entries can be saved • Only for the duration of the query! Not across queries
70 Scalar Subquery Caching ops$tkyte%ORA11GR2> create table t 2 as 3 select * 4 from all_objects; Table created.
71 Scalar Subquery Caching ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user,'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
72 Scalar Subquery Caching ops$tkyte%ORA11GR2> create or replace function f( x in varchar2 ) 2 return number 3 as 4 begin 5 dbms_application_info.set_client_info 6 ( to_number(userenv('client_info'))+1 ); 7 8 return length(x); 9 end; 10 / Function created.
73 Scalar Subquery Caching ops$tkyte%ORA11GR2> variable startcpu number; ops$tkyte%ORA11GR2> begin 2 dbms_application_info.set_client_info(0); 3 :startcpu := dbms_utility.get_cpu_time; 4 end; 5 / PL/SQL procedure successfully completed. This is run before every subsequent query…
74 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 72233 115
75 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, (select f(owner) from dual) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 69 25
76 Scalar Subquery Caching ops$tkyte%ORA11GR2> create or replace function f( x in varchar2 ) 2 return number 3 DETERMINISTIC 4 as … 10 end; 11 / Function created.
77 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 8278 74
78 Scalar Subquery Caching ops$tkyte%ORA11GR2> create or replace function f( x in varchar2 ) 2 return number 3 RESULT_CACHE 4 as 5 begin … 10 end; 11 / Function created.
79 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 35 60
80 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 0 62
81 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, (select f(owner) from dual) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 0 22
82 Scalar Subquery Caching Select * from T where owner = g(‘scott’); How many times will g(‘scott’) be invoked? It depends of course…
83 Scalar Subquery Caching Select * from T where owner = (select g(‘scott’) from dual); Now How many times will g(‘scott’) be invoked? It won’t depend this time…
84 Statement Level non-Deterministic Functions
85 <Insert Picture Here> Statement level non-deterministic functions • What is a deterministic function? • What is a statement level deterministic function? • Why do we care?
86 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> create table t 2 as 3 select * 4 from all_users 5 where rownum <= 5; Table created.
87 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> create or replace function f 2 return number 3 as 4 pragma autonomous_transaction; 5 l_cnt number; 6 begin 7 select count(*) into l_cnt from t; 8 9 insert into t (username, user_id, created ) 10 values ( 'hello', 123, sysdate ); 11 commit; 12 13 return l_cnt; 14 end; 15 /
88 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> select count(*) over () cnt1, 2 (select count(*) from t) cnt2, 3 f() cnt3, 4 (select f() from dual) cnt4 5 from t;
89 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> select count(*) over () cnt1, 2 (select count(*) from t) cnt2, 3 f() cnt3, 4 (select f() from dual) cnt4 5 from t; CNT1 CNT2 CNT3 CNT4 ---------- ---------- ---------- ---------- 5 5 5 6 5 5 7 6 5 5 8 6 5 5 9 6 5 5 10 6
90 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> create or replace function f(p_scn in number) 2 return number 3 as 4 pragma autonomous_transaction; 5 l_cnt number; 6 begin 7 select count(*) into l_cnt from t 8 as of scn p_scn; 9 10 insert into t (username, user_id, created ) 11 values ( 'hello', 123, sysdate ); 12 commit; 13 14 return l_cnt; 15 end; 16 /
91 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> variable scn number ops$tkyte%ORA11GR2> exec :scn := dbms_flashback.get_system_change_number PL/SQL procedure successfully completed.
92 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> select count(*) over () cnt1, 2 (select count(*) from t) cnt2, 3 f(:scn) cnt3, 4 (select f(:scn) from dual) cnt4 5 from t; CNT1 CNT2 CNT3 CNT4 ---------- ---------- ---------- ---------- 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
93 <Insert Picture Here> Five things you probably didn’t know about SQL • SQLNet Compression • NULLS and Indexes and Cardinality • You are being watched! • Scalar Subquery Caching • Calling statement level non-deterministic functions
94 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Q&A
95 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8
96 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8

More Related Content

PPTX
Cardinality and the Optimizer
Connor McDonald
 
PDF
Oracle b tree index internals - rebuilding the thruth
Xavier Davias
 
PDF
Oracle Diagnostics : Explain Plans (Simple)
Hemant K Chitale
 
PPTX
More than 12 More things about Oracle Database 12c
Guatemala User Group
 
PPT
11 Things About11g
fcamachob
 
PPTX
Indexes From the Concept to Internals
Deiby Gómez
 
DOCX
Exadata - Smart Scan Testing
Monowar Mukul
 
PPTX
A few things about the Oracle optimizer - 2013
Connor McDonald
 
Cardinality and the Optimizer
Connor McDonald
 
Oracle b tree index internals - rebuilding the thruth
Xavier Davias
 
Oracle Diagnostics : Explain Plans (Simple)
Hemant K Chitale
 
More than 12 More things about Oracle Database 12c
Guatemala User Group
 
11 Things About11g
fcamachob
 
Indexes From the Concept to Internals
Deiby Gómez
 
Exadata - Smart Scan Testing
Monowar Mukul
 
A few things about the Oracle optimizer - 2013
Connor McDonald
 

What's hot (12)

PPTX
Oracle 122 partitioning_in_action_slide_share
Thomas Teske
 
PPTX
12c Mini Lesson - Better Defaults
Connor McDonald
 
PDF
UKOUG - 25 years of hints and tips
Connor McDonald
 
PDF
My SQL Idiosyncrasies That Bite OTN
Ronald Bradford
 
PDF
Flashback ITOUG
Connor McDonald
 
PPTX
Everything you always wanted to know about datetime types but didn’t have tim...
MartinHanssonOracle
 
PDF
Learn to use Stratio Crossdata
Álvaro Agea Herradón
 
PPTX
Using SQL to process hierarchies
Connor McDonald
 
PDF
Demystifying cost based optimization
Riyaj Shamsudeen
 
DOCX
Sql seuence and sub queries
pooja kumari
 
PDF
Sangam 19 - PLSQL still the coolest
Connor McDonald
 
PDF
Presentation indexing new features oracle 11g release 1 and release 2
xKinAnx
 
Oracle 122 partitioning_in_action_slide_share
Thomas Teske
 
12c Mini Lesson - Better Defaults
Connor McDonald
 
UKOUG - 25 years of hints and tips
Connor McDonald
 
My SQL Idiosyncrasies That Bite OTN
Ronald Bradford
 
Flashback ITOUG
Connor McDonald
 
Everything you always wanted to know about datetime types but didn’t have tim...
MartinHanssonOracle
 
Learn to use Stratio Crossdata
Álvaro Agea Herradón
 
Using SQL to process hierarchies
Connor McDonald
 
Demystifying cost based optimization
Riyaj Shamsudeen
 
Sql seuence and sub queries
pooja kumari
 
Sangam 19 - PLSQL still the coolest
Connor McDonald
 
Presentation indexing new features oracle 11g release 1 and release 2
xKinAnx
 
Ad

Similar to 5 Cool Things About SQL (20)

PPTX
Five more things about Oracle SQL and PLSQL
Connor McDonald
 
PPTX
Writing efficient sql
j9soto
 
PPT
11thingsabout11g 12659705398222 Phpapp01
Karam Abuataya
 
PDF
SQLチューニング総合診療Oracle CloudWorld出張所
Hiroshi Sekiguchi
 
PPTX
Understanding Query Optimization with ‘regular’ and ‘Exadata’ Oracle
Guatemala User Group
 
PDF
Oracle Database 12c Application Development
Saurabh K. Gupta
 
PPT
Metadata Matters
afa reg
 
PPTX
Wellington APAC Groundbreakers tour - Upgrading to the 12c Optimizer
Connor McDonald
 
PPTX
Oracle dbms_xplan.display_cursor format
Franck Pachot
 
PPTX
OpenWorld Sep14 12c for_developers
Connor McDonald
 
DOCX
Checking clustering factor to detect row migration
Heribertus Bramundito
 
PDF
New index features in MySQL 8
Erik Frøseth
 
PDF
On Seeing Double in V$SQL_Thomas_Kytepdf
cookie1969
 
PDF
Managing Statistics for Optimal Query Performance
Karen Morton
 
PPTX
Oracle 12c SPM
Anton Bushmelev
 
PDF
The Cost Based Optimiser in 11gR2
Sage Computing Services
 
PPTX
12c Mini Lesson - Data Redaction
Connor McDonald
 
PPTX
Date data type and Globalization in Oracle
Masoud Haji Hassan Pour
 
PDF
Chasing the optimizer
Mauro Pagano
 
PDF
Oracle Diagnostics : Joins - 1
Hemant K Chitale
 
Five more things about Oracle SQL and PLSQL
Connor McDonald
 
Writing efficient sql
j9soto
 
11thingsabout11g 12659705398222 Phpapp01
Karam Abuataya
 
SQLチューニング総合診療Oracle CloudWorld出張所
Hiroshi Sekiguchi
 
Understanding Query Optimization with ‘regular’ and ‘Exadata’ Oracle
Guatemala User Group
 
Oracle Database 12c Application Development
Saurabh K. Gupta
 
Metadata Matters
afa reg
 
Wellington APAC Groundbreakers tour - Upgrading to the 12c Optimizer
Connor McDonald
 
Oracle dbms_xplan.display_cursor format
Franck Pachot
 
OpenWorld Sep14 12c for_developers
Connor McDonald
 
Checking clustering factor to detect row migration
Heribertus Bramundito
 
New index features in MySQL 8
Erik Frøseth
 
On Seeing Double in V$SQL_Thomas_Kytepdf
cookie1969
 
Managing Statistics for Optimal Query Performance
Karen Morton
 
Oracle 12c SPM
Anton Bushmelev
 
The Cost Based Optimiser in 11gR2
Sage Computing Services
 
12c Mini Lesson - Data Redaction
Connor McDonald
 
Date data type and Globalization in Oracle
Masoud Haji Hassan Pour
 
Chasing the optimizer
Mauro Pagano
 
Oracle Diagnostics : Joins - 1
Hemant K Chitale
 
Ad

More from Connor McDonald (20)

PDF
Sangam 19 - Analytic SQL
Connor McDonald
 
PDF
Sangam 19 - Successful Applications on Autonomous
Connor McDonald
 
PDF
Sangam 2019 - The Latest Features
Connor McDonald
 
PDF
UKOUG 2019 - SQL features
Connor McDonald
 
PDF
APEX tour 2019 - successful development with autonomous
Connor McDonald
 
PDF
APAC Groundbreakers 2019 - Perth/Melbourne
Connor McDonald
 
PDF
OOW19 - Flashback, not just for DBAs
Connor McDonald
 
PDF
OOW19 - Read consistency
Connor McDonald
 
PDF
OOW19 - Slower and less secure applications
Connor McDonald
 
PDF
OOW19 - Killing database sessions
Connor McDonald
 
PDF
OOW19 - Ten Amazing SQL features
Connor McDonald
 
PDF
Latin America Tour 2019 - 18c and 19c featues
Connor McDonald
 
PDF
Latin America tour 2019 - Flashback
Connor McDonald
 
PDF
Latin America Tour 2019 - 10 great sql features
Connor McDonald
 
PDF
Latin America Tour 2019 - pattern matching
Connor McDonald
 
PDF
Latin America Tour 2019 - slow data and sql processing
Connor McDonald
 
PDF
ANSI vs Oracle language
Connor McDonald
 
PDF
OG Yatra - upgrading to the new 12c+ optimizer
Connor McDonald
 
PDF
OG Yatra - 25 years of hints and tips
Connor McDonald
 
PDF
OG Yatra - Flashback, not just for developers
Connor McDonald
 
Sangam 19 - Analytic SQL
Connor McDonald
 
Sangam 19 - Successful Applications on Autonomous
Connor McDonald
 
Sangam 2019 - The Latest Features
Connor McDonald
 
UKOUG 2019 - SQL features
Connor McDonald
 
APEX tour 2019 - successful development with autonomous
Connor McDonald
 
APAC Groundbreakers 2019 - Perth/Melbourne
Connor McDonald
 
OOW19 - Flashback, not just for DBAs
Connor McDonald
 
OOW19 - Read consistency
Connor McDonald
 
OOW19 - Slower and less secure applications
Connor McDonald
 
OOW19 - Killing database sessions
Connor McDonald
 
OOW19 - Ten Amazing SQL features
Connor McDonald
 
Latin America Tour 2019 - 18c and 19c featues
Connor McDonald
 
Latin America tour 2019 - Flashback
Connor McDonald
 
Latin America Tour 2019 - 10 great sql features
Connor McDonald
 
Latin America Tour 2019 - pattern matching
Connor McDonald
 
Latin America Tour 2019 - slow data and sql processing
Connor McDonald
 
ANSI vs Oracle language
Connor McDonald
 
OG Yatra - upgrading to the new 12c+ optimizer
Connor McDonald
 
OG Yatra - 25 years of hints and tips
Connor McDonald
 
OG Yatra - Flashback, not just for developers
Connor McDonald
 

Recently uploaded (20)

PPTX
Programs and apps: productivity, graphics, security and other tools
4mqw9zch22
 
PDF
Encapsulation_ Review paper, used for researhc scholars
gurumoop
 
PDF
Encapsulation theory and applications.pdf
gurumoop
 
PDF
Zenith AI: Advanced Artificial Intelligence
Biz Preneurs
 
PDF
Approach and Philosophy of On baking technology
30anthuong17
 
PDF
Accuracy of neural networks in brain wave diagnosis of schizophrenia
IAESIJAI
 
PDF
Heart disease approach using modified random forest and particle swarm optimi...
IAESIJAI
 
PPTX
TLE Review Electricity (Electricity).pptx
JayPolicarpio2
 
PDF
Microsoft Solutions Partner Drive Digital Transformation with D365.pdf
Microsoft Dynamics
 
PDF
project resource management chapter-09.pdf
ssuser00e6e21
 
PDF
DP Operators-handbook-extract for the Mautical Institute
LeonelMejiaLarios
 
PDF
Web App vs Mobile App What Should You Build First.pdf
KodekX
 
PDF
Building Integrated photovoltaic BIPV_UPV.pdf
SandeepSingh286037
 
PDF
Unlocking AI with Model Context Protocol (MCP)
Brian McKeiver
 
PPTX
Group 1 Presentation -Planning and Decision Making .pptx
MatthewLewis227954
 
PDF
Mushroom cultivation and it's methods.pdf
mailmeonrishi
 
PPTX
cloud_computing_Infrastucture_as_cloud_p
mainakbhattacharya20
 
PPTX
Chapter 5: Probability Theory and Statistics
RandyFin
 
PDF
From MVP to Full-Scale Product A Startup’s Software Journey.pdf
KodekX
 
PDF
ENT215_Completing-a-large-scale-migration-and-modernization-with-AWS.pdf
ssuser28425f
 
Programs and apps: productivity, graphics, security and other tools
4mqw9zch22
 
Encapsulation_ Review paper, used for researhc scholars
gurumoop
 
Encapsulation theory and applications.pdf
gurumoop
 
Zenith AI: Advanced Artificial Intelligence
Biz Preneurs
 
Approach and Philosophy of On baking technology
30anthuong17
 
Accuracy of neural networks in brain wave diagnosis of schizophrenia
IAESIJAI
 
Heart disease approach using modified random forest and particle swarm optimi...
IAESIJAI
 
TLE Review Electricity (Electricity).pptx
JayPolicarpio2
 
Microsoft Solutions Partner Drive Digital Transformation with D365.pdf
Microsoft Dynamics
 
project resource management chapter-09.pdf
ssuser00e6e21
 
DP Operators-handbook-extract for the Mautical Institute
LeonelMejiaLarios
 
Web App vs Mobile App What Should You Build First.pdf
KodekX
 
Building Integrated photovoltaic BIPV_UPV.pdf
SandeepSingh286037
 
Unlocking AI with Model Context Protocol (MCP)
Brian McKeiver
 
Group 1 Presentation -Planning and Decision Making .pptx
MatthewLewis227954
 
Mushroom cultivation and it's methods.pdf
mailmeonrishi
 
cloud_computing_Infrastucture_as_cloud_p
mainakbhattacharya20
 
Chapter 5: Probability Theory and Statistics
RandyFin
 
From MVP to Full-Scale Product A Startup’s Software Journey.pdf
KodekX
 
ENT215_Completing-a-large-scale-migration-and-modernization-with-AWS.pdf
ssuser28425f
 

5 Cool Things About SQL

  • 1. 1 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8
  • 2. 2 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Five things you probably didn’t know about SQL Thomas Kyte http://asktom.oracle.com/
  • 3. 3 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8
  • 4. 4 The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle’s products remains at the sole discretion of Oracle.
  • 5. 5 Who am I • Been with Oracle since 1993 • User of Oracle since 1987 • The “Tom” behind AskTom in Oracle Magazine www.oracle.com/oramag • Expert Oracle Database Architecture • Effective Oracle by Design • Expert One on One Oracle • Beginning Oracle
  • 6. 6 <Insert Picture Here> Five things you probably didn’t know about SQL • SQLNet Compression • NULLS and Indexes and Cardinality • You are being watched! • Scalar Subquery Caching • Calling statement level non-deterministic functions
  • 8. 8 SQLNet Compression • How you retrieve the data matters • Not all result sets are the same – even if they have the same data
  • 9. 9 SQLNet Compression ops$tkyte%ORA11GR2> create table t 2 as 3 select * 4 from all_objects; Table created. ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
  • 10. 10 SQLNet Compression ops$tkyte%ORA11GR2> set arraysize 15 ops$tkyte%ORA11GR2> set autotrace traceonly statistics
  • 11. 11 SQLNet Compression ops$tkyte%ORA11GR2> select * from t; 72228 rows selected. Statistics -------------------------------------------------------- 5794 consistent gets 8015033 bytes sent via SQL*Net to client 53385 bytes received via SQL*Net from client 4817 SQL*Net roundtrips to/from client 72228 rows processed
  • 12. 12 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp; 72228 rows selected. Statistics -------------------------------------------------------- 1031 consistent gets 3427630 bytes sent via SQL*Net to client 53385 bytes received via SQL*Net from client 4817 SQL*Net roundtrips to/from client 72228 rows processed
  • 13. 13 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp, object_type, owner; 72228 rows selected. Statistics ---------------------------------------------------------- 1031 consistent gets 3280011 bytes sent via SQL*Net to client 53385 bytes received via SQL*Net from client 4817 SQL*Net roundtrips to/from client 72228 rows processed
  • 14. 14 SQLNet Compression ops$tkyte%ORA11GR2> set arraysize 100 ops$tkyte%ORA11GR2> set autotrace traceonly statistics
  • 15. 15 SQLNet Compression ops$tkyte%ORA11GR2> select * from t; 72228 rows selected. Statistics -------------------------------------------------------- 1842 consistent gets 7482943 bytes sent via SQL*Net to client 8362 bytes received via SQL*Net from client 724 SQL*Net roundtrips to/from client 72228 rows processed
  • 16. 16 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp; 72228 rows selected. Statistics ------------------------------------------------------- 1031 consistent gets 2907819 bytes sent via SQL*Net to client 8362 bytes received via SQL*Net from client 724 SQL*Net roundtrips to/from client 72228 rows processed
  • 17. 17 SQLNet Compression ops$tkyte%ORA11GR2> select * from t order by timestamp, object_type, owner; 72228 rows selected. Statistics ---------------------------------------------------------- 1031 consistent gets 2760200 bytes sent via SQL*Net to client 8362 bytes received via SQL*Net from client 724 SQL*Net roundtrips to/from client 72228 rows processed
  • 18. 18 SQLNet Compression No Order 15 Some Order 15 Very Ordered 15 No Order 100 Some Order 100 Very Ordered 100 Bytes Sent 8.01 m 3.42 m 3.28 m 7.48 m 2.90 m 2.76 m % of original 100% 43% 41% 93% 36% 34% Consistent Gets 5832 1033 1033 1741 1033 1033 ops$tkyte%ORA11GR2> select round(1033*8/1024,2) from dual; ROUND(1033*8/1024,2) -------------------- 8.07
  • 19. 19 SQLNet Compression No Order 1000 Some Order 1000 Very Ordered 1000 No Order 100 Some Order 100 Very Ordered 100 Bytes Sent 7.39 m 2.82 m 2.67 m 7.48 m 2.90 m 2.76 m % of original 92% 35% 33% 93% 36% 34% Consistent Gets 1105 1033 1033 1741 1033 1033 ops$tkyte%ORA11GR2> select round(1033*8/1024,2) from dual; ROUND(1033*8/1024,2) -------------------- 8.07
  • 21. 21 “Wrong cardinality = Wrong Plan”
  • 22. 22 NULLs and Cardinality ops$tkyte%ORA11GR2> create table t 2 pctfree 20 3 as 4 select a.*, 5 case when mod(rownum,100) <= 50 6 then last_ddl_time 7 end end_date 8 from all_objects a; Table created.
  • 23. 23 NULLs and Cardinality ops$tkyte%ORA11GR2> create index t_idx 2 on t(end_date); Index created.
  • 24. 24 NULLs and Cardinality ops$tkyte%ORA11GR2> select count(*) 2 from t 3 where end_date 4 between to_date('01-sep-2010', 'dd-mon-yyyy') 5 and to_date('30-sep-2010', 'dd-mon-yyyy'); COUNT(*) ---------- 36267
  • 25. 25 NULLs and Cardinality ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats(user, 'T'); 3 end; 4 / PL/SQL procedure successfully completed.
  • 26. 26 NULLs and Cardinality ops$tkyte%ORA11GR2> select count(*), 2 count(distinct end_date), 3 count(end_date), 4 min(end_date), 5 max(end_date) 6 from t; CNT CNTD CNT2 MIN MAX ---------- ---------- ---------- --------- --------- 72228 703 36850 01-OCT-02 30-SEP-11
  • 27. 27 NULLs and Cardinality ops$tkyte%ORA11GR2> set autotrace traceonly explain ops$tkyte%ORA11GR2> select * 2 from t 3 where end_date 4 between to_date( '01-sep-2010', 'dd-mon-yyyy' ) 5 and to_date( '30-sep-2010', 'dd-mon-yyyy' ); Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873
  • 28. 28 NULLs and Cardinality -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 36024 | 3588K| 339 (1)| 00:00:05 | |* 1 | TABLE ACCESS FULL| T | 36024 | 3588K| 339 (1)| 00:00:05 | -------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 1 - filter("END_DATE"<=TO_DATE(' 2010-09-30 00:00:00', 'syyyy-mm-dd hh24:mi:ss') AND "END_DATE">=TO_DATE(' 2010-09-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))
  • 29. 29 NULLs and Cardinality ops$tkyte%ORA11GR2> update t 2 set end_date = 3 to_date( '01-jan-9999','dd-mon-yyyy' ) 4 where end_date is null; 35378 rows updated. ops$tkyte%ORA11GR2> commit; Commit complete.
  • 30. 30 NULLs and Cardinality ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats(user, 'T'); 3 end; 4 / PL/SQL procedure successfully completed.
  • 31. 31 NULLs and Cardinality ops$tkyte%ORA11GR2> select * 2 from t 3 where end_date 4 between to_date('01-sep-2010', 'dd-mon-yyyy') 5 and to_date('30-sep-2010', 'dd-mon-yyyy'); Execution Plan ------------------------------------------------- Plan hash value: 470836197
  • 32. 32 NULLs and Cardinality --------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| --------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 175 | 18375 | 10 (0)| | 1 | TABLE ACCESS BY INDEX ROWID| T | 175 | 18375 | 10 (0)| |* 2 | INDEX RANGE SCAN | T_IDX | 175 | | 2 (0)| --------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 1 - filter("END_DATE"<=TO_DATE(' 2010-09-30 00:00:00', 'syyyy-mm-dd hh24:mi:ss') AND "END_DATE">=TO_DATE(' 2010-09-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss'))
  • 33. 33 “Wrong cardinality = Wrong Plan”
  • 34. 34 <Insert Picture Here> Nulls and Indexes • There is a pervasive myth that indexes and NULLs are like matter and anti-matter • There is the thought that “where column is null” cannot use an index • There is a thought that NULLs are not indexed • None of that is true…
  • 35. 35 NULLs and Indexes ops$tkyte%ORA11GR2> create table t 2 as 3 select a.*, 4 case when mod(rownum,100) > 1 5 then object_type 6 end otype 7 from all_objects a; Table created.
  • 36. 36 NULLs and Indexes ops$tkyte%ORA11GR2> select count(*) from t where otype is null; COUNT(*) ---------- 1445
  • 37. 37 NULLs and Indexes ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
  • 38. 38 NULLs and Indexes ops$tkyte%ORA11GR2> create index t_idx on t(otype,owner); Index created.
  • 39. 39 NULLs and Indexes ops$tkyte%ORA11GR2> set autotrace traceonly explain ops$tkyte%ORA11GR2> select * from t where otype is null; Execution Plan -------------------------------------------------------- Plan hash value: 470836197
  • 40. 40 NULLs and Indexes ------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | ------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1445 | 149K| 96 (0)| 00:00:02 | | 1 | TABLE ACCESS BY INDEX ROWID| T | 1445 | 149K| 96 (0)| 00:00:02 | |* 2 | INDEX RANGE SCAN | T_IDX | 1445 | | 7 (0)| 00:00:01 | ------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - access("OTYPE" IS NULL)
  • 41. 41 NULLs and Indexes ops$tkyte%ORA11GR2> drop index t_idx; Index dropped. ops$tkyte%ORA11GR2> create index t_idx on t(otype,0); Index created.
  • 42. 42 NULLs and Indexes ------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | ------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1445 | 149K| 96 (0)| 00:00:02 | | 1 | TABLE ACCESS BY INDEX ROWID| T | 1445 | 149K| 96 (0)| 00:00:02 | |* 2 | INDEX RANGE SCAN | T_IDX | 1445 | | 7 (0)| 00:00:01 | ------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - access("OTYPE" IS NULL)
  • 43. 43 <Insert Picture Here> Nulls and Indexes • What is true is that entirely NULL key entries are not made in B*Tree indexes • Therefore, an index on just OTYPE cannot be used to find NULLs • But – what about B*Tree cluster indexes and Bitmap indexes?
  • 45. 45 <Insert Picture Here> You are being WATCHED! • 9i and before – V$ tables • 10g – ASH and AWR are obvious • But there is more – We watch what you ask for and change how statistics are gathered based on that.
  • 46. 46 You are being WATCHED! ops$tkyte%ORA11GR2> create table t 2 as 3 select a.*, 4 case when rownum < 500 5 then 1 6 else 99 7 end some_status 8 from all_objects a 9 / Table created.
  • 47. 47 You are being WATCHED! ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats(user,'T'); 3 end; 4 / PL/SQL procedure successfully completed.
  • 48. 48 You are being WATCHED! ops$tkyte%ORA11GR2> select histogram 2 from user_tab_cols 3 where table_name = 'T' 4 and column_name = 'SOME_STATUS'; HISTOGRAM --------------- NONE
  • 49. 49 You are being WATCHED! ops$tkyte%ORA11GR2> create index t_idx on t(some_status); Index created.
  • 50. 50 You are being WATCHED! ops$tkyte%ORA11GR2> set autotrace traceonly explain ops$tkyte%ORA11GR2> select * from t where some_status = 1; Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 36115 | 3526K| 300 (1)| 00:00:04 | |* 1 | TABLE ACCESS FULL| T | 36115 | 3526K| 300 (1)| 00:00:04 | --------------------------------------------------------------------------
  • 51. 51 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status = 99; Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 36115 | 3526K| 300 (1)| 00:00:04 | |* 1 | TABLE ACCESS FULL| T | 36115 | 3526K| 300 (1)| 00:00:04 | --------------------------------------------------------------------------
  • 52. 52 You are being WATCHED! ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
  • 53. 53 You are being WATCHED! ops$tkyte%ORA11GR2> select histogram 2 from user_tab_cols 3 where table_name = 'T' 4 and column_name = 'SOME_STATUS'; HISTOGRAM --------------- FREQUENCY
  • 54. 54 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status = 1; Execution Plan ---------------------------------------------------------- Plan hash value: 470836197 -------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time -------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 539 | 53900 | 10 (0)| 00:00 | 1 | TABLE ACCESS BY INDEX ROWID| T | 539 | 53900 | 10 (0)| 00:00 |* 2 | INDEX RANGE SCAN | T_IDX | 539 | | 2 (0)| 00:00 --------------------------------------------------------------------------------
  • 55. 55 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status = 99; Execution Plan ---------------------------------------------------------- Plan hash value: 1601196873 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 71683 | 7000K| 300 (1)| 00:00:04 | |* 1 | TABLE ACCESS FULL| T | 71683 | 7000K| 300 (1)| 00:00:04 | --------------------------------------------------------------------------
  • 56. 56 You are being WATCHED! ops$tkyte%ORA11GR2> select * 2 from 3 ( 4 select * 5 from sys.col_usage$ 6 where obj# = (select object_id 7 from dba_objects 8 where object_name = 'T' 9 and owner = 'OPS$TKYTE' ) 10 ) 11 unpivot (value for x in 12 ( EQUALITY_PREDS, EQUIJOIN_PREDS, NONEQUIJOIN_PREDS, 13 RANGE_PREDS, LIKE_PREDS, NULL_PREDS ) ) 14 /
  • 57. 57 You are being WATCHED! OBJ# INTCOL# TIMESTAMP X VALUE ---------- ---------- --------- ----------------- ---------- 98040 16 30-SEP-11 EQUALITY_PREDS 1 98040 16 30-SEP-11 EQUIJOIN_PREDS 0 98040 16 30-SEP-11 NONEQUIJOIN_PREDS 0 98040 16 30-SEP-11 RANGE_PREDS 0 98040 16 30-SEP-11 LIKE_PREDS 0 98040 16 30-SEP-11 NULL_PREDS 0 6 rows selected.
  • 58. 58 You are being WATCHED! ops$tkyte%ORA11GR2> select * from t where some_status > 100; no rows selected ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user, 'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
  • 59. 59 You are being WATCHED! OBJ# INTCOL# TIMESTAMP X VALUE ---------- ---------- --------- ----------------- ---------- 98040 16 30-SEP-11 EQUALITY_PREDS 2 98040 16 30-SEP-11 EQUIJOIN_PREDS 0 98040 16 30-SEP-11 NONEQUIJOIN_PREDS 0 98040 16 30-SEP-11 RANGE_PREDS 1 98040 16 30-SEP-11 LIKE_PREDS 0 98040 16 30-SEP-11 NULL_PREDS 0 6 rows selected.
  • 60. 60 <Insert Picture Here> You are being WATCHED! • You can ‘seed’ column stats pre-emptively • Adds more “watching” • Suggests possible extended statistics as well
  • 61. 61 You are being WATCHED! ops$tkyte%ORA11GR2> begin 2 dbms_stats.seed_col_usage( null, null, 10 ); 3 end; 4 / PL/SQL procedure successfully completed.
  • 62. 62 You are being WATCHED! ops$tkyte%ORA11GR2> select * 2 from t 3 where owner = 'SYS' 4 and object_type = 'DIMENSION'; no rows selected
  • 63. 63 You are being WATCHED! ops$tkyte%ORA11GR2> select dbms_stats.report_col_usage( user, 'T' ) 2 from dual; DBMS_STATS.REPORT_COL_USAGE(USER,'T') -------------------------------------------------------------------------------- LEGEND: ....... EQ : Used in single table EQuality predicate RANGE : Used in single table RANGE predicate LIKE : Used in single table LIKE predicate NULL : Used in single table is (not) NULL predicate EQ_JOIN : Used in EQuality JOIN predicate NONEQ_JOIN : Used in NON EQuality JOIN predicate FILTER : Used in single table FILTER predicate JOIN : Used in JOIN predicate GROUP_BY : Used in GROUP BY expression
  • 64. 64 You are being WATCHED! ops$tkyte%ORA11GR2> select dbms_stats.report_col_usage( user, 'T' ) 2 from dual; … COLUMN USAGE REPORT FOR OPS$TKYTE.T ................................... 1. OBJECT_TYPE : EQ 2. OWNER : EQ 3. SOME_STATUS : EQ RANGE 4. (OWNER, OBJECT_TYPE) : FILTER ###############################################################################
  • 66. 66 <Insert Picture Here> Scalar Subquery Caching • A scalar subquery is a query that returns zero or one rows and a single column • Can be used anywhere an expression can be used • Is executed conceptually once for each row it is processed against • For example:
  • 67. 67 Scalar Subquery Caching Select dname, (select count(*) from emp where emp.deptno = dept.deptno) from dept; Is a lot like….
  • 68. 68 Scalar Subquery Caching Begin for x in (select dname, deptno from dept) loop select count(*) into cnt from emp where deptno = X.DEPTNO; dbms_output.put_line ( x.dname || ‘ ‘ || x.cnt ); end loop; End;
  • 69. 69 <Insert Picture Here> Scalar Subquery Caching • Conceptually it is like that… • In reality there is caching going on • Up to 255 entries can be saved • Only for the duration of the query! Not across queries
  • 70. 70 Scalar Subquery Caching ops$tkyte%ORA11GR2> create table t 2 as 3 select * 4 from all_objects; Table created.
  • 71. 71 Scalar Subquery Caching ops$tkyte%ORA11GR2> begin 2 dbms_stats.gather_table_stats( user,'T' ); 3 end; 4 / PL/SQL procedure successfully completed.
  • 72. 72 Scalar Subquery Caching ops$tkyte%ORA11GR2> create or replace function f( x in varchar2 ) 2 return number 3 as 4 begin 5 dbms_application_info.set_client_info 6 ( to_number(userenv('client_info'))+1 ); 7 8 return length(x); 9 end; 10 / Function created.
  • 73. 73 Scalar Subquery Caching ops$tkyte%ORA11GR2> variable startcpu number; ops$tkyte%ORA11GR2> begin 2 dbms_application_info.set_client_info(0); 3 :startcpu := dbms_utility.get_cpu_time; 4 end; 5 / PL/SQL procedure successfully completed. This is run before every subsequent query…
  • 74. 74 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 72233 115
  • 75. 75 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, (select f(owner) from dual) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 69 25
  • 76. 76 Scalar Subquery Caching ops$tkyte%ORA11GR2> create or replace function f( x in varchar2 ) 2 return number 3 DETERMINISTIC 4 as … 10 end; 11 / Function created.
  • 77. 77 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 8278 74
  • 78. 78 Scalar Subquery Caching ops$tkyte%ORA11GR2> create or replace function f( x in varchar2 ) 2 return number 3 RESULT_CACHE 4 as 5 begin … 10 end; 11 / Function created.
  • 79. 79 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 35 60
  • 80. 80 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, f(owner) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 0 62
  • 81. 81 Scalar Subquery Caching ops$tkyte%ORA11GR2> select owner, (select f(owner) from dual) from t; 72233 rows selected. ops$tkyte%ORA11GR2> select userenv('client_info') ci, 2 dbms_utility.get_cpu_time-:startcpu cpu 3 from dual; CI CPU ---------- ---------- 0 22
  • 82. 82 Scalar Subquery Caching Select * from T where owner = g(‘scott’); How many times will g(‘scott’) be invoked? It depends of course…
  • 83. 83 Scalar Subquery Caching Select * from T where owner = (select g(‘scott’) from dual); Now How many times will g(‘scott’) be invoked? It won’t depend this time…
  • 85. 85 <Insert Picture Here> Statement level non-deterministic functions • What is a deterministic function? • What is a statement level deterministic function? • Why do we care?
  • 86. 86 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> create table t 2 as 3 select * 4 from all_users 5 where rownum <= 5; Table created.
  • 87. 87 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> create or replace function f 2 return number 3 as 4 pragma autonomous_transaction; 5 l_cnt number; 6 begin 7 select count(*) into l_cnt from t; 8 9 insert into t (username, user_id, created ) 10 values ( 'hello', 123, sysdate ); 11 commit; 12 13 return l_cnt; 14 end; 15 /
  • 88. 88 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> select count(*) over () cnt1, 2 (select count(*) from t) cnt2, 3 f() cnt3, 4 (select f() from dual) cnt4 5 from t;
  • 89. 89 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> select count(*) over () cnt1, 2 (select count(*) from t) cnt2, 3 f() cnt3, 4 (select f() from dual) cnt4 5 from t; CNT1 CNT2 CNT3 CNT4 ---------- ---------- ---------- ---------- 5 5 5 6 5 5 7 6 5 5 8 6 5 5 9 6 5 5 10 6
  • 90. 90 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> create or replace function f(p_scn in number) 2 return number 3 as 4 pragma autonomous_transaction; 5 l_cnt number; 6 begin 7 select count(*) into l_cnt from t 8 as of scn p_scn; 9 10 insert into t (username, user_id, created ) 11 values ( 'hello', 123, sysdate ); 12 commit; 13 14 return l_cnt; 15 end; 16 /
  • 91. 91 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> variable scn number ops$tkyte%ORA11GR2> exec :scn := dbms_flashback.get_system_change_number PL/SQL procedure successfully completed.
  • 92. 92 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Statement level non-deterministic functions ops$tkyte%ORA11GR2> select count(*) over () cnt1, 2 (select count(*) from t) cnt2, 3 f(:scn) cnt3, 4 (select f(:scn) from dual) cnt4 5 from t; CNT1 CNT2 CNT3 CNT4 ---------- ---------- ---------- ---------- 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
  • 93. 93 <Insert Picture Here> Five things you probably didn’t know about SQL • SQLNet Compression • NULLS and Indexes and Cardinality • You are being watched! • Scalar Subquery Caching • Calling statement level non-deterministic functions
  • 94. 94 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8 Q&A
  • 95. 95 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8
  • 96. 96 Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 8