Advanced PLSQL Optimizing for Better Performance

1
Advanced PL/SQL
Optimizing for Better Performance
111
Zohar Elkayam
CTO, Brillix
Zohar@Brillix.co.il
www.realdbamagic.com
Twitter: @realmgic

22
• Zohar Elkayam, COO/CTO at Brillix-DBAces
• Oracle ACE Associate
• DBA, team leader, Oracle University instructor, public
speaker, and a senior consultant for over 18 years
• Editor of ilDBA – Israel Database Community website
• Blogger – www.realdbamagic.com
Who am I?

3
We are committed to provide the highest quality of services
delivered by our dedicated team of highly trained and
experienced industry’s top data experts. We offer:
 Complete integrated end-to-end solutions based on best-of-
breed innovations in database, security and big data
technologies
 On-site professional customized trainings led by our team of
Oracle ACEs and Oracle Certified Professionals
 Comprehensive security solutions and services for leading
database platforms and business applications, leveraging a
world-class team of security experts
About Brillix

44
• Developing PL/SQL: Composite datatypes,
advanced cursors, dynamic SQL, tracing, and
more…
• Compiling PL/SQL: dependencies, optimization
levels, and DBMS_WARNING
• Tuning PL/SQL: GTT, Result cache and Memory
handling
• Oracle 11g and 12c new useful features
• SQLcl – New replacement tool for SQL*Plus
Agenda

5
The REAL Agenda
10:30-10:45‫הפסקה‬
12:30-13:30‫משתתפ‬ ‫לכל‬ ‫צהריים‬ ‫ארוחת‬‫הכנס‬ ‫י‬
15:00-15:15‫מתוקה‬ ‫הפסקה‬‫ומשובים‬
16:30‫הביתה‬ ‫הולכים‬

77
• Composite datatypes
• Advanced Cursors and Bulk operations
• Dynamic SQL and SQL Injection
• Autonomous Transactions
• 11g mentionable features
• 12c new development features
Developing PL/SQL

9
• Collections
– Nested table, varray
– Associative arrays/PLSQL tables
• Use collections methods
• Manipulate collections
• Distinguish between the different types of
collections and when to use them
Composite Datatypes

10
• A collection is a group of elements, all of the
same type.
• Collections work like arrays.
• Collections can store instances of an object type and,
conversely, can be attributes of an object type.
• Types of collections in PL/SQL:
– Associative arrays
• String-indexed collections
• INDEX BY pls_integer or BINARY_INTEGER
– Nested tables
– Varrays
Understanding Collections

11
Collection Types
Nested table Varray
Associative array
1 2 3 4 5 6 a f i o t w
Index by
PLS_INTEGER
Index by
VARCHAR2

13
Associative arrays:
– That are indexed by strings can improve performance
– Are pure memory structures that are much faster than
schema-level tables
– Provide significant additional flexibility
Using Associative Arrays
Associative arrays
1 2 3 4 5 6 a f i o t w
Index by
PLS_INTEGER
Index by
VARCHAR2

15
Associative array in PL/SQL (string-indexed):
Creating the Array
TYPE type_name IS TABLE OF element_type
INDEX BY VARCHAR2(size)
CREATE OR REPLACE PROCEDURE report_credit
(p_last_name customers.cust_last_name%TYPE,
p_credit_limit customers.credit_limit%TYPE)
IS
TYPE typ_name IS TABLE OF customers%ROWTYPE
INDEX BY customers.cust_email%TYPE;
v_by_cust_email typ_name;
i VARCHAR2(30);
PROCEDURE load_arrays IS
BEGIN
FOR rec IN (SELECT * FROM customers WHERE cust_email IS NOT NULL)
LOOP
-- Load up the array in single pass to database table.
v_by_cust_email (rec.cust_email) := rec;
END LOOP;
END;
...
Create the string-indexed
associative array type.
Create the string-indexed
associative array variable.
Populate the string-indexed
associative array variable.

16
Traversing the Array
...
BEGIN
load_arrays;
i:= v_by_cust_email.FIRST;
dbms_output.put_line ('For credit amount of: ' || p_credit_limit);
WHILE i IS NOT NULL LOOP
IF v_by_cust_email(i).cust_last_name = p_last_name
AND v_by_cust_email(i).credit_limit > p_credit_limit
THEN dbms_output.put_line ( 'Customer '||
v_by_cust_email(i).cust_last_name || ': ' ||
v_by_cust_email(i).cust_email || ' has credit limit of: ' ||
v_by_cust_email(i).credit_limit);
END IF;
i := v_by_cust_email.NEXT(i);
END LOOP;
END report_credit;
/
EXECUTE report_credit('Walken', 1200)
For credit amount of: 1200
Customer Walken: Emmet.Walken@LIMPKIN.COM has credit limit of: 3600
Customer Walken: Prem.Walken@BRANT.COM has credit limit of: 3700

18
Nested table characteristics:
– A table within a table
– Unbounded
– Available in both SQL and PL/SQL as well as the database
– Array-like access to individual rows
Using Nested Tables
Nested table:

19
pOrder nested table:
Nested tables are stored out-of-line in storage tables.
Nested Table Storage
Storage table:
ORDID SUPPLIER REQUESTER ORDERED ITEMS
500 50 5000 30-OCT-07
800 80 8000 31-OCT-07
NESTED_TABLE_ID PRODID PRICE
55 555
56 566
57 577
NESTED_TABLE_ID PRODID PRICE
88 888

20
To create a nested table in the database:
To create a nested table in PL/SQL:
Creating Nested Tables
CREATE [OR REPLACE] TYPE type_name AS TABLE OF
Element_datatype [NOT NULL];
TYPE type_name IS TABLE OF element_datatype [NOT NULL];

21
– First, define an object type:
– Second, declare a column of that collection type:
Declaring Collections: Nested Table
CREATE TYPE typ_item AS OBJECT --create object
(prodid NUMBER(5),
price NUMBER(7,2) )
/
CREATE TYPE typ_item_nst -- define nested table type
AS TABLE OF typ_item
/
CREATE TABLE pOrder ( -- create database table
ordid NUMBER(5),
supplier NUMBER(5),
requester NUMBER(4),
ordered DATE,
items typ_item_nst)
NESTED TABLE items STORE AS item_stor_tab
/
1
2
3

22
Using Nested Tables
Add data to the nested table:
INSERT INTO pOrder
VALUES (500, 50, 5000, sysdate, typ_item_nst(
typ_item(55, 555),
typ_item(56, 566),
typ_item(57, 577)));
INSERT INTO pOrder
VALUES (800, 80, 8000, sysdate,
typ_item_nst (typ_item (88, 888)));
500 50 5000 30-OCT-07
800 80 8000 31-OCT-07
PRODID PRICE
55 555
56 566
57 577
PRODID PRICE
88 888
1
2
1
2
pOrder nested table

23
• Querying the results:
• Querying the results with the TABLE function:
Using Nested Tables
SELECT * FROM porder;
ORDID SUPPLIER REQUESTER ORDERED
---------- ---------- ---------- ---------
ITEMS(PRODID, PRICE)
-----------------------------------------------------------------
500 50 5000 31-OCT-07
TYP_ITEM_NST(TYP_ITEM(55, 555), TYP_ITEM(56, 566), TYP_ITEM(57, 577))
800 80 8000 31-OCT-07
TYP_ITEM_NST(TYP_ITEM(88, 888))
SELECT p2.ordid, p1.*
FROM porder p2, TABLE(p2.items) p1;
ORDID PRODID PRICE
---------- ---------- ----------
800 88 888
500 57 577
500 55 555
500 56 566

24
Use the collection name and a subscript to reference a
collection element:
– Syntax:
– Example:
– To reference a field in a collection:
Referencing Collection Elements
collection_name(subscript)
v_with_discount(i)
p_new_items(i).prodid

25
Using Nested Tables in PL/SQL
CREATE OR REPLACE PROCEDURE add_order_items
(p_ordid NUMBER, p_new_items typ_item_nst)
IS
v_num_items NUMBER;
v_with_discount typ_item_nst;
BEGIN
v_num_items := p_new_items.COUNT;
v_with_discount := p_new_items;
IF v_num_items > 2 THEN
--ordering more than 2 items gives a 5% discount
FOR i IN 1..v_num_items LOOP
v_with_discount(i) :=
typ_item(p_new_items(i).prodid,
p_new_items(i).price*.95);
END LOOP;
END IF;
UPDATE pOrder
SET items = v_with_discount
WHERE ordid = p_ordid;
END;

26
Using Nested Tables in PL/SQL
-- caller pgm:
DECLARE
v_form_items typ_item_nst:= typ_item_nst();
BEGIN
-- let's say the form holds 4 items
v_form_items.EXTEND(4);
v_form_items(1) := typ_item(1804, 65);
add_order_items(800, v_form_items);
END;
PRODID PRICE
1804 65
3172 42
3337 800
2144 14
v_form_items variable
500 50 5000 30-OCT-07
800 80 8000 31-OCT-07
Resulting data in the pOrder nested table
PRODID PRICE
1804 61.75
3172 39.9
3337 760
2144 13.3
The prices are added
after discounts.

27
– To create a varray in the database:
– To create a varray in PL/SQL:
Understanding Varrays
Varray:
CREATE [OR REPLACE] TYPE type_name AS VARRAY
(max_elements) OF element_datatype [NOT NULL];
TYPE type_name IS VARRAY (max_elements) OF
element_datatype [NOT NULL];

28
CREATE TABLE department ( -- create database table
dept_id NUMBER(2),
name VARCHAR2(25),
budget NUMBER(12,2),
projects typ_ProjectList) -- declare varray as column
/
– First, define a collection type:
– Second, declare a collection of that type:
Declaring Collections: Varray
CREATE TYPE typ_Project AS OBJECT( --create object
project_no NUMBER(4),
title VARCHAR2(35),
cost NUMBER(12,2))
/
CREATE TYPE typ_ProjectList AS VARRAY (50) OF typ_Project
-- define VARRAY type
/
1
2
3

29
Add data to the table containing a varray column:
Using Varrays
INSERT INTO department
VALUES (10, 'Exec Admn', 30000000,
typ_ProjectList(
typ_Project(1001, 'Travel Monitor', 400000),
typ_Project(1002, 'Open World', 10000000)));
VALUES (20, 'IT', 5000000,
typ_ProjectList(
typ_Project(2001, 'DB11gR2', 900000)));
1
2
DEPT_ID NAME BUDGET PROJECTS
PROJECT_NO TITLE COSTS
10 Exec Admn 30000000 1001 Travel Monitor 400000
1002 Open World 10000000
20 IT 5000000 2001 DB11gR2 900000
1
2
DEPARTMENT table

30
– Querying the results:
– Querying the results with the TABLE function:
Using Varrays
SELECT * FROM department;
DEPT_ID NAME BUDGET
---------- ------------------------- ----------
PROJECTS(PROJECT_NO, TITLE, COST)
-----------------------------------------------------------------
10 Executive Administration 30000000
TYP_PROJECTLIST(TYP_PROJECT(1001, 'Travel Monitor', 400000),
TYP_PROJECT(1002, 'Open World', 10000000))
20 Information Technology 5000000
TYP_PROJECTLIST(TYP_PROJECT(2001, 'DB11gR2', 900000))
SELECT d2.dept_id, d2.name, d1.*
FROM department d2, TABLE(d2.projects) d1;
DEPT_ID NAME PROJECT_NO TITLE COST
------- ------------------------ ---------- -------------- --------
10 Executive Administration 1001 Travel Monitor 400000
10 Executive Administration 1002 Open World 10000000
20 Information Technology 2001 DB11gR2 900000

31
• You can declare collections as the formal parameters of procedures
and functions.
• You can specify a collection type in the RETURN clause of a
function specification.
• Collections follow the usual scoping and instantiation rules.
Working with Collections in PL/SQL
CREATE OR REPLACE PACKAGE manage_dept_proj
AS
PROCEDURE allocate_new_proj_list
(p_dept_id NUMBER, p_name VARCHAR2, p_budget NUMBER);
FUNCTION get_dept_project (p_dept_id NUMBER)
RETURN typ_projectlist;
PROCEDURE update_a_project
(p_deptno NUMBER, p_new_project typ_Project,
p_position NUMBER);
FUNCTION manipulate_project (p_dept_id NUMBER)
RETURN typ_projectlist;
FUNCTION check_costs (p_project_list typ_projectlist)
RETURN boolean;
END manage_dept_proj;

34
Three ways to initialize:
– Use a constructor.
– Fetch from the database.
– Assign another collection variable directly.
Initializing Collections
PROCEDURE allocate_new_proj_list
(p_dept_id NUMBER, p_name VARCHAR2, p_budget NUMBER)
IS
v_accounting_project typ_projectlist;
BEGIN
-- this example uses a constructor
v_accounting_project :=
typ_ProjectList
(typ_Project (1, 'Dsgn New Expense Rpt', 3250),
typ_Project (2, 'Outsource Payroll', 12350),
typ_Project (3, 'Audit Accounts Payable',1425));
VALUES(p_dept_id, p_name, p_budget, v_accounting_project);
END allocate_new_proj_list;

35
FUNCTION get_dept_project (p_dept_id NUMBER)
RETURN typ_projectlist
IS
BEGIN -- this example uses a fetch from the database
SELECT projects INTO v_accounting_project
FROM department WHERE dept_id = p_dept_id;
RETURN v_accounting_project;
END get_dept_project;
Initializing Collections
FUNCTION manipulate_project (p_dept_id NUMBER)
RETURN typ_projectlist
IS
v_changed_list typ_projectlist;
BEGIN
SELECT projects INTO v_accounting_project
FROM department WHERE dept_id = p_dept_id;
-- this example assigns one collection to another
v_changed_list := v_accounting_project;
RETURN v_changed_list;
END manipulate_project;
1
2

36
Referencing Collection Elements
-- sample caller program to the manipulate_project function
DECLARE
v_result_list typ_projectlist;
BEGIN
v_result_list := manage_dept_proj.manipulate_project(10);
FOR i IN 1..v_result_list.COUNT LOOP
dbms_output.put_line('Project #: '
||v_result_list(i).project_no);
dbms_output.put_line('Title: '||v_result_list(i).title);
dbms_output.put_line('Cost: ' ||v_result_list(i).cost);
END LOOP;
END;
Project #: 1001
Title: Travel Monitor
Cost: 400000
Project #: 1002
Title: Open World
Cost: 10000000

37
– EXISTS
– COUNT
– LIMIT
– FIRST and LAST
– PRIOR and NEXT
– EXTEND
– TRIM
– DELETE
Using Collection Methods
collection_name.method_name [(parameters)]

39
FUNCTION check_costs (p_project_list typ_projectlist)
RETURN boolean
IS
c_max_allowed NUMBER := 10000000;
i INTEGER;
v_flag BOOLEAN := FALSE;
BEGIN
i := p_project_list.FIRST ;
WHILE i IS NOT NULL LOOP
IF p_project_list(i).cost > c_max_allowed then
v_flag := TRUE;
dbms_output.put_line (p_project_list(i).title || '
exceeded allowable budget.');
RETURN TRUE;
END IF;
i := p_project_list.NEXT(i);
END LOOP;
RETURN null;
END check_costs;
Traverse collections with the following methods:

40
-- sample caller program to check_costs
set serverout on
DECLARE
v_project_list typ_projectlist;
BEGIN
v_project_list := typ_ProjectList(
typ_Project (1,'Dsgn New Expense Rpt', 3250),
typ_Project (2, 'Outsource Payroll', 120000),
typ_Project (3, 'Audit Accounts Payable',14250000));
IF manage_dept_proj.check_costs(v_project_list) THEN
dbms_output.put_line('Project rejected: overbudget');
ELSE
dbms_output.put_line('Project accepted, fill out forms.');
END IF;
END;
Audit Accounts Payable exceeded allowable budget.
Project rejected: overbudget
PROJECT_NO TITLE COSTS
1 Dsgn New Expense Rpt 3250
2 Outsource Payroll 120000
3 Audit Accounts Payable 14250000
V_PROJECT_LIST variable:

41
PROCEDURE update_a_project
(p_deptno NUMBER, p_new_project typ_Project, p_position NUMBER)
IS
v_my_projects typ_ProjectList;
BEGIN
v_my_projects := get_dept_project (p_deptno);
v_my_projects.EXTEND; --make room for new project
/* Move varray elements forward */
FOR i IN REVERSE p_position..v_my_projects.LAST - 1 LOOP
v_my_projects(i + 1) := v_my_projects(i);
END LOOP;
v_my_projects(p_position) := p_new_project; -- insert new one
UPDATE department SET projects = v_my_projects
WHERE dept_id = p_deptno;
END update_a_project;
Manipulating Individual Elements

42
Manipulating Individual Elements
-- check the table prior to the update:
------- ------------------------- ---------- ----------------------------- --
-- caller program to update_a_project
BEGIN
manage_dept_proj.update_a_project(20,
typ_Project(2002, 'AQM', 80000), 2);
END;
------- ------------------------- ---------- ----------------------------- --
20 Information Technology 2002 AQM 80000
-- check the table after the update:

43
Listing Characteristics for Collections
PL/SQL
Nested
Tables
DB
Nested
Tables
PL/SQL
Varrays
DB
Varrays
PL/SQL
Associative
Arrays
Maximum
size
No No Yes Yes Dynamic
Sparsity Can be No Dense Dense Yes
Storage N/A Stored out-of-
line
N/A Stored inline
(if < 4,000
bytes)
N/A
Ordering Does not
retain ordering
and subscripts
Does not
retain ordering
and subscripts
Retains
ordering and
subscripts
Retains
ordering and
subscripts
Retains
ordering and
subscripts

44
• Varrays involve fewer disk accesses and are more
efficient.
• Use nested tables for storing large amounts of data.
• Use varrays to preserve the order of elements in the
collection column.
• If you do not have a requirement to delete elements in
the middle of a collection, favor varrays.
• Varrays do not allow piecewise updates.
• After deleting the elements, release the unused
memory with
DBMS_SESSION.FREE_UNUSED_USER_MEMORY
Guidelines for Using Collections Effectively

46
• Fetch into a record when fetching from a cursor.
• Benefit
– No individual variables declaration is needed.
– You can automatically use the structure of the SELECT
column list.
Cursor Design: Use Records
DECLARE
CURSOR cur_cust IS
SELECT customer_id, cust_last_name, cust_email
FROM customers WHERE credit_limit = 1200;
v_cust_record cur_cust%ROWTYPE;
BEGIN
OPEN cur_cust;
LOOP
FETCH cur_cust INTO v_cust_record;
...

47
• Reference implicit cursor attributes immediately after the SQL
statement executes.
• Benefit
– Doing so ensures that you are dealing with the result of the
correct SQL statement.
Guidelines for Good Cursor Design
BEGIN
UPDATE customers
SET credit_limit = p_credit_limit
WHERE customer_id = p_cust_id;
get_avg_order(p_cust_id); -- procedure call
IF SQL%NOTFOUND THEN
...

48
• Create cursors with parameters.
• Benefit
– Parameters increase the cursor’s flexibility and reusability.
– Parameters help avoid scoping problems.
Use Cursor Parameters
CURSOR cur_cust
(p_crd_limit NUMBER, p_acct_mgr NUMBER)
IS
FROM customers
WHERE credit_limit = p_crd_limit
AND account_mgr_id = p_acct_mgr;
BEGIN
OPEN cur_cust(p_crd_limit_in, p_acct_mgr_in);
...
CLOSE cur_cust;
...
OPEN cur_cust(v_credit_limit, 145);
...
END;

49
– Simplify coding with cursor FOR loops.
– Benefit
• Reduces the volume of code
• Automatically handles the open, fetch, and close operations, and
defines a record type that matches the cursor definition
Using the Cursor For Loop
CREATE OR REPLACE PROCEDURE cust_pack
(p_crd_limit_in NUMBER, p_acct_mgr_in NUMBER)
IS
v_credit_limit NUMBER := 1500;
CURSOR cur_cust
(p_crd_limit NUMBER, p_acct_mgr NUMBER)
IS
FROM customers WHERE credit_limit = p_crd_limit
AND account_mgr_id = p_acct_mgr;
BEGIN
FOR cur_rec IN cur_cust (p_crd_limit_in, p_acct_mgr_in)
LOOP -- implicit open and fetch
...
END LOOP; -- implicit close
...
END;

50
CREATE OR REPLACE PROCEDURE cust_list
IS
CURSOR cur_cust IS
SELECT customer_id, cust_last_name, credit_limit*1.1
FROM customers;
cust_record cur_cust%ROWTYPE;
BEGIN
OPEN cur_cust;
LOOP
FETCH cur_cust INTO cust_record;
DBMS_OUTPUT.PUT_LINE('Customer ' ||
cust_record.cust_last_name || ' wants credit '
|| cust_record.(credit_limit * 1.1));
EXIT WHEN cur_cust%NOTFOUND;
END LOOP;
...
• Make a DBA happy: Close a cursor when it is no longer needed.
• Use column aliases in cursors for calculated columns fetched into
records declared with %ROWTYPE.
More Guidelines for Good Cursor Design
Use col. alias

5151
• A Ref Cursor is a Cursor variable
• It hold a pointer to the result set of a
previously opened cursor
• The actual SQL statement of the cursor is
dynamic and determined at execution time
• A single Ref Cursor can point to different
result sets at different times
Returning Result Sets From PL/SQL

52
Memory
1 Southlake, Texas 1400
2 San Francisco 1500
3 New Jersey 1600
4 Seattle, Washington 1700
5 Toronto 1800
Cursor Variables: Overview
REF
CURSOR
memory
locator

53
Working with Cursor Variables
Define and
declare the
cursor
variable.
Open the
cursor
variable.
Fetch rows
from the
result set.
Close the
cursor
variable.
1 2 3 4

54
• Strong REF CURSOR:
– Is restrictive
– Specifies a RETURN type
– Associates only with type-compatible queries
– Is less error prone
• Weak REF CURSOR:
– Is nonrestrictive
– Associates with any query
– Is very flexible
Strong Versus Weak REF CURSOR Variables

55
DECLARE
TYPE rt_cust IS REF CURSOR
RETURN customers%ROWTYPE;
...
Define a REF CURSOR type:
– ref_type_name is a type specified in subsequent
declarations.
– return_type represents a record type.
– RETURN keyword indicates a strong cursor.
Step 1: Defining a REF CURSOR Type
TYPE ref_type_name IS REF CURSOR
[RETURN return_type];

56
Declare a cursor variable of a cursor type:
– cursor_variable_name is the name of the cursor
variable.
– ref_type_name is the name of a REF CURSOR type.
Step 1: Declaring a Cursor Variable
DECLARE
TYPE rt_cust IS REF CURSOR
RETURN customers%ROWTYPE;
cv_cust rt_cust;
cursor_variable_name ref_type_name;

57
Options:
– Use %TYPE and %ROWTYPE.
– Specify a user-defined record in the RETURN
clause.
– Declare the cursor variable as the formal
parameter of a stored procedure or function.
Step 1: Declaring a REF CURSOR Return Type

58
• Associate a cursor variable with a multiple-row
SELECT statement.
• Execute the query.
• Identify the result set:
– cursor_variable_name is the name of the
cursor variable.
– select_statement is the SQL SELECT
statement.
Step 2: Opening a Cursor Variable
OPEN cursor_variable_name
FOR select_statement;

60
• Retrieve rows from the result set one at a
time.
• The return type of the cursor variable must be
compatible with the variables named in the
INTO clause of the FETCH statement.
Step 3: Fetching from a Cursor Variable
FETCH cursor_variable_name
INTO variable_name1
[,variable_name2,. . .]
| record_name;

61
• Disable a cursor variable.
• The result set is undefined.
• Accessing the cursor variable after it is closed
raises the INVALID_CURSOR predefined
exception.
Step 4: Closing a Cursor Variable
CLOSE cursor_variable_name;

62
You can pass query result sets among PL/SQL-
stored subprograms and various clients.
Passing Cursor Variables as Arguments
Pointer
to the
result
set
Access by a host variable
on the client side

63
Passing Cursor Variables as Arguments

65
Using the SYS_REFCURSOR Predefined Type
CREATE OR REPLACE PROCEDURE REFCUR
(p_num IN NUMBER)
IS
refcur sys_refcursor;
empno emp.empno%TYPE;
ename emp.ename%TYPE;
BEGIN
IF p_num = 1 THEN
OPEN refcur FOR SELECT empno, ename FROM emp;
DBMS_OUTPUT.PUT_LINE('Employee# Name');
DBMS_OUTPUT.PUT_LINE('----- -------');
LOOP
FETCH refcur INTO empno, ename;
EXIT WHEN refcur%NOTFOUND;
DBMS_OUTPUT.PUT_LINE(empno || ' ' || ename);
END LOOP;
ELSE
....
SYS_REFCURSOR is a built-in
REF CURSOR type that allows
any result set to be associated
with it.

67
• You cannot use cursor variables with remote
subprograms on another server.
• You cannot use comparison operators to test
cursor variables.
• You cannot assign a null value to cursor variables.
• You cannot use REF CURSOR types in CREATE
TABLE or VIEW statements.
• Cursors and cursor variables are not
interoperable.
Rules for Cursor Variables

68
Cursor variables have the following benefits:
– Are dynamic and ensure more flexibility
– Are not tied to a single SELECT statement
– Hold the value of a pointer
– Can reduce network traffic
– Give access to query work areas after a
block completes
Comparing Cursor Variables
with Static Cursors

7070
• Use explicit locking to deny access to other
sessions for the duration of a transaction
• Lock the rows before the update or delete
• Syntax:
Using the For Update Clause
SELECT ...
FROM ...
FOR UPDATE [OF column_reference][NOWAIT | WAIT n];

7171
• Select for Update default is to wait for locked
rows
• We can set the wait time to be limited with a
time frame or fail if rows are already locked
• We can also ask the query to skip the locked
rows and return the unlocked rows:
Handling Locked Rows
SELECT ...
FROM ...
FOR UPDATE SKIP LOCKED;

7272
• When using cursor with FOR UPDATE clause,
we might want to change that same record
• We can use the WHERE CURRENT OF cursor to
get a fast ROWID access to that row
Using WHERE CURRENT OF Clause
UPDATE employees
SET salary = 12000
WHERE CURRENT OF emp_cursor;

7373
• Include the RETURNING clause with an
INSERT, UPDATE, and DELETE to return column
values.
• Benefit:
– Eliminates the need to SELECT the row after DML
– Fewer network round trips
– Less server CPU time
– Fewer cursors
– Less server memory is required
Using the RETURNING Clause

7474
Using RETURNING Clause Example
DECLARE
v_emp_sal employees.salary%type;
BEGIN
UPDATE employees e
set salary = salary * 1.2
WHERE e.employee_id = 202
RETURNING e.salary into v_emp_sal;
dbms_output.put_line('The new salary is ' ||
v_emp_sal);
END;
/
The new salary is 7200

75
Use bulk binds to reduce context switches
between the PL/SQL engine and the SQL engine.
Using Bulk Binding
SQL enginePL/SQL run-time engine
PL/SQL block
FORALL j IN 1..1000
INSERT …
(OrderId(j),
OrderDate(j), …);
SQL
statement
executor
Procedural
statement
executor

76
Bind whole arrays of values simultaneously, rather
than looping to perform fetch, insert, update, and
delete on multiple rows.
– Instead of:
– Use:
Using Bulk Binding
...
FOR i IN 1 .. 50000 LOOP
INSERT INTO bulk_bind_example_tbl
VALUES(...);
END LOOP; ...
...
FORALL i IN 1 .. 50000
INSERT INTO bulk_bind_example_tbl
VALUES(...);
...

78
Use BULK COLLECT to improve performance:
Using Bulk Binding
CREATE OR REPLACE PROCEDURE process_customers
(p_account_mgr customers.account_mgr_id%TYPE)
IS
TYPE typ_numtab IS TABLE OF
customers.customer_id%TYPE;
TYPE typ_chartab IS TABLE OF
customers.cust_last_name%TYPE;
TYPE typ_emailtab IS TABLE OF
customers.cust_email%TYPE;
v_custnos typ_numtab;
v_last_names typ_chartab;
v_emails typ_emailtab;
BEGIN
BULK COLLECT INTO v_custnos, v_last_names, v_emails
FROM customers
WHERE account_mgr_id = p_account_mgr;
...
END process_customers;

79
Use the RETURNING clause to retrieve information
about the rows that are being modified:
Using Bulk Binding
DECLARE
TYPE typ_replist IS VARRAY(100) OF NUMBER;
TYPE typ_numlist IS TABLE OF
orders.order_total%TYPE;
repids typ_replist :=
typ_replist(153, 155, 156, 161);
totlist typ_numlist;
c_big_total CONSTANT NUMBER := 60000;
BEGIN
FORALL i IN repids.FIRST..repids.LAST
UPDATE orders
SET order_total = .95 * order_total
WHERE sales_rep_id = repids(i)
AND order_total > c_big_total
RETURNING order_total BULK COLLECT INTO Totlist;
END;

81
• You can use the SAVE EXCEPTIONS keyword in your
FORALL statements:
• Exceptions raised during execution are saved in the
%BULK_EXCEPTIONS cursor attribute.
• The attribute is a collection of records with two fields:
– Note that the values always refer to the most recently executed FORALL
statement.
Using SAVE EXCEPTIONS
FORALL index IN lower_bound..upper_bound
SAVE EXCEPTIONS
{insert_stmt | update_stmt | delete_stmt}
Field Definition
ERROR_INDEX Holds the iteration of the FORALL statement where the
exception was raised
ERROR_CODE Holds the corresponding Oracle error code

82
Handling FORALL Exceptions
DECLARE
TYPE NumList IS TABLE OF NUMBER;
num_tab NumList :=
NumList(100,0,110,300,0,199,200,0,400);
bulk_errors EXCEPTION;
PRAGMA EXCEPTION_INIT (bulk_errors, -24381 );
BEGIN
FORALL i IN num_tab.FIRST..num_tab.LAST
SAVE EXCEPTIONS
DELETE FROM orders WHERE order_total < 500000/num_tab(i);
EXCEPTION WHEN bulk_errors THEN
DBMS_OUTPUT.PUT_LINE('Number of errors is: '
|| SQL%BULK_EXCEPTIONS.COUNT);
FOR j in 1..SQL%BULK_EXCEPTIONS.COUNT
LOOP
DBMS_OUTPUT.PUT_LINE (
TO_CHAR(SQL%BULK_EXCEPTIONS(j).error_index) ||
' / ' ||
SQLERRM(-SQL%BULK_EXCEPTIONS(j).error_code) );
END LOOP;
END;
/

Dynamic SQL and SQL Injection
83

8484
• Used for two main reasons:
– Modify commands dynamically to retrieve or filter
columns based on input
– Run DDL and DCL commands from PL/SQL
– PLEASE don’t use it for running DMLs!
• Two main ways to run dynamic commands:
– EXECUTE IMMDIATE
– DBMS_SQL
Dynamic SQL

8585
• Easy syntax:
• Accepts all command types
• Accept bind variables with the USING clause
• Allows dynamic PL/SQL creation
Advantages of Native Dynamic SQL
EXECUTE_IMMEDIATE dynamic_string
{
INTO { define_variable [, define_variable ...] | record_name }
| BULK COLLECT INTO { collection_name [, collection_name ...] |
:host_array_name }
}
[ USING [ IN | OUT | IN OUT ] bind_argument
[, [ IN | OUT | IN OUT ] bind_argument] ... ] [ returning_clause ] ;

8686
• DBMS_SQL supports statements with unknown
number of inputs or outputs
• We can use DESCRIBE_COLUMNS procedure in
the DBMS_SQL package to describe columns for
a cursor opened/parsed through DBMS_SQL
• DBMS_SQL Supports SQL statements larger than
32 KB
• DBMS_SQL Lets you reuse SQL statements
Advantages of DBMS_SQL

8787
Using Execute Immediate
CREATE OR REPLACE PROCEDURE del_rows
(p_condition varchar2, p_rows_deld out number)
IS
BEGIN
EXECUTE IMMEDIATE 'DELETE FROM employees ' || p_condition;
p_rows_deld:=sql%ROWCOUNT;
END;
/
DECLARE
cnt number;
BEGIN
del_rows(‘where employee_id = 201’, cnt);
END;
/

8888
Using DBMS_SQL
CREATE OR REPLACE
PROCEDURE del_rows
(p_condition in varchar2, p_rows_del out number)
is
cursor_id integer;
BEGIN
cursor_id := dbms_sql.open_cursor;
dbms_sql.parse (cursor_id,'DELETE FROM employees ‘ ||p_condition,
dbms_sql.native);
p_rows_del := dbms_sql.execute (cursor_id);
dbms_sql.close_cursor (cursor_id);
END;
/

8989
• Starting Oracle 11g, we can transform a
DBMS_SQL cursor into a PL/SQL REF CURSOR
and vice versa
Transforming DBMS_SQL
Cursor into a REF CURSOR
DBMS_SQL.TO_REFCURSOR (cursor_number IN INTEGER) RETURN SYS_REFCURSOR;
DBMS_SQL.TO_CURSOR_NUMBER (rc IN OUT SYS_REFCURSOR) RETURN INTEGER;

9090
• Dynamic SQL is not checked until runtime
– Syntax
– Structure validity
– Permissions
• Dynamic SQL can be used as SQL injection
entry point
Disadvantages of Dynamic SQL

91
SQL injection is a technique for maliciously
exploiting applications that use client-supplied data
in SQL statements.
– Attackers trick the SQL engine into executing
unintended commands.
– SQL injection techniques may differ, but they all
exploit a single vulnerability in the application.
– To immunize your code against SQL injection attacks,
use bind arguments or validate and sanitize all input
concatenated to dynamic SQL.
Understanding SQL Injection

SQL Injection: Example
-- First order attack
CREATE OR REPLACE PROCEDURE GET_EMAIL
(p_last_name VARCHAR2 DEFAULT NULL)
AS
TYPE cv_custtyp IS REF CURSOR;
cv cv_custtyp;
v_email customers.cust_email%TYPE;
v_stmt VARCHAR2(400);
BEGIN
v_stmt := 'SELECT cust_email FROM customers
WHERE cust_last_name = '''|| p_last_name || '''';
DBMS_OUTPUT.PUT_LINE('SQL statement: ' || v_stmt);
OPEN cv FOR v_stmt;
LOOP
FETCH cv INTO v_email;
EXIT WHEN cv%NOTFOUND;
DBMS_OUTPUT.PUT_LINE('Email: '||v_email);
END LOOP;
CLOSE cv;
EXCEPTION WHEN OTHERS THEN
dbms_output.PUT_LINE(sqlerrm);
dbms_output.PUT_LINE('SQL statement: ' || v_stmt);
END;
String literals that are incorrectly
validated or not validated are
concatenated into a dynamic SQL
statement, and interpreted as code by th
SQL engine.

Assessing Vulnerability
Code
Uses bind
arguments for all
dynamic
components
Vulnerable
Safe
Contains dynamic
SQL?
Filters input
correctly?
No
No
No
Yes
Yes
YesLateral injection
vulnerability
No

94
Use the following strategies to reduce attack
surface:
– Use invoker’s rights.
– Reduce arbitrary inputs.
– Use Bind arguments.
– The Filter pitfall.
Reducing the Attack Surface

95
• Give out privileges appropriately.
• Run code with invoker’s rights when possible.
• Ensure that the database privilege model is
upheld when using definer’s rights.
Avoiding Privilege Escalation
Invoker’s rights
Definer’s rights

96
• Using invoker’s rights:
– Helps to limit the privileges
– Helps to minimize the security exposure.
• The following example does not use invoker's rights:
Using Invoker’s Rights
CREATE OR REPLACE
PROCEDURE change_password(p_username VARCHAR2 DEFAULT NULL,
p_new_password VARCHAR2 DEFAULT NULL)
IS
v_sql_stmt VARCHAR2(500);
BEGIN
v_sql_stmt := 'ALTER USER '||p_username ||' IDENTIFIED BY '
|| p_new_password;
EXECUTE IMMEDIATE v_sql_stmt;
END change_password;
GRANT EXECUTE ON change_password to OE, HR, SH;
1
2
Note the use of dynamic SQL with
concatenated input values.

97
• OE is successful at changing the SYS password,
because, by default, CHANGE_PASSWORD executes
with SYS privileges:
• Add the AUTHID to change the privileges to the invokers:
EXECUTE sys.change_password ('SYS', 'mine')
Using Invoker’s Rights
CREATE OR REPLACE
PROCEDURE change_password(p_username VARCHAR2 DEFAULT NULL,
p_new_password VARCHAR2 DEFAULT NULL)
AUTHID CURRENT_USER
IS
v_sql_stmt VARCHAR2(500);
BEGIN
v_sql_stmt := 'ALTER USER '||p_username ||' IDENTIFIED BY '
|| p_new_password;
EXECUTE IMMEDIATE v_sql_stmt;
END change_password;

98
• Reduce the end-user interfaces to only those that are actually
needed.
– In a Web application, restrict users to accessing specified Web
pages.
– In a PL/SQL API, expose only those routines that are intended for
customer use.
• Where user input is required, make use of language features
to ensure that only data of the intended type can be
specified.
– Do not specify a VARCHAR2 parameter when it will be used as a
number.
– Do not use numbers if you need only positive integers; use natural
instead.
Reducing Arbitrary Inputs

99
DBMS_ASSERT functions:
Understanding DBMS_ASSERT
Function Description
ENQUOTE_LITERAL Encloses string literal in single quotes
SIMPLE_SQL_NAME Verifies that the string is a simple SQL name

100
Formatting Oracle Identifiers
– Example 1: The object name used as an identifier:
SELECT count(*) records FROM orders;
– Example 2: The object name used as a literal:
SELECT num_rows FROM user_tables
WHERE table_name = 'ORDERS';
– Example 3: The object name used as a quoted (normal
format) identifier:
• The "orders" table referenced in example 3 is a different
table compared to the orders table in examples 1 and 2.
• It is vulnerable to SQL injection.
SELECT count(*) records FROM "orders";

101
– For your identifiers, determine:
1.Where will the input come from: user or data dictionary?
2.What verification is required?
3.How will the result be used, as an identifier or a literal
value?
– These three factors affect:
• What preprocessing is required (if any) prior to calling
the verification functions
• Which DBMS_ASSERT verification function is required
• What post-processing is required before the identifier
can actually be used
Working with Identifiers in Dynamic SQL

102
Avoiding Injection by Using
DBMS_ASSERT.ENQUOTE_LITERAL
CREATE OR REPLACE PROCEDURE Count_Rows(w in varchar2)
authid definer as
Quote constant varchar2(1) := '''';
Quote_Quote constant varchar2(2) := Quote||Quote;
Safe_Literal varchar2(32767) :=
Quote||replace(w,Quote,Quote_Quote)||Quote;
Stmt constant varchar2(32767) :=
'SELECT count(*) FROM t WHERE a='||
DBMS_ASSERT.ENQUOTE_LITERAL(Safe_Literal);
Row_Count number;
BEGIN
EXECUTE IMMEDIATE Stmt INTO Row_Count;
DBMS_OUTPUT.PUT_LINE(Row_Count||' rows');
END;
/
Verify whether the literal
is well-formed.

105
Avoiding Injection by Using
DBMS_ASSERT.SIMPLE_SQL_NAME
CREATE OR REPLACE PROCEDURE show_col2 (p_colname varchar2,
p_tablename varchar2)
AS
type t is varray(200) of varchar2(25);
Results t;
Stmt CONSTANT VARCHAR2(4000) :=
'SELECT '||dbms_assert.simple_sql_name( p_colname ) || ' FROM
'|| dbms_assert.simple_sql_name( p_tablename ) ;
BEGIN
DBMS_Output.Put_Line ('SQL Stmt: ' || Stmt);
EXECUTE IMMEDIATE Stmt bulk collect into Results;
for j in 1..Results.Count() loop
DBMS_Output.Put_Line(Results(j));
end loop;
END show_col2;
Verify that the input string conforms to
the basic characteristics of a simple
SQL name.

107
• Most common vulnerability:
– Dynamic SQL with string concatenation
• Your code design must:
– Avoid input string concatenation in dynamic SQL
– Use bind arguments, whether automatically via static
SQL or explicitly via dynamic SQL statements
Using Bind Arguments

108
• Filter parameter:
– P_WHERE_CLAUSE is a filter.
– It is difficult to protect against SQL injection.
• Prevention methods:
– Do not specify APIs that allow arbitrary query
parameters to be exposed.
– Any existing APIs with this type of functionality must
be deprecated and replaced with safe alternatives.
Beware of Filter Parameters
stmt := 'SELECT session_id FROM sessions
WHERE' || p_where_clause;

109109
• An Autonomous Transaction is an independent
Transaction started by another transaction.
• The main transaction will hold for the AT and wait
until it is completed
• Uses PRAGMA compiler directive
• Allowed in individual routines
• Commonly used for loggers, progress bars, and
concurrent operations
Autonomous Transactions

110110
AT Example: Main Transaction
DECLARE
tmp NUMBER;
BEGIN
FOR i IN 1..10
LOOP
tmp := i;
INSERT INTO t
VALUES (TRUNC(i/2));
END LOOP;
EXCEPTION
WHEN DUP_VAL_ON_INDEX THEN
logger('Failed on ' || TO_CHAR(tmp));
ROLLBACK;
RAISE;
END;
/

111111
AT Example: Logger (AT)
PROCEDURE logger (message IN VARCHAR2) IS
PRAGMA AUTONOMOUS_TRANSACTION
BEGIN
INSERT INTO logger VALUES (sysdate, message);
COMMIT;
END;
/

11g Useful Features
for Developers
112

113113
• For a specified measure, LISTAGG orders data
within each group specified in the ORDER BY
clause and then concatenates the values of
the measure column
• Limited to output of 4000 chars
The LISTAGG Function
LISTAGG(measure_expr [, 'delimiter'])
WITHIN GROUP (order_by_clause) [OVER
query_partition_clause]

114114
Using the LISTAGG Function Example
SELECT department_id "Dept", hire_date "Date“,
last_name "Name", LISTAGG(last_name, ', ')
WITHIN GROUP (ORDER BY hire_date, last_name)
OVER (PARTITION BY department_id) as "Emp_list"
FROM employees
WHERE hire_date < '01-SEP-2003'
ORDER BY "Dept", "Date", "Name";

115115
• Returns the N-th values in an ordered set of
values
• Different default window: RANGE BETWEEN
UNBOUNDED PRECEDING AND CURRENT
ROW
The NTH_VALUE Analytic Function
NTH_VALUE (measure_expr, n)
[ FROM { FIRST | LAST } ][ { RESPECT | IGNORE } NULLS ]
OVER (analytic_clause)

116116
Using NTH_VALUE Analytic Function
Example
SELECT prod_id, channel_id, MIN(amount_sold),
NTH_VALUE ( MIN(amount_sold), 2) OVER (PARTITION BY
prod_id ORDER BY channel_id
ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED
FOLLOWING) nv
FROM sales
WHERE prod_id BETWEEN 13 and 16
GROUP BY prod_id, channel_id;

117117
• Virtual columns are dynamic (not stored) columns of a table (no
views required)
• Virtual columns obtain their value be evaluating as expression:
– Columns from the same table
– Constraints
– Function Calls (user defined)
• Might be used for
– Eliminate views
– Control table partitioning
– Manage “Binary” XMLType data
– Index values (function based index)
Virtual Columns

118118
Virtual Columns Example
column_name [datatype] [GENERATED ALWAYS] AS (expression) [VIRTUAL]
CREATE TABLE employees (
id NUMBER,
first_name VARCHAR2(10),
last_name VARCHAR2(10),
salary NUMBER(9,2),
comm1 NUMBER(3),
comm2 NUMBER(3),
salary1 AS (ROUND(salary*(1+comm1/100),2)),
salary2 NUMBER GENERATED ALWAYS AS
(ROUND(salary*(1+comm2/100),2)) VIRTUAL,
CONSTRAINT employees_pk PRIMARY KEY (id)
);

120
• A single trigger on a table that allows you to
specify actions for each of the following four
timing points:
– Before the firing statement
– Before each row that the firing statement affects
– After each row that the firing statement affects
– After the firing statement
What Is a Compound Trigger?

121
• The compound trigger body supports a
common PL/SQL state that the code for each
timing point can access.
• The compound trigger common state is:
– Established when the triggering statement starts
– Destroyed when the triggering statement
completes
• A compound trigger has a declaration section
and a section for each of its timing points.
Working with Compound Triggers

122
• You can use compound triggers to:
– Program an approach where you want the actions
you implement for the various timing points to
share common data.
– Accumulate rows destined for a second table so
that you can periodically bulk-insert them
– Avoid the mutating-table error (ORA-04091)by
allowing rows destined for a second table to
accumulate and then bulk-inserting them
The Benefits of Using a Compound Trigger

123
• A compound trigger defined on a table has
one or more of the following timing-point
sections. Timing-point sections must appear in
the order shown in the table.
Timing-Point Sections of a
Table Compound Trigger
Timing Point Compound Trigger Section
Before the triggering statement executes BEFORE statement
After the triggering statement executes AFTER statement
Before each row that the triggering statement affects BEFORE EACH ROW
After each row that the triggering statement affects AFTER EACH ROW

124
Compound Trigger Structure for Tables
CREATE OR REPLACE TRIGGER schema.trigger
FOR dml_event_clause ON schema.table
COMPOUND TRIGGER
-- Initial section
-- Declarations
-- Subprograms
-- Optional section
BEFORE STATEMENT IS ...;
-- Optional section
BEFORE EACH ROW IS ...;
-- Optional section
AFTER EACH ROW IS ...;
-- Optional section
AFTER STATEMENT IS ...;
1
2

125
• A mutating table is:
– A table that is being modified by an UPDATE, DELETE, or
INSERT statement, or
– A table that might be updated by the effects of a DELETE
CASCADE constraint
• The session that issued the triggering statement cannot
query or modify a mutating table.
• This restriction prevents a trigger from seeing an
inconsistent set of data.
• This restriction applies to all triggers that use the FOR
EACH ROW clause.
• Views being modified in the INSTEAD OF triggers are not
considered mutating.
Trigger Restrictions on Mutating Tables

126
Using a Compound Trigger to
Resolve the Mutating Table Error
CREATE OR REPLACE TRIGGER check_salary
FOR INSERT OR UPDATE OF salary, job_id
ON employees
WHEN (NEW.job_id <> 'AD_PRES')
COMPOUND TRIGGER
TYPE salaries_t IS TABLE OF employees.salary%TYPE;
min_salaries salaries_t;
max_salaries salaries_t;
TYPE department_ids_t IS TABLE OF employees.department_id%TYPE;
department_ids department_ids_t;
TYPE department_salaries_t IS TABLE OF employees.salary%TYPE
INDEX BY VARCHAR2(80);
department_min_salaries department_salaries_t;
department_max_salaries department_salaries_t;
-- example continues on next slide

127
Using a Compound Trigger to Resolve
the Mutating Table Error
. . .
BEFORE STATEMENT IS
BEGIN
SELECT MIN(salary), MAX(salary), NVL(department_id, -1)
BULK COLLECT INTO min_Salaries, max_salaries, department_ids
FROM employees
GROUP BY department_id;
FOR j IN 1..department_ids.COUNT() LOOP
department_min_salaries(department_ids(j)) := min_salaries(j);
department_max_salaries(department_ids(j)) := max_salaries(j);
END LOOP;
END BEFORE STATEMENT;
AFTER EACH ROW IS
BEGIN
IF :NEW.salary < department_min_salaries(:NEW.department_id)
OR :NEW.salary > department_max_salaries(:NEW.department_id) THEN
RAISE_APPLICATION_ERROR(-20505,'New Salary is out of acceptable
range');
END IF;
END AFTER EACH ROW;
END check_salary;

128
• A compound trigger must be a DML trigger and defined on
either a table or a view
• An exception that occurs in one section must be handled in
that section. It cannot transfer control to another section
• :OLD and :NEW cannot appear in the declaration, BEFORE
STATEMENT, or the AFTER STATEMENT sections
• Only the BEFORE EACH ROW section can change the value of
:NEW
• The firing order of compound triggers is not guaranteed
unless you use the FOLLOWS clause
Compound Trigger Restrictions

129129
• To ensure that a trigger fires after certain
other triggers on the same object, use the
FOLLOWS clause
• Lets you order the executions of multiple
triggers relative to each other
• Applies to both compound and simple triggers
• Applies only to the section of the compound
trigger with the same timing point as the
simple trigger
FOLLOWS Clause

130130
• Consider two AFTER ROW ... FOR UPDATE triggers
defined on the same table. One trigger needs to
reference the :OLD value and the other trigger
needs to change the :OLD value. In this case, you
can use the FOLLOWS clause to order the firing
sequence
FOLLOWS Clause Example
CREATE OR REPLACE TRIGGER change_product
AFTER UPDATE of product_id ON order_items
FOR EACH ROW
FOLLOWS oe1.compute_total
BEGIN
dbms_output.put_line ('Do processing here…');
END;

Pivot and Unpivot
Turning things around!
131

PIVOT and UNPIVOT Clauses
of the SELECT Statement
• You can use the PIVOT operator of the
SELECT statement to write cross-tabulation
queries that rotate the column values into new
columns, aggregating data in the process.
• You can use the UNPIVOT operator of the
SELECT statement to rotate columns into
values of a column.
PIVOT UNPIVOT

133
Pivoting on the QUARTER
Column: Conceptual Example
30,000
40,000
60,000
30,000
40,000
20,000
AMOUNT_
SOLD
2,500Q1IUSAKids Jeans
2,000Q2CJapanKids Jeans
2,000Q3SUSAShorts
I
P
C
CHANNEL
Kids Jeans
Shorts
Shorts
PRODUCT
1,000Q2Germany
1,500Q4USA
Q2
QUARTER
2,500Poland
QUANTITY_
SOLD
COUNTRY
2,000
Q3
Kids Jeans
Shorts
PRODUCT
3,500
2,000
Q2
1,5002,500
Q4Q1

134134
• Pivoting the data before 11g was a complex
query which required the use of the CASE or
DECODE functions
Pivoting Before Oracle 11g
select product,
sum(case when quarter = 'Q1' then amount_sold else null end) Q1,
sum(case when quarter = 'Q4' then amount_sold else null end) Q4
from sales
group by product;

135
Pivot Clause Syntax
table_reference PIVOT [ XML ]
( aggregate_function ( expr ) [[AS] alias ]
[, aggregate_function ( expr ) [[AS] alias ] ]...
pivot_for_clause
pivot_in_clause )
-- Specify the column(s) to pivot whose values are to
-- be pivoted into columns.
pivot_for_clause =
FOR { column |( column [, column]... ) }
-- Specify the pivot column values from the columns you
-- specified in the pivot_for_clause.
pivot_in_clause =
IN ( { { { expr | ( expr [, expr]... ) } [ [ AS] alias] }...
| subquery | { ANY | ANY [, ANY]...} } )

137
Creating a New View: Example
CREATE OR REPLACE VIEW sales_view AS
SELECT
prod_name AS product,
country_name AS country,
channel_id AS channel,
SUBSTR(calendar_quarter_desc, 6,2) AS quarter,
SUM(amount_sold) AS amount_sold,
SUM(quantity_sold) AS quantity_sold
FROM sales, times, customers, countries, products
WHERE sales.time_id = times.time_id AND
sales.prod_id = products.prod_id AND
sales.cust_id = customers.cust_id AND
customers.country_id = countries.country_id
GROUP BY prod_name, country_name, channel_id,
SUBSTR(calendar_quarter_desc, 6, 2);

139
Selecting the SALES VIEW Data
SELECT product, country, channel, quarter, quantity_sold
FROM sales_view;
PRODUCT COUNTRY CHANNEL QUARTER QUANTITY_SOLD
------------ ------------ ---------- -------- -------------
Y Box Italy 4 01 21
Y Box Italy 4 02 17
Y Box Italy 4 03 20
. . .
Y Box Japan 2 01 35
Y Box Japan 2 02 39
Y Box Japan 2 03 36
Y Box Japan 2 04 46
Y Box Japan 3 01 65
. . .
Bounce Italy 2 01 34
Bounce Italy 2 02 43
. . .
9502 rows selected.

140
Pivoting the QUARTER Column
in the SH Schema: Example
SELECT *
FROM
(SELECT product, quarter, quantity_sold
FROM sales_view) PIVOT (sum(quantity_sold)
FOR quarter IN ('01', '02', '03', '04'))
ORDER BY product DESC;
. . .

142
Unpivoting the QUARTER Column:
Conceptual Example
2,000
Q3
Kids Jeans
Shorts
PRODUCT
3,500
2,000
Q2
1,5002,500
Q4Q1
2,500Q1Kids Jeans
2,000Q2Kids Jeans
3,500Q2Shorts
1,500Q4Kids Jeans
Q3
QUARTER
2,000Shorts
SUM_OF_QUANTITYPRODUCT

143143
• Univoting the data before 11g requires
multiple queries on the table using the
UNION ALL operator
Unpivoting Before Oracle 11g
SELECT *
FROM (
SELECT product, '01' AS quarter, Q1_value FROM sales
UNION ALL
UNION ALL
UNION ALL
);

144
• An UNPIVOT operation does not reverse a
PIVOT operation; instead, it rotates data found
in multiple columns of a single row into multiple
rows of a single column.
• If you are working with pivoted data, UNPIVOT
cannot reverse any aggregations that have been
made by PIVOT or any other means.
Using the UNPIVOT Operator
UNPIVOT

145
• The UNPIVOT clause rotates columns from a
previously pivoted table or a regular table into
rows. You specify:
– The measure column or columns to be unpivoted
– The name or names for the columns that result from
the UNPIVOT operation
– The columns that are unpivoted back into values of
the column specified in pivot_for_clause
• You can use an alias to map the column name to
another value.
Using the UNPIVOT Clause

146
UNPIVOT Clause Syntax
table_reference UNPIVOT [{INCLUDE|EXCLUDE} NULLS]
-- specify the measure column(s) to be unpivoted.
( { column | ( column [, column]... ) }
unpivot_for_clause
unpivot_in_clause )
-- Specify one or more names for the columns that will
-- result from the unpivot operation.
unpivot_for_clause =
FOR { column | ( column [, column]... ) }
-- Specify the columns that will be unpivoted into values of
-- the column specified in the unpivot_for_clause.
unpivot_in_clause =
( { column | ( column [, column]... ) }
[ AS { constant | ( constant [, constant]... ) } ]
[, { column | ( column [, column]... ) }
[ AS { constant | ( constant [, constant]...) } ] ]...)

147
Creating a New Pivot Table: Example
. . .
CREATE TABLE pivotedtable AS
SELECT *
FROM
(SELECT product, quarter, quantity_sold
FROM sales_view) PIVOT (sum(quantity_sold)
FOR quarter IN ('01' AS Q1, '02' AS Q2,
'03' AS Q3, '04' AS Q4));
SELECT * FROM pivotedtable
ORDER BY product DESC;

148
Unpivoting the QUARTER Column
in the SH Schema: Example
SELECT *
FROM pivotedtable
UNPIVOT (quantity_sold For Quarter IN (Q1, Q2, Q3, Q4))
ORDER BY product DESC, quarter;
. . .

Oracle 12c New Features
for Developers
149

150150
• Oracle 12c allows SQL VARCHAR2 to be the same
size as PL/SQL VARCHAR2
• This is not the default behavior and needs to be
turned on:
• Create table with 32k varchar2:
32K VARCHAR2/NVARCHAR2
ALTER SYSTEM set MAX_STRING_SIZE = EXTENDED scope = SPFILE;
CREATE TABLE Applicants
(id NUMBER GENERATED AS IDENTITY,
first_name varchar2(30),
last_name varchar2(30),
application date,
CV varchar2(32767)
);

151151
• Columns might be marked “invisible”
• The invisible column will not be available
unless explicitly named
• This is very useful when doing application
migration
Invisible Columns

152152
Invisible Column Example
CREATE TABLE tab1 (
id NUMBER,
description VARCHAR2(50) INVISIBLE
);
INSERT INTO tab1 VALUES (1);
SELECT * FROM tab1;
ID
----------
1
INSERT INTO tab1 (id, description) VALUES (2, 'TWO');
COMMIT;
SELECT id, description
FROM tab1;
ID DESCRIPTION
---------- ----------------------------------------------
1
2 TWO

153
• In previous releases, there was no direct
equivalent of the AutoNumber or Identity
functionality of other database engines
• This behavior had to be implemented using a
combination of sequences and trigger
• Oracle 12c introduces the ability to define an
identity clause against a table column defined
using a numeric type
• User should have the create sequence privilege
• Identity columns are always not null
Identity Column Type

154
• ALWAYS
– Forces the use of the identity. If an insert statement
references the identity column, an error is produced
• BY DEFAULT
– Allows using the identity if the column isn't referenced
in the insert statement. If the column is referenced,
the specified value will be used in place of the identity
• BY DEFAULT ON NULL
– Allows the identity to be used if the identity column is
referenced, but a value of NULL is specified
Identity Column Type – Options
Identity.sql

155
• You can specify only one identity column per table
• When specifying identity clause, you must specify a
numeric data type for datatype in the column
definition clause
• When specifying identity clause, you cannot specify
the DEFAULT clause in the column definition clause
• When specifying identity clause,
the NOT NULL constraint is implicitly specified
• CREATE TABLE AS SELECT will not inherit the identity
property on a column
Identity Column Type – Restrictions

156
• You can specify CURRVAL and NEXTVAL as default
values for a column
• Default value is used when the column is not
referenced by the insert or when the DEFAULT
keyword is used
• Gives you the ability to auto-populate master-
detail relationships
– Only makes sense if you can guarantee the inserts into
the detail table would always immediately follow the
insert into the master table
Default Value Using a Sequence
Default_with_Sequence.sql

157
• In an insert statement, when the column is
explicitly referenced, even when using the
value NULL, the default value is not used
• Oracle database 12c allows you to modify this
behavior using the ON NULL clause in the
default definition
Default Value on Explicit Nulls
Default_with_Null.sql

158
• You can define PL/SQL functions and procedures in the
WITH clause of a subquery and then use them as you
would any other built-in or user-defined function
• The “;” does not work as a terminator to the SQL
statement when the PL/SQL declaration is included in
the WITH clause
• Functions defined in the PL/SQL declaration section of
the WITH clause take precedence over objects with the
same name defined at the schema level
• Provides better performance as compared with schema
level functions
Calling PL/SQL from SQL
PLSQL_from_SQL.sql

159
Top-N Queries
• A Top-N query is used to retrieve the top or
bottom N rows from an ordered set
• Combining two Top-N queries gives you the
ability to page through an ordered set
• Oracle 12c has introduced the row limiting
clause to simplify Top-N queries

160160
• This is ANSI syntax
• The default offset is 0
• Null values in offset, rowcount or percent
will return no rows
Top-N in 12c
[ OFFSET offset { ROW | ROWS } ]
[ FETCH { FIRST | NEXT } [ { rowcount | percent PERCENT } ]
{ ROW | ROWS } { ONLY | WITH TIES } ]

161161
Top-N Examples
SELECT last_name, salary
FROM hr.employees
ORDER BY salary DESC
FETCH FIRST 5 ROWS ONLY;
FROM hr.employees
FETCH FIRST 5 ROWS WITH TIES;
FROM hr.employees
FETCH FIRST 10 PERCENT ROWS ONLY;

162162
• Before 12c we had to use the rownum pseudo
column to filter out rows
• That will require sorting the entire rowset
Paging Before 12c
SELECT val
FROM (SELECT val, rownum AS rnum
FROM (SELECT val
FROM rownum_order_test
ORDER BY val)
WHERE rownum <= 10)
WHERE rnum >= 5;

163163
• After 12c we have a syntax improvement for
paging using the Top-N queries
• This will use ROW_NUMBER and RANK in the
background – there is no real optimization
improvements
Paging in Oracle 12c
SELECT val
FROM rownum_order_test
ORDER BY val
OFFSET 4 ROWS FETCH NEXT 5 ROWS ONLY;

Pattern Matching in
Oracle 12c
164

165165
• Identify and group rows with consecutive values
• Consecutive in this regards – row after row
• Uses regular expression like syntax to find
patterns
What is Pattern Matching

166166
• Finding sequences of events in security
applications
• Locating dropped calls in a CDR listing
• Financial price behaviors (V-shape, W-shape
U-shape, etc.)
• Fraud detection and sensor data analysis
Common Business Challenges

167167
MATCH_RECOGNIZE Syntax
SELECT
FROM [row pattern input table]
MATCH_RECOGNIZE
( [ PARTITION BY <cols> ]
[ ORDER BY <cols> ]
[ MEASURES <cols> ]
[ ONE ROW PER MATCH | ALL ROWS PER MATCH ]
[ SKIP_TO_option]
PATTERN ( <row pattern> )
DEFINE <definition list>
)

169169
• Find Simple V-Shape with 1 row output per
match
MATCH_RECOGNIZE Example
SELECT *
FROM Ticker MATCH_RECOGNIZE (
PARTITION BY symbol
ORDER BY tstamp
MEASURES STRT.tstamp AS start_tstamp,
LAST(DOWN.tstamp) AS bottom_tstamp,
LAST(UP.tstamp) AS end_tstamp
ONE ROW PER MATCH
AFTER MATCH SKIP TO LAST UP
PATTERN (STRT DOWN+ UP+)
DEFINE
DOWN AS DOWN.price < PREV(DOWN.price),
UP AS UP.price > PREV(UP.price)
) MR
ORDER BY MR.symbol, MR.start_tstamp;

171171
• Our goal: find uninterrupted sequences in a
book
• This can be useful for detecting missing
records or sequential behavior
Pages in a Book Example

172
Building Our Query
1. Define input
2. Pattern Matching
3. Order input
4. Process pattern
5. using defined conditions
6. Output: rows per match
7. Output: columns per row
8. Where to go after match?
SELECT *
FROM book_pages
MATCH_RECOGNIZE (
ORDER BY page
PATTERN (A B*)
DEFINE B AS page = PREV(page)+1
ONE ROW PER MATCH
MEASURES
A.page firstpage,
LAST(page) lastpage,
COUNT(*) cnt
AFTER MATCH SKIP PAST LAST ROW
);
SELECT *
FROM book_pages
MATCH_RECOGNIZE (
ORDER BY page
MEASURES
A.page firstpage,
LAST(page) lastpage,
COUNT(*) cnt
ONE ROW PER MATCH
AFTER MATCH SKIP PAST LAST ROW
PATTERN (A B*)
DEFINE B AS page = PREV(page)+1
);

173173173173
And The Output…
FIRSTPAGE LASTPAGE CNT
---------- ---------- ----------
1 3 3
5 7 3
10 15 6
42 42 1

174174
• Session-specific sequence
• Truncate CASCADE command
• Temporal Validity
• Temporary Undo
• Online DML Operations
• And tons of new features for DBAs too!
More 12c Developers’ Features…

175175
• More information and examples could be
found on my Blog:
http://www.realdbamagic.com/he/pivot-a-table/
http://www.realdbamagic.com/he/12c-top-n-query/
http://www.realdbamagic.com/he/with-pl-sql-oracle-
12c/
http://www.realdbamagic.com/he/session-level-
sequence-12c/
More Examples…

177177
• Native and Interpreted Compilation
• Fine-grain Dependency management
• Code inlining
Compiling PL/SQL

178
Two compilation methods:
– Interpreted compilation
• Default compilation method
• Interpreted at run time
– Native compilation
• Compiles into native code
• Stored in the SYSTEM tablespace
Native and Interpreted Compilation

179
• Use the interpreted mode when (typically during
development):
– You are using a debugging tool, such as SQL Developer
– You need the code compiled quickly
• Use the native mode when (typically post
development):
– Your code is heavily PL/SQL based
– You are looking for increased performance in
production
Deciding on a Compilation Method
Native
Interpreted

180
• PLSQL_CODE_TYPE: Specifies the compilation mode for the
PL/SQL library units
• PLSQL_OPTIMIZE_LEVEL: Specifies the optimization level
to be used to compile the PL/SQL library units
• In general, for fastest performance, use the following setting:
Setting the Compilation Method
PLSQL_CODE_TYPE = { INTERPRETED | NATIVE }
PLSQL_OPTIMIZE_LEVEL = { 0 | 1 | 2 | 3}
PLSQL_CODE_TYPE = NATIVE
PLSQL_OPTIMIZE_LEVEL = 2

182
Use the USER|ALL|DBA_PLSQL_OBJECT_SETTINGS data
dictionary views to display the settings for a PL/SQL object:
Viewing the Compilation Settings
DESCRIBE ALL_PLSQL_OBJECT_SETTINGS
Name Null? Type
------------------------- -------- --------------------
OWNER NOT NULL VARCHAR2(30)
NAME NOT NULL VARCHAR2(30)
TYPE VARCHAR2(12)
PLSQL_OPTIMIZE_LEVEL NUMBER
PLSQL_CODE_TYPE VARCHAR2(4000)
PLSQL_DEBUG VARCHAR2(4000)
PLSQL_WARNINGS VARCHAR2(4000)
NLS_LENGTH_SEMANTICS VARCHAR2(4000)
PLSQL_CCFLAGS VARCHAR2(4000)
PLSCOPE_SETTINGS VARCHAR2(4000)

183
Viewing the Compilation Settings
SELECT name, plsql_code_type, plsql_optimize_level
FROM user_plsql_object_settings;
NAME PLSQL_CODE_TYP PLSQL_OPTIMIZE_LEVEL
-------------------- -------------- --------------------
ACTIONS_T INTERPRETED 2
ACTION_T INTERPRETED 2
ACTION_V INTERPRETED 2
ADD_ORDER_ITEMS INTERPRETED 2
CATALOG_TYP INTERPRETED 2
CATEGORY_TYP INTERPRETED 2
CATEGORY_TYP INTERPRETED 2
COMPOSITE_CATEGORY_TYP INTERPRETED 2
...

184
• This requires DBA privileges.
• The PLSQL_CODE_TYPE compilation
parameter must be set to NATIVE.
• The benefits apply to all the built-in PL/SQL
packages that are used for many database
operations.
Setting Up a Database for Native Compilation
ALTER SYSTEM SET PLSQL_CODE_TYPE = NATIVE;

185
Compiling a Program Unit for Native
Compilation
FROM user_plsql_object_settings
WHERE name = 'ADD_ORDER_ITEMS';
NAME PLSQL_CODE_T PLSQL_OPTIMIZE_LEVEL
---------------------- ------------ --------------------
ADD_ORDER_ITEMS INTERPRETED 2
ALTER SESSION SET PLSQL_CODE_TYPE = 'NATIVE';
ALTER PROCEDURE add_order_items COMPILE;
FROM user_plsql_object_settings
WHERE name = 'ADD_ORDER_ITEMS';
NAME PLSQL_CODE_T PLSQL_OPTIMIZE_LEVEL
---------------------- ------------ --------------------
ADD_ORDER_ITEMS NATIVE 2
1
2
3
4

186186
• We can turn on checking for certain warning conditions
• Warning messages can be issued during compilation of
PL/SQL subprograms (not for anonymous blocks )
• Use the SQL*Plus SHOW ERRORS command or query
the USER_ERRORS data dictionary view, to see any
warnings generated during compilation
• PL/SQL warning messages use the prefix PLW
• Use PLSQL_WARNINGS initialization parameter, or the
DBMS_WARNING package
PL/SQL Compile-Time Warnings

187187
• SEVERE: Messages for conditions that might cause
unexpected behavior or wrong results, such as aliasing
problems with parameters
• PERFORMANCE: Messages for conditions that might cause
performance problems, such as passing a VARCHAR2 value
to a NUMBER column in an INSERT statement.
• INFORMATIONAL: Messages for conditions that do not
have an effect on performance or correctness, but that you
might want to change to make the code more
maintainable, such as unreachable code that can never be
executed.
• All: refer to all warning messages
PL/SQL Warning Categories

188188
• Can be set at
– System level
– Session level
– Single compilation level
PLSQL_WARNINGS Parameter
ALTER SYSTEM SET PLSQL_WARNINGS='ENABLE:PERFORMANCE';
ALTER SESSION SET PLSQL_WARNINGS='DISABLE:ALL';
ALTER SESSION SET PLSQL_WARNINGS='ENABLE:SEVERE',
'DISABLE:PERFORMANCE', 'ERROR:07204';
ALTER PROCEDURE query_emp COMPILE
PLSQL_WARNINGS='ENABLE:ALL';

189189
• This warning means that the OTHERS handler
of PL/SQL subroutine can exit without
executing some form of RAISE or a call to the
standard RAISE_APPLICATION_ERROR
procedure.
• Good programming practices suggest that the
OTHERS handler should pass an exception
upward to avoid the risk of having exceptions
go unnoticed
PLW-06009 Warning Message

Fine Grain Dependency
Management
191

192
• Procedure A is a direct dependent of View B. View B is a direct
dependent of Table C. Procedure A is an indirect dependent of
Table C.
• Direct dependents are invalidated only by changes to the
referenced object that affect them.
• Indirect dependents can be invalidated by changes to the reference
object that do not affect them.
Invalidation of Dependent Objects
View B Table CProcedure A

193
• Before 11g, adding column D to table T invalidated the dependent
objects.
• Oracle Database 11g records additional, finer-grained dependency
management:
– Adding column D to table T does not impact view V and does not
invalidate the dependent objects
More Precise Dependency Metadata
in Oracle Database 11g
Procedure P Function FView V
Columns: A,B
Table T
Columns: A,B
Add column D

194
• In Oracle Database 11g, dependencies are now
tracked at the level of element within unit.
• Element-based dependency tracking covers the
following:
– Dependency of a single-table view on its base table
– Dependency of a PL/SQL program unit (package
specification, package body, or subprogram) on the
following:
• Other PL/SQL program units
• Tables
• Views
Fine-Grained Dependency Management

195
Fine-Grained Dependency Management:
Example 1
CREATE TABLE t2 (col_a NUMBER, col_b NUMBER, col_c NUMBER);
CREATE VIEW v AS SELECT col_a, col_b FROM t2;
ALTER TABLE t2 ADD (col_d VARCHAR2(20));
SELECT ud.name, ud.type, ud.referenced_name,
ud.referenced_type, uo.status
FROM user_dependencies ud, user_objects uo
WHERE ud.name = uo.object_name AND ud.name = 'V';
SELECT ud.name, ud.type, ud.referenced_name,
ud.referenced_type, uo.status

196
Example 1
ALTER TABLE t2 MODIFY (col_a VARCHAR2(20));
SELECT ud.name, ud.referenced_name, ud.referenced_type,
uo.status

197
Example 2
CREATE PACKAGE pkg IS
PROCEDURE proc_1;
END pkg;
/
CREATE OR REPLACE PROCEDURE p IS
BEGIN
pkg.proc_1();
END p;
/
CREATE OR REPLACE PACKAGE pkg
IS
PROCEDURE proc_1;
PROCEDURE unheard_of;
END pkg;
/

198
• To reduce invalidation of dependent objects:
Guidelines for Reducing Invalidation
Add new items to the
end of the package
Reference each table
through a view

199
• An object that is not valid when it is referenced
must be validated before it can be used.
• Validation occurs automatically when an object
is referenced; it does not require explicit user
action.
• If an object is not valid, its status is either
COMPILED WITH ERRORS,
UNAUTHORIZED, or INVALID.
Object Revalidation

201
• Definition:
– Inlining is defined as the replacement of a call to
subroutine with a copy of the body of the subroutine
that is called.
– The copied procedure generally runs faster than the
original.
– The PL/SQL compiler can automatically find the calls
that should be inlined.
• Benefits:
– Inlining can provide large performance gains when
applied judiciously by a factor of 2–10 times.
Intra Unit Inlining

202
• Influence implementing inlining via two methods:
– Oracle parameter PLSQL_OPTIMIZE_LEVEL
– PRAGMA INLINE
• Recommend that you:
– Inline small programs
– Inline programs that are frequently executed
• Use performance tools to identify hot spots
suitable for inline applications:
– plstimer
Use of Inlining

203
• Noninlined program:
Inlining Concepts
CREATE OR REPLACE PROCEDURE small_pgm
IS
a NUMBER;
b NUMBER;
PROCEDURE touch(x IN OUT NUMBER, y NUMBER)
IS
BEGIN
IF y > 0 THEN
x := x*x;
END IF;
END;
BEGIN
a := b;
FOR I IN 1..10 LOOP
touch(a, -17);
a := a*b;
END LOOP;
END small_pgm;

204
• Examine the loop after inlining:
Inlining Concepts
...
BEGIN
a := b;
FOR i IN 1..10 LOOP
IF –17 > 0 THEN
a := a*a;
END IF;
a := a*b;
END LOOP;
END small_pgm;
...

205
• The loop is transformed in several steps:
Inlining Concepts
a := b;
FOR i IN 1..10 LOOP ...
IF false THEN
a := a*a;
END IF;
a := a*b;
END LOOP;
a := b;
a := a*b;
END LOOP;
a := b;
a := a*b;
END LOOP;
a := b*b;
END LOOP;

206
• Set the PLSQL_OPTIMIZE_LEVEL session-level
parameter to a value of 2 or 3:
– Setting it to 2 means no automatic inlining is attempted.
– Setting it to 3 means automatic inlining is attempted and
no pragmas are necessary.
• Within a PL/SQL subroutine, use PRAGMAINLINE
– NO means no inlining occurs regardless of the level and
regardless of the YES pragmas.
– YES means inline at level 2 of a particular call and increase
the priority of inlining at level 3 for the call.
Inlining: Example
ALTER PROCEDURE small_pgm COMPILE
PLSQL_OPTIMIZE_LEVEL = 3 REUSE SETTINGS;

207
• After setting the PLSQL_OPTIMIZE_LEVEL
parameter, use a pragma:
Inlining: Example
CREATE OR REPLACE PROCEDURE small_pgm
IS
a PLS_INTEGER;
FUNCTION add_it(a PLS_INTEGER, b PLS_INTEGER)
RETURN PLS_INTEGER
IS
BEGIN
RETURN a + b;
END;
BEGIN
pragma INLINE (small_pgm, 'YES');
a := add_it(3, 4) + 6;
END small_pgm;

208
• Pragmas apply only to calls in the next statement
following the pragma.
• Programs that make use of smaller helper
subroutines are good candidates for inlining.
• Only local subroutines can be inlined.
• You cannot inline an external subroutine.
• Cursor functions should not be inlined.
• Inlining can increase the size of a unit.
• Be careful about suggesting to inline functions
that are deterministic.
Inlining: Guidelines

210210
• Handing PL/SQL in memory
• Global Temporary Tables (GTT)
• PL/SQL result cache
• Tips and Tricks
Tuning PL/SQL

211211
• Create packages that contain logically related
program units
• Reserve space for large allocations:
– Set the SHARED_POOL_RESERVED_SIZE
initialization parameter
• Prevent large or frequently used objects from
being aged out:
– Use the DBMS_SHARED_POOL package
Packages: Memory Issues
ORA-04031: unable to allocate 4160 bytes of shared memory..

212212
• Use dbms_shared_pool package:
• Flags:
– P – Package, Procedure or Function
– T – Type
– R – Trigger
– Q – Sequence
Pinning Objects
DBMS_SHARED_POOL.KEEP(object_name, flag)
DBMS_SHARED_POOL.UNKEEP(object_name, flag)

213213
• Memory is used more efficiently for scalability
(more users consume more memory)
• Package global memory is kept in the SGA
(instead of the UGA) and is reused for different
users
• Package global memory is only used within a unit
of work (a client-server call or a server to
different server call)
• Memory can be released and reused by another
user
Reusing Package Memory
Pragma SERIALLY_REUSABLE

214214
SERIALLY_REUSABLE - Example
CREATE OR REPLACE PACKAGE maintain_state
IS
pragma serially_reusable;
num1 number:= 0;
END maintain_state;
/
CREATE OR REPLACE PACKAGE regular_state
IS
num1 number:= 0;
END regular_state;
/

215215
BEGIN
dbms_output.put_line(chr(10) || 'THE MAINTAIN PACKAGE');
dbms_output.put_line('Original Value: ' || maintain_state.num1);
maintain_state.num1 := maintain_state.num1 + 10;
dbms_output.put_line('New Value: ' || maintain_state.num1 || chr(10));
dbms_output.put_line('THE REGULAR PACKAGE');
dbms_output.put_line('Original Value: ' || regular_state.num1);
regular_state.num1 := regular_state.num1 + 10;
dbms_output.put_line('New Value: ' || regular_state.num1 || chr(10));
dbms_output.put_line(chr(10)||chr(10));
dbms_output.put_line('THE MAINTAIN PACKAGE');
dbms_output.put_line('Original Value: ' || maintain_state.num1);
maintain_state.num1 := maintain_state.num1 + 10;
dbms_output.put_line('New Value: ' || maintain_state.num1 || chr(10));
dbms_output.put_line('THE REGULAR PACKAGE');
dbms_output.put_line('Original Value: ' || regular_state.num1);
regular_state.num1 := regular_state.num1 + 10;
dbms_output.put_line('New Value: ' || regular_state.num1 || chr(10));
END;
/

216
First Run Second Run
216
THE MAINTAIN PACKAGE
Original Value: 0
New Value: 10
THE REGULAR PACKAGE
Original Value: 0
New Value: 10
Original Value: 10
New Value: 20
THE REGULAR PACKAGE
Original Value: 10
New Value: 20
Original Value: 0
New Value: 10
THE REGULAR PACKAGE
Original Value: 20
New Value: 30
Original Value: 10
New Value: 20
THE REGULAR PACKAGE
Original Value: 30
New Value: 40

217217217217
• Since we’re giving up state managing, we can
now avoid ORA-4068 when compiling
• For more information, visit my blog:
http://www.realdbamagic.com/he/solving-ora-04068/
SERIALLY_REUSABLE – Side Effect

218
• The flexibility built into PL/SQL enables you to
pass:
– Simple scalar variables
– Complex data structures
• You can use the NOCOPY hint to improve
performance with the IN OUT parameters.
Passing Data Between PL/SQL Programs

219219
• The hint enables the PL/SQL compiler to pass
OUT and IN OUT parameters by reference, as
opposed to passing by value
• Enhances performance by reducing overhead
when passing parameters since less memory is
being used
• The Compiler will ignore the hint if it is not
possible to reference the original structure (type
conversion, constraints, for loop variable, etc.)
NOCOPY Hint

220220
NOCOPY - Example
CREATE OR REPLACE PACKAGE show_emp_pkg
IS
TYPE EmpTabTyp IS TABLE OF emp%ROWTYPE INDEX BY BINARY_INTEGER;
PROCEDURE show_emp (p_Deptno IN NUMBER,
p_EmpTab OUT NOCOPY EmpTabTyp);
END;
/

221
• Can be used in functions as an optimization
hint
• Indicates that a function can be used in a
parallelized query or parallelized DML
statement
Using the PARALLEL_ENABLE Hint
CREATE OR REPLACE FUNCTION f2 (p_p1 NUMBER)
RETURN NUMBER PARALLEL_ENABLE IS
BEGIN
RETURN p_p1 * 2;
END f2;

222222
• Large result sets can be stored in Table Variables or Temporary
Tables
• Temporary tables can be created to hold session-private data
that exists only for the duration of a transaction or session.
• Each session sees its own separate set of rows
• DML locks are not acquired on the data
• We can create indexes, views, and triggers on temporary tables
• Using temporary tables instead of program variables for large
• record sets, can reduce memory consumption
Global Temporary Tables
CREATE GLOBAL TEMPORARY TABLE hr.employees_temp
ON COMMIT PRSERVE ROWS;

224
– The result cache allows SQL query and PL/SQL function
results to be stored in cache memory.
– Subsequent executions of the same query or function can
be served directly out of the cache, improving response
times.
– This technique can be especially effective for SQL queries
and PL/SQL functions that are executed frequently.
– Cached query results become invalid when the database
data accessed by the query is modified.
What Is Result Caching?
Data dictionary
cache
Library
cache
SGA
Result
cache
Shared pool

225
• You can increase the small, default result cache
memory size by using the
RESULT_CACHE_MAX_SIZE initialization
parameter.
Increasing Result Cache Memory Size
SGA
Default
result
cache
Shared pool
Increased
result
cache

226
• Set Result_Cache_Max_Size from the command
line or in an initialization file created by a DBA.
• The cache size is dynamic and can be changed either
permanently or until the instance is restarted.
Setting Result_Cache_Max_Size
SQL> ALTER SYSTEM SET result_cache_max_size = 2M SCOPE =
MEMORY;
System altered.
SQL> SELECT name, value
2 FROM v$parameter
3 WHERE name = 'result_cache_max_size';
NAME VALUE
---------------------------------------- ------------------
result_cache_max_size 2097152
1 row selected.

227
• Use the RESULT_CACHE_MODE initialization
parameter in the database initialization
parameter file.
• RESULT_CACHE_MODE can be set to:
– MANUAL (default): You must add the RESULT_CACHE
hint to your queries for the results to be cached.
– FORCE: Results are always stored in the result
cache memory, if possible.
Enabling Query Result Cache

228
• Definition:
– Cache the results of the current query or query
fragment in memory, and then use the cached results
in future executions of the query or query fragments.
– Cached results reside in the result cache memory
portion of the SGA.
• Benefits:
– Improved performance
SQL Query Result Cache

229
• Scenario:
– You need to find the greatest average value of credit limit grouped by
state over the whole population.
– The query returns a large number of rows being analyzed to yield a
few or one row.
– In your query, the data changes fairly slowly (say every hour) but the
query is repeated fairly often (say every second).
• Solution:
– Use the new optimizer hint /*+ result_cache */ in your query:
SQL Query Result Cache
SELECT /*+ result_cache */
AVG(cust_credit_limit), cust_state_province
FROM sh.customers
GROUP BY cust_state_province;

230
Clearing the Shared Pool and Result Cache
--- flush.sql
--- Start with a clean slate. Flush the cache and shared
pool.
--- Verify that memory was released.
SET ECHO ON
SET FEEDBACK 1
SET SERVEROUTPUT ON
execute dbms_result_cache.flush
alter system flush shared_pool
/

231
– Definition:
• Enables data that is stored in cache to be shared across
sessions
• Stores the function result cache in an SGA, making it
available to any session that runs your application
– Benefits:
• Improved performance
• Improved scalability
PL/SQL Function Result Cache

232
• Scenario:
– You need a PL/SQL function that derives a complex metric.
– The data that your function calculates changes slowly, but
the function is frequently called.
• Solution:
– Use the new RESULT_CACHE clause in your function
definition.
– You can also have the cache purged when a dependent
table experiences a DML operation, by using the
RELIES_ON clause.
Marking PL/SQL Function Results to Be Cached

233
CREATE OR REPLACE FUNCTION ORD_COUNT(cust_no number)
RETURN NUMBER
RESULT_CACHE RELIES_ON (orders)
IS
V_COUNT NUMBER;
BEGIN
SELECT COUNT(*) INTO V_COUNT
FROM orders
WHERE customer_id = cust_no;
return v_count;
end;
• Include the RESULT_CACHE option in the function definition.
• Optionally, include the RELIES_ON clause.
Creating a PL/SQL Function Using the
RESULT_CACHE Clause
Specifies that the result
should be cached
Specifies the table upon which
the function relies
(not needed in 11.2+)

234
• Specify DETERMINISTIC to indicate that the
function returns the same result value whenever
it is called with the same values for its arguments.
• This helps the optimizer avoid redundant function
calls.
• If a function was called previously with the same
arguments, the optimizer can elect to use the
previous result.
• Do not specify DETERMINISTIC for a function
whose result depends on the state of session
variables or schema objects.
Using the DETERMINISTIC
Clause with Functions

235
Calling the PL/SQL Function Inside a Query
select cust_last_name, ord_count(customer_id) no_of_orders
from customers
where cust_last_name = 'MacGraw'

236
Verifying Memory Allocation
--- Establish the cache content
set serveroutput on
execute dbms_result_cache.memory_report

237
Viewing Cache Results Created
col name format a55
select * from v$result_cache_statistics
/

238
Calling the PL/SQL Function Again
select cust_last_name, ord_count(customer_id) no_of_orders
from customers
where cust_last_name = 'MacGraw'

239
Viewing Cache Results Found
col name format a55
select * from v$result_cache_statistics
/

240
Confirming That the Cached Result Was Used
select type, namespace,status, scan_count,name
from v$result_cache_objects
/

241241
• Beware of result caching of timed actions:
DBMS_LOCK.SLEEP will also be cached
overriding the sleep
• Cannot be used with invoker's rights or in an
anonymous block
• Cannot be used with pipelined table function
• Cannot be used with OUT or IN OUT
parameters.
PL/SQL Result Cache Pitfall

243
You can tune your PL/SQL code by:
– Identifying the data type and constraint issues
• Data type conversion
• The NOT NULL constraint
• PLS_INTEGER
• SIMPLE_INTEGER
– Writing smaller executable sections of code
– Comparing SQL with PL/SQL
– Rephrasing conditional statements
Tuning PL/SQL Code

244
DECLARE
n NUMBER;
BEGIN
n := n + 15; -- converted
n := n + 15.0; -- not converted
...
END;
– PL/SQL performs implicit conversions between
structurally different data types.
– Example: When assigning a PLS_INTEGER
variable to a NUMBER variable
Avoiding Implicit Data Type Conversion
strings
dates
numbers

245
Understanding the NOT NULL Constraint
PROCEDURE calc_m IS
m NUMBER; --no constraint
...
BEGIN
m := a + b;
IF m IS NULL THEN
-- raise error
END IF;
END;
PROCEDURE calc_m IS
m NUMBER NOT NULL:=0;
a NUMBER;
b NUMBER;
BEGIN
m := a + b;
END;
The value of the expression a + b is
assigned to a temporary variable,
which is then tested for nullity.
A better way to check nullity; no
performance overhead

246
Use PLS_INTEGER when dealing with integer
data.
– It is an efficient data type for integer variables.
– It requires less storage than INTEGER or
NUMBER.
– Its operations use machine arithmetic, which is
faster than library arithmetic.
Using the PLS_INTEGER Data Type for Integers

247
– Definition:
• Is a predefined subtype
• Has the range –2147483648 .. 2147483648
• Does not include a null value
• Is allowed anywhere in PL/SQL where the PLS_INTEGER
data type is allowed
– Benefits:
• Eliminates the overhead of overflow
checking
• Is estimated to be 2–10 times faster
when compared with the PLS_INTEGER
type with native PL/SQL compilation
Using the SIMPLE_INTEGER Data Type

248
Each has its own benefits:
• SQL:
– Accesses data in the database
– Treats data as sets
• PL/SQL:
– Provides procedural capabilities
– Has more flexibility built into the language
Comparing SQL with PL/SQL

249
• Some simple set processing is markedly faster than the
equivalent PL/SQL.
• Avoid using procedural code when it may be better to use
SQL.
...FOR I IN 1..5600 LOOP
counter := counter + 1;
SELECT product_id, warehouse_id
INTO v_p_id, v_wh_id
FROM big_inventories WHERE v_p_id = counter;
INSERT INTO inventories2 VALUES(v_p_id, v_wh_id);
END LOOP;...
Comparing SQL with PL/SQL
BEGIN
INSERT INTO inventories2
SELECT product_id, warehouse_id
FROM main_inventories;
END;

250
If your business logic results in one condition being
true, use the ELSIF syntax for mutually exclusive
clauses:
Rephrasing Conditional
Control Statements
IF v_acct_mgr = 145 THEN
process_acct_145;
END IF;
process_acct_147;
END IF;
process_acct_148;
END IF;
process_acct_149;
END IF;
IF v_acct_mgr = 145
THEN
process_acct_145;
ELSIF v_acct_mgr = 147
THEN
process_acct_147;
THEN
process_acct_148;
THEN
process_acct_149;
END IF;

SQLcl Introduction
The Next Generation of SQL*Plus?
251

252252
• Introduced in Oracle 5 (1985)
• Looks very simple but has tight integration with
other Oracle infrastructure and tools
• Very good for reporting, scripting, and
automation
• Replaced old CLI tool called …
UFI (“User Friendly Interface”)
SQL*Plus

253253
• Nothing really wrong with SQL*Plus – it is being
updated constantly but it is missing a lot of
functionality
• SQL*Plus forces us to use GUI tools to complete
some basic tasks
• Easy to understand, a bit hard to use
• Not easy for new users or developers
What’s Wrong With SQL*Plus?

254254
• SQLcl is a new command line interface (CLI) for SQL
users and DBAs
• It is part of the SQL Developer suite – developed by the
same team: Oracle Database Development Tools Team
• Can do most of what SQL*Plus does and much more
• Minimal installation, minimal requirements
Introducing SQLcl

255255
• It’s still in the early adopter version (current
version: 4.2.0.15.295.1605 RC, October 23,
2015)
• Uses Java, one version for Windows, Linux and
OS X
• Planned to be shipped out with Oracle
Database 12cR2 (“within the next 12 month”)
Introducing SQLcl (cont.)

256256
• Early Adopter version/RC
• QA still logging bugs from SQL*Plus regression tests
• Adding support for existing SQL*Plus commands and
syntax
• Adding new commands
• But can it do...?
– Yes
– Not yet
– No
Current Status

257257
• Download from: SQL Developer Download
page
• Unzip the file
• Run it
Installing

259259
• Use the tab key to complete commands
• Can be used to list tables, views or other queriable
objects
• Can be used to replace the * with actual column names
• Use the arrow keys to move around the command
• Use CTRL+W and CTRL+S to jump to the beginning/end
of commands
Object Completion and Easy Edit

260260
• 100 command history buffer
• Commands are persistent between sessions (watch out for
security!)
• Use UP and DOWN arrow keys to access old commands
• Usage:
history
history usage
History script
history full
History clear [session?]
• Load from history into command buffer:
history <number>
Command History

261261
• Describe lists the column of the tables just like SQL*Plus
• Information shows column names, default values, indexes
and constraints.
• In 12c database information shows table statistics and In
memory status
• Works for table, views, sequences, and code objects
• Info+ shows additional information regarding column
statistics and column histograms
Describe, Information and info+

262262
• Outputting query results becomes easier with the
“set sqlformat” command (also available in SQL
Developer)
• We can create a query in the “regular” way and
then switch between the different output styles:
– ANSIConsole
– Fixed column size output
– XML or JSON output
– HTML output generates a built in search field and a responsive html output for
the result only
Generating Pretty Output

263263
• We can generate loader ready output (with “|” as a
delimiter)
• We can generate insert commands
• We can easily generate CSV output
• Usage:
set sqlformat {
csv,html,xml,json,ansiconsole,insert,
loader,fixed,default}
Generating Other Useful Outputs

264264
• Loads a comma separated value (csv) file into a
table
• The first row of the file must be a header row and
the file must be encoded UTF8
• The load is processed with 50 rows per batch
• Usage:
LOAD [schema.]table_name[@db_link] file_name
Load Data From CSV File

265265
• We talked about Developing PL/SQL: composite
datatypes, cursors, dynamic SQL
• We went over Oracle 11g and Oracle 12c
important features
• We looked at the compilation process of PL/SQL
and how to make it more efficient
• We talked about tuning our PL/SQL code and
some dedicated features for that
• SQLcl – New replacement tool for SQL*Plus
Summary

266266
• There are a lot of changes in Oracle 12c. We
only looked on a few developer oriented
features
• New XML and JSON handling functions
• Edition Based Redefinition
• Pipeline Table Functions and Parallel executions
• SQL Performance and new Optimizer changes
• Exceptions and Error handling
What Did We Not Talk About?

Q&AAny Questions? Now will be the time!
267

268268
Zohar Elkayam
twitter: @realmgic
Zohar@Brillix.co.il
www.ilDBA.co.il
www.realdbamagic.com

Advanced PLSQL Optimizing for Better Performance

Recommended

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Advanced PLSQL Optimizing for Better Performance (20)

More from Zohar Elkayam (20)

Recently uploaded (20)

Advanced PLSQL Optimizing for Better Performance