Infolinks

Friday 29 June 2012

Fundamentals of PL/SQL

Fundamentals of PL/SQL

There are six essentials in painting. The first is called spirit; the second, rhythm; the third, thought; the fourth, scenery; the fifth, the brush; and the last is the ink. --Ching Hao
The previous chapter provided an overview of PL/SQL. This chapter focuses on the small-scale aspects of the language. Like every other programming language, PL/SQL has a character set, reserved words, punctuation, datatypes, rigid syntax, and fixed rules of usage and statement formation. You use these basic elements of PL/SQL to represent real-world objects and operations.
This chapter discusses the following topics:
Character Set
Lexical Units
Declarations
PL/SQL Naming Conventions
Scope and Visibility of PL/SQL Identifiers
Variable Assignment
PL/SQL Expressions and Comparisons
Built-In Functions

Character Set

You write a PL/SQL program as lines of text using a specific set of characters. The PL/SQL character set includes
the upper- and lower-case letters A .. Z and a .. z
the numerals 0 .. 9
the symbols ( ) + - * / < > = ! ~ ^ ; : . ' @ % , " # $ & _ | { } ? [ ]
tabs, spaces, and carriage returns
PL/SQL is not case sensitive, so lower-case letters are equivalent to corresponding upper-case letters except within string and character literals.

Lexical Units

A line of PL/SQL text contains groups of characters known as lexical units, which can be classified as follows:
delimiters (simple and compound symbols)
identifiers, which include reserved words
literals
comments
To improve readability, you can separate lexical units by spaces. In fact, you must separate adjacent identifiers by a space or punctuation. The following line is not allowed because the reserved words END and IF are joined:
IF x > y THEN high := x; ENDIF;  -- not allowed

However, you cannot embed spaces in lexical units except for string literals and comments. For example, the following line is not allowed because the compound symbol for assignment (:=) is split:
count : = count + 1;  -- not allowed

To show structure, you can divide lines using carriage returns and indent lines using spaces or tabs. Compare these IF statements for readability:
IF x>y THEN max:=x;ELSE max:=y;END IF;     |     IF x > y THEN
                                           |        max := x;
                                           |     ELSE
                                           |        max := y;
                                           |     END IF;

Delimiters

A delimiter is a simple or compound symbol that has a special meaning to PL/SQL. For example, you use delimiters to represent arithmetic operations such as addition and subtraction. Simple symbols consist of one character. A list follows:
Symbol Meaning
+
addition operator
%
attribute indicator
'
character string delimiter
.
component selector
/
division operator
(
expression or list delimiter
)
expression or list delimiter
:
host variable indicator
,
item separator
*
multiplication operator
"
quoted identifier delimiter
=
relational operator
<
relational operator
>
relational operator
@
remote access indicator
;
statement terminator
-
subtraction/negation operator
Compound symbols consist of two characters. A list follows:
Symbol Meaning
:=
assignment operator
=>
association operator
||
concatenation operator
**
exponentiation operator
<<
label delimiter (begin)
>>
label delimiter (end)
/*
multi-line comment delimiter (begin)
*/
multi-line comment delimiter (end)
..
range operator
<>
relational operator
!=
relational operator
~=
relational operator
^=
relational operator
<=
relational operator
>=
relational operator
- -
single-line comment indicator

Identifiers

You use identifiers to name PL/SQL program items and units, which include constants, variables, exceptions, cursors, cursor variables, subprograms, and packages. Some examples of identifiers follow:
X
t2
phone#
credit_limit
LastName
oracle$number

An identifier consists of a letter optionally followed by more letters, numerals, dollar signs, underscores, and number signs. Other characters such as hyphens, slashes, and spaces are not allowed, as the following examples show:
mine&yours    -- not allowed because of ampersand
debit-amount  -- not allowed because of hyphen
on/off        -- not allowed because of slash
user id       -- not allowed because of space

The next examples show that adjoining and trailing dollar signs, underscores, and number signs are allowed:
money$$$tree
SN## 
try_again_

You can use upper, lower, or mixed case to write identifiers. PL/SQL is not case sensitive except within string and character literals. So, if the only difference between identifiers is the case of corresponding letters, PL/SQL considers the identifiers to be the same, as the following example shows:
lastname
LastName  -- same as lastname
LASTNAME  -- same as lastname and LastName

The size of an identifier cannot exceed 30 characters. But, every character, including dollar signs, underscores, and number signs, is significant. For example, PL/SQL considers the following identifiers to be different:
lastname
last_name

Identifiers should be descriptive. So, avoid obscure names such as cpm. Instead, use meaningful names such as cost_per_thousand.

Reserved Words

Some identifiers, called reserved words, have a special syntactic meaning to PL/SQL and so should not be redefined. For example, the words BEGIN and END, which bracket the executable part of a block or subprogram, are reserved. As the next example shows, if you try to redefine a reserved word, you get a compilation error:
DECLARE
   end BOOLEAN;  -- not allowed; causes compilation error

However, you can embed reserved words in an identifier, as the following example shows:
DECLARE
   end_of_game BOOLEAN;  -- allowed

Often, reserved words are written in upper case to promote readability. However, like other PL/SQL identifiers, reserved words can be written in lower or mixed case. For a list of reserved words, see Appendix F.

Predefined Identifiers

Identifiers globally declared in package STANDARD, such as the exception INVALID_NUMBER, can be redeclared. However, redeclaring predefined identifiers is error prone because your local declaration overrides the global declaration.

Quoted Identifiers

For flexibility, PL/SQL lets you enclose identifiers within double quotes. Quoted identifiers are seldom needed, but occasionally they can be useful. They can contain any sequence of printable characters including spaces but excluding double quotes. Thus, the following identifiers are valid:
"X+Y"
"last name"
"on/off switch"
"employee(s)"
"*** header info ***"

The maximum size of a quoted identifier is 30 characters not counting the double quotes. Though allowed, using PL/SQL reserved words as quoted identifiers is a poor programming practice.
Some PL/SQL reserved words are not reserved by SQL. For example, you can use the PL/SQL reserved word TYPE in a CREATE TABLE statement to name a database column. But, if a SQL statement in your program refers to that column, you get a compilation error, as the following example shows:
SELECT acct, type, bal INTO ...  -- causes compilation error

To prevent the error, enclose the uppercase column name in double quotes, as follows:
SELECT acct, "TYPE", bal INTO ...

The column name cannot appear in lower or mixed case (unless it was defined that way in the CREATE TABLE statement). For example, the following statement is invalid:
SELECT acct, "type", bal INTO ...  -- causes compilation error

Alternatively, you can create a view that renames the troublesome column, then use the view instead of the base table in SQL statements.

Literals

A literal is an explicit numeric, character, string, or Boolean value not represented by an identifier. The numeric literal 147 and the Boolean literal FALSE are examples.

Numeric Literals

Two kinds of numeric literals can be used in arithmetic expressions: integers and reals. An integer literal is an optionally signed whole number without a decimal point. Some examples follow:
030   6   -14   0   +32767

A real literal is an optionally signed whole or fractional number with a decimal point. Several examples follow:
6.6667   0.0   -12.0   3.14159   +8300.00   .5   25.

PL/SQL considers numbers such as 12.0 and 25. to be reals even though they have integral values.
Numeric literals cannot contain dollar signs or commas, but can be written using scientific notation. Simply suffix the number with an E (or e) followed by an optionally signed integer. A few examples follow:
2E5   1.0E-7   3.14159e0   -1E38   -9.5e-3

E stands for "times ten to the power of." As the next example shows, the number after E is the power of ten by which the number before E must be multiplied (the double asterisk (**) is the exponentiation operator):
5E3 = 5 * 10**3 = 5 * 1000 = 5000

The number after E also corresponds to the number of places the decimal point shifts. In the last example, the implicit decimal point shifted three places to the right. In this example, it shifts three places to the left:
5E-3 = 5 * 10**-3 = 5 * 0.001 = 0.005

As the following example shows, if the value of a numeric literal falls outside the range 1E-130 .. 10E125, you get a compilation error:
DECLARE
   n NUMBER;
BEGIN
   n := 10E127;  -- causes a 'numeric overflow or underflow' error

Character Literals

A character literal is an individual character enclosed by single quotes (apostrophes). Character literals include all the printable characters in the PL/SQL character set: letters, numerals, spaces, and special symbols. Some examples follow:
'Z'   '%'   '7'   ' '   'z'   '('

PL/SQL is case sensitive within character literals. For example, PL/SQL considers the literals 'Z' and 'z' to be different. Also, the character literals '0'..'9' are not equivalent to integer literals but can be used in arithmetic expressions because they are implicitly convertible to integers.

String Literals

A character value can be represented by an identifier or explicitly written as a string literal, which is a sequence of zero or more characters enclosed by single quotes. Several examples follow:
'Hello, world!'
'XYZ Corporation'
'10-NOV-91'
'He said "Life is like licking honey from a thorn."'
'$1,000,000'

All string literals except the null string ('') have datatype CHAR.
Given that apostrophes (single quotes) delimit string literals, how do you represent an apostrophe within a string? As the next example shows, you write two single quotes, which is not the same as writing a double quote:
'Don''t leave without saving your work.'

PL/SQL is case sensitive within string literals. For example, PL/SQL considers the following literals to be different:
'baker'
'Baker'

Boolean Literals

Boolean literals are the predefined values TRUE, FALSE, and NULL (which stands for a missing, unknown, or inapplicable value). Remember, Boolean literals are values, not strings. For example, TRUE is no less a value than the number 25.

Datetime Literals

Datetime literals have various formats depending on the datatype. For example:
DECLARE
d1 DATE := DATE '1998-12-25';
t1 TIMESTAMP := TIMESTAMP '1997-10-22 13:01:01';
t2 TIMESTAMP WITH TIME ZONE := TIMESTAMP '1997-01-31 09:26:56.66 
+02:00';
-- Three years and two months
-- (For greater precision, we would use the day-to-second interval)
i1 INTERVAL YEAR TO MONTH := INTERVAL '3-2' YEAR TO MONTH;
-- Five days, four hours, three minutes, two and 1/100 seconds
i2 INTERVAL DAY TO SECOND := INTERVAL '5 04:03:02.01' DAY TO SECOND;
...

You can also specify whether a given interval value is YEAR TO MONTH or DAY TO SECOND. For example, current_timestamp - current_timestamp produces a value of type INTERVAL DAY TO SECOND by default. You can specify the type of the interval using the formats:
  • (interval_expression) DAY TO SECOND
  • (interval_expression) YEAR TO MONTH
For details on the syntax for the date and time types, see the Oracle9i SQL Reference. For examples of performing date/time arithmetic, see Oracle9i Application Developer's Guide - Fundamentals.

Comments

The PL/SQL compiler ignores comments, but you should not. Adding comments to your program promotes readability and aids understanding. Generally, you use comments to describe the purpose and use of each code segment. PL/SQL supports two comment styles: single-line and multi-line.

Single-Line Comments

Single-line comments begin with a double hyphen (--) anywhere on a line and extend to the end of the line. A few examples follow:
-- begin processing
SELECT sal INTO salary FROM emp  -- get current salary
   WHERE empno = emp_id;
bonus := salary * 0.15;  -- compute bonus amount

Notice that comments can appear within a statement at the end of a line.
While testing or debugging a program, you might want to disable a line of code. The following example shows how you can "comment-out" the line:
-- DELETE FROM emp WHERE comm IS NULL;

Multi-line Comments

Multi-line comments begin with a slash-asterisk (/*), end with an asterisk-slash (*/), and can span multiple lines. Some examples follow:
BEGIN
   ...
   /* Compute a 15% bonus for top-rated employees. */
   IF rating > 90 THEN
      bonus := salary * 0.15 /* bonus is based on salary */
   ELSE
      bonus := 0;
   END IF;
   ...
   /* The following line computes the area of a 
      circle using pi, which is the ratio between 
      the circumference and diameter. */
   area := pi * radius**2;
END;

You can use multi-line comment delimiters to comment-out whole sections of code, as the following example shows:
/* 
LOOP
   FETCH c1 INTO emp_rec;
   EXIT WHEN c1%NOTFOUND;
   ...
END LOOP;
*/

Restrictions on Comments

You cannot nest comments. Also, you cannot use single-line comments in a PL/SQL block that will be processed dynamically by an Oracle Precompiler program because end-of-line characters are ignored. As a result, single-line comments extend to the end of the block, not just to the end of a line. So, use multi-line comments instead.

Declarations

Your program stores values in variables and constants. As the program executes, the values of variables can change, but the values of constants cannot.
You can declare variables and constants in the declarative part of any PL/SQL block, subprogram, or package. Declarations allocate storage space for a value, specify its datatype, and name the storage location so that you can reference it.
A couple of examples follow:
birthday  DATE;
emp_count SMALLINT := 0;

The first declaration names a variable of type DATE. The second declaration names a variable of type SMALLINT and uses the assignment operator to assign an initial value of zero to the variable.
The next examples show that the expression following the assignment operator can be arbitrarily complex and can refer to previously initialized variables:
pi     REAL := 3.14159;
radius REAL := 1;
area   REAL := pi * radius**2;

By default, variables are initialized to NULL. So, these declarations are equivalent:
birthday DATE;
birthday DATE := NULL;

In the declaration of a constant, the keyword CONSTANT must precede the type specifier, as the following example shows:
credit_limit CONSTANT REAL := 5000.00;

This declaration names a constant of type REAL and assigns an initial (also final) value of 5000 to the constant. A constant must be initialized in its declaration. Otherwise, you get a compilation error when the declaration is elaborated. (The processing of a declaration by the PL/SQL compiler is called elaboration.)

Using DEFAULT

You can use the keyword DEFAULT instead of the assignment operator to initialize variables. For example, the declaration
blood_type CHAR := 'O';

can be rewritten as follows:
blood_type CHAR DEFAULT 'O';

Use DEFAULT for variables that have a typical value. Use the assignment operator for variables (such as counters and accumulators) that have no typical value. A couple of examples follow:
hours_worked   INTEGER DEFAULT 40;
employee_count INTEGER := 0;

You can also use DEFAULT to initialize subprogram parameters, cursor parameters, and fields in a user-defined record.

Using NOT NULL

Besides assigning an initial value, declarations can impose the NOT NULL constraint, as the following example shows:
acct_id INTEGER(4) NOT NULL := 9999;

You cannot assign nulls to a variable defined as NOT NULL. If you try, PL/SQL raises the predefined exception VALUE_ERROR. The NOT NULL constraint must be followed by an initialization clause. For example, the following declaration is not allowed:
acct_id INTEGER(5) NOT NULL;  -- not allowed; not initialized

PL/SQL provide subtypes NATURALN and POSITIVEN that are predefined as NOT NULL. For instance, the following declarations are equivalent:
emp_count NATURAL NOT NULL := 0;
emp_count NATURALN := 0;

In NATURALN and POSITIVEN declarations, the type specifier must be followed by an initialization clause. Otherwise, you get a compilation error. For example, the following declaration is not allowed:
line_items POSITIVEN;  -- not allowed; not initialized

Using %TYPE

The %TYPE attribute provides the datatype of a variable or database column. In the following example, %TYPE provides the datatype of a variable:
credit REAL(7,2);
debit  credit%TYPE;

Variables declared using %TYPE are treated like those declared using a datatype specifier. For example, given the previous declarations, PL/SQL treats debit like a REAL(7,2) variable. The next example shows that a %TYPE declaration can include an initialization clause:
balance         NUMBER(7,2);
minimum_balance balance%TYPE := 10.00;

The %TYPE attribute is particularly useful when declaring variables that refer to database columns. You can reference a table and column, or you can reference an owner, table, and column, as in
my_dname scott.dept.dname%TYPE;

Using %TYPE to declare my_dname has two advantages. First, you need not know the exact datatype of dname. Second, if the database definition of dname changes, the datatype of my_dname changes accordingly at run time.
However, %TYPE variables do not inherit the NOT NULL column constraint. In the next example, even though the database column empno is defined as NOT NULL, you can assign a null to the variable my_empno:
DECLARE
   my_empno emp.empno%TYPE;
   ...
BEGIN
   my_empno := NULL;  -- this works

Using %ROWTYPE

The %ROWTYPE attribute provides a record type that represents a row in a table (or view). The record can store an entire row of data selected from the table or fetched from a cursor or strongly typed cursor variable. In the example below, you declare two records. The first record stores a row selected from the emp table. The second record stores a row fetched from cursor c1.
DECLARE
   emp_rec emp%ROWTYPE;
   CURSOR c1 IS SELECT deptno, dname, loc FROM dept;
   dept_rec c1%ROWTYPE;

Columns in a row and corresponding fields in a record have the same names and datatypes. However, fields in a %ROWTYPE record do not inherit the NOT NULL column constraint.
In the following example, you select column values into record emp_rec:
BEGIN
   SELECT * INTO emp_rec FROM emp WHERE ...

The column values returned by the SELECT statement are stored in fields. To reference a field, you use dot notation. For example, you might reference the deptno field as follows:
IF emp_rec.deptno = 20 THEN ...

Also, you can assign the value of an expression to a specific field, as the following examples show:
emp_rec.ename := 'JOHNSON';
emp_rec.sal := emp_rec.sal * 1.15;

In the final example, you use %ROWTYPE to define a packaged cursor:
CREATE PACKAGE emp_actions AS
   CURSOR c1 RETURN emp%ROWTYPE;  -- declare cursor specification
   ...
END emp_actions;

CREATE PACKAGE BODY emp_actions AS
   CURSOR c1 RETURN emp%ROWTYPE IS  -- define cursor body
      SELECT * FROM emp WHERE sal > 3000;
   ...
END emp_actions;

Aggregate Assignment

A %ROWTYPE declaration cannot include an initialization clause. However, there are two ways to assign values to all fields in a record at once. First, PL/SQL allows aggregate assignment between entire records if their declarations refer to the same table or cursor. For example, the following assignment is allowed:
DECLARE
   dept_rec1 dept%ROWTYPE;
   dept_rec2 dept%ROWTYPE;
   CURSOR c1 IS SELECT deptno, dname, loc FROM dept;
   dept_rec3 c1%ROWTYPE;
BEGIN
   ...
   dept_rec1 := dept_rec2;

However, because dept_rec2 is based on a table and dept_rec3 is based on a cursor, the following assignment is not allowed:
dept_rec2 := dept_rec3;  -- not allowed

Second, you can assign a list of column values to a record by using the SELECT or FETCH statement, as the example below shows. The column names must appear in the order in which they were defined by the CREATE TABLE or CREATE VIEW statement.
DECLARE
   dept_rec dept%ROWTYPE;
   ...
BEGIN
   SELECT * INTO dept_rec FROM dept WHERE deptno = 30;
   ...
END;

However, you cannot assign a list of column values to a record by using an assignment statement. So, the following syntax is not allowed:
record_name := (value1, value2, value3, ...);  -- not allowed

Using Aliases

Select-list items fetched from a cursor associated with %ROWTYPE must have simple names or, if they are expressions, must have aliases. In the following example, you use an alias called wages:
-- available online in file 'examp4'
DECLARE
   CURSOR my_cursor IS 
      SELECT sal + NVL(comm, 0) wages, ename FROM emp;
   my_rec  my_cursor%ROWTYPE;
BEGIN
   OPEN my_cursor;
   LOOP
      FETCH my_cursor INTO my_rec;
      EXIT WHEN my_cursor%NOTFOUND;
      IF my_rec.wages > 2000 THEN
         INSERT INTO temp VALUES (NULL, my_rec.wages, my_rec.ename);
      END IF;
   END LOOP;
   CLOSE my_cursor;
END;

Restrictions on Declarations

PL/SQL does not allow forward references. You must declare a variable or constant before referencing it in other statements, including other declarative statements. For example, the following declaration of maxi is not allowed:
maxi INTEGER := 2 * mini;  -- not allowed
mini INTEGER := 15;

However, PL/SQL does allow the forward declaration of subprograms. For more information, see "Declaring PL/SQL Subprograms".
Some languages allow you to declare a list of variables that have the same datatype. PL/SQL does not allow this. For example, the following declaration is not allowed:
i, j, k SMALLINT;  -- not allowed

You must declare each variable separately:
i SMALLINT;
j SMALLINT;
k SMALLINT;

PL/SQL Naming Conventions

The same naming conventions apply to all PL/SQL program items and units including constants, variables, cursors, cursor variables, exceptions, procedures, functions, and packages. Names can be simple, qualified, remote, or both qualified and remote. For example, you might use the procedure name raise_salary in any of the following ways:
raise_salary(...);                      -- simple
emp_actions.raise_salary(...);          -- qualified
raise_salary@newyork(...);              -- remote
emp_actions.raise_salary@newyork(...);  -- qualified and remote

In the first case, you simply use the procedure name. In the second case, you must qualify the name using dot notation because the procedure is stored in a package called emp_actions. In the third case, using the remote access indicator (@), you reference the database link newyork because the procedure is stored in a remote database. In the fourth case, you qualify the procedure name and reference a database link.

Synonyms

You can create synonyms to provide location transparency for remote schema objects such as tables, sequences, views, standalone subprograms, packages, and object types. However, you cannot create synonyms for items declared within subprograms or packages. That includes constants, variables, cursors, cursor variables, exceptions, and packaged subprograms.

Scoping

Within the same scope, all declared identifiers must be unique. So, even if their datatypes differ, variables and parameters cannot share the same name. In the following example, the second declaration is not allowed:
DECLARE
   valid_id BOOLEAN;
   valid_id VARCHAR2(5);  -- not allowed duplicate identifier

For the scoping rules that apply to identifiers, see "Scope and Visibility of PL/SQL Identifiers".

Case Sensitivity

Like all identifiers, the names of constants, variables, and parameters are not case sensitive. For instance, PL/SQL considers the following names to be the same:
DECLARE
   zip_code INTEGER;
   Zip_Code INTEGER;  -- same as zip_code

Name Resolution

In potentially ambiguous SQL statements, the names of database columns take precedence over the names of local variables and formal parameters. For example, the following DELETE statement removes all employees from the emp table, not just 'KING', because Oracle assumes that both enames in the WHERE clause refer to the database column:
DECLARE
   ename VARCHAR2(10) := 'KING';
BEGIN
   DELETE FROM emp WHERE ename = ename;
   ...

In such cases, to avoid ambiguity, prefix the names of local variables and formal parameters with my_, as follows:
DECLARE
   my_ename VARCHAR2(10);

Or, use a block label to qualify references, as in
<<main>>
DECLARE
   ename VARCHAR2(10) := 'KING';
BEGIN
   DELETE FROM emp WHERE ename = main.ename;
   ...

The next example shows that you can use a subprogram name to qualify references to local variables and formal parameters:
FUNCTION bonus (deptno IN NUMBER, ...) RETURN REAL IS
   job CHAR(10);
BEGIN
   SELECT ... WHERE deptno = bonus.deptno AND job = bonus.job;
   ...

For a full discussion of name resolution, see Appendix D.

Scope and Visibility of PL/SQL Identifiers

References to an identifier are resolved according to its scope and visibility. The scope of an identifier is that region of a program unit (block, subprogram, or package) from which you can reference the identifier. An identifier is visible only in the regions from which you can reference the identifier using an unqualified name. Figure 2-1 shows the scope and visibility of a variable named x, which is declared in an enclosing block, then redeclared in a sub-block.
Identifiers declared in a PL/SQL block are considered local to that block and global to all its sub-blocks. If a global identifier is redeclared in a sub-block, both identifiers remain in scope. Within the sub-block, however, only the local identifier is visible because you must use a qualified name to reference the global identifier.
Although you cannot declare an identifier twice in the same block, you can declare the same identifier in two different blocks. The two items represented by the identifier are distinct, and any change in one does not affect the other. However, a block cannot reference identifiers declared in other blocks at the same level because those identifiers are neither local nor global to the block.

Figure 2-1 Scope and Visibility

Text description of pls81007_scope_and_visibility.gif follows
Text description of the illustration pls81007_scope_and_visibility.gif


The example below illustrates the scope rules. Notice that the identifiers declared in one sub-block cannot be referenced in the other sub-block. That is because a block cannot reference identifiers declared in other blocks nested at the same level.
DECLARE
   a CHAR;
   b REAL;
BEGIN
   -- identifiers available here: a (CHAR), b
   DECLARE
      a INTEGER;
      c REAL;
   BEGIN
      -- identifiers available here: a (INTEGER), b, c
   END;
   DECLARE
      d REAL;
   BEGIN
      -- identifiers available here: a (CHAR), b, d
   END;
   -- identifiers available here: a (CHAR), b
END;

Recall that global identifiers can be redeclared in a sub-block, in which case the local declaration prevails and the sub-block cannot reference the global identifier unless you use a qualified name. The qualifier can be the label of an enclosing block, as the following example shows:
<<outer>>
DECLARE
   birthdate DATE;
BEGIN
   DECLARE
      birthdate DATE;
   BEGIN
      ...
      IF birthdate = outer.birthdate THEN ...
   END;
   ...
END;

As the next example shows, the qualifier can also be the name of an enclosing subprogram:
PROCEDURE check_credit (...) IS
   rating NUMBER;
   FUNCTION valid (...) RETURN BOOLEAN IS
      rating NUMBER;
   BEGIN
      ...
      IF check_credit.rating < 3 THEN ...
   END;
BEGIN
   ...
END;

However, within the same scope, a label and a subprogram cannot have the same name.

Variable Assignment

Variables and constants are initialized every time a block or subprogram is entered. By default, variables are initialized to NULL. Unless you expressly initialize a variable, its value is undefined:
DECLARE
   count INTEGER;
BEGIN
-- COUNT began with a value of NULL.
-- Thus the expression 'COUNT + 1' is also null.
-- So after this assignment, COUNT is still NULL.
   count := count + 1;

To avoid unexpected results, never reference a variable before you assign it a value.
You can use assignment statements to assign values to a variable. For example, the following statement assigns a new value to the variable bonus, overwriting its old value:
bonus := salary * 0.15;

The expression following the assignment operator can be arbitrarily complex, but it must yield a datatype that is the same as or convertible to the datatype of the variable.

Assigning Boolean Values

Only the values TRUE, FALSE, and NULL can be assigned to a Boolean variable. For example, given the declaration
DECLARE
   done BOOLEAN;

the following statements are allowed:
BEGIN
   done := FALSE;
   WHILE NOT done LOOP 
      ...
   END LOOP;

When applied to an expression, the relational operators return a Boolean value. So, the following assignment is allowed:
done := (count > 500);

Assigning a SQL Query Result to a PL/SQL Variable

You can use the SELECT statement to have Oracle assign values to a variable. For each item in the select list, there must be a corresponding, type-compatible variable in the INTO list. An example follows:
DECLARE
   emp_id   emp.empno%TYPE;
   emp_name emp.ename%TYPE;
   wages    NUMBER(7,2);
BEGIN
   -- assign a value to emp_id here
   SELECT ename, sal + comm 
      INTO emp_name, wages FROM emp
      WHERE empno = emp_id;
   ...
END;

However, you cannot select column values into a Boolean variable.

PL/SQL Expressions and Comparisons

Expressions are constructed using operands and operators. An operand is a variable, constant, literal, or function call that contributes a value to an expression. An example of a simple arithmetic expression follows:
-X / 2 + 3

Unary operators such as the negation operator (-) operate on one operand; binary operators such as the division operator (/) operate on two operands. PL/SQL has no ternary operators.
The simplest expressions consist of a single variable, which yields a value directly. PL/SQL evaluates (finds the current value of) an expression by combining the values of the operands in ways specified by the operators. This always yields a single value and datatype. PL/SQL determines the datatype by examining the expression and the context in which it appears.

Operator Precedence

The operations within an expression are done in a particular order depending on their precedence (priority). Table 2-1 shows the default order of operations from first to last (top to bottom).
Table 2-1 Order of Operations
Operator Operation
**
exponentiation
+, -
identity, negation
*, /
multiplication, division
+, -, ||
addition, subtraction, concatenation
=, <, >, <=, >=, <>, !=, ~=, ^=, IS NULL, LIKE, BETWEEN, IN
comparison
NOT
logical negation
AND
conjunction
OR
inclusion


Operators with higher precedence are applied first. In the example below, both expressions yield 8 because division has a higher precedence than addition. Operators with the same precedence are applied in no particular order.
5 + 12 / 4
12 / 4 + 5

You can use parentheses to control the order of evaluation. For example, the following expression yields 7, not 11, because parentheses override the default operator precedence:
(8 + 6) / 2

In the next example, the subtraction is done before the division because the most deeply nested subexpression is always evaluated first:
100 + (20 / 5 + (7 - 3))

The following example shows that you can always use parentheses to improve readability, even when they are not needed:
(salary * 0.05) + (commission * 0.25)

Logical Operators

The logical operators AND, OR, and NOT follow the tri-state logic shown in Table 2-2. AND and OR are binary operators; NOT is a unary operator.
Table 2-2 Logic Truth Table
x y x AND y x OR y NOT x
TRUE
TRUE
TRUE
TRUE
FALSE
TRUE
FALSE
FALSE
TRUE
FALSE
TRUE
NULL
NULL
TRUE
FALSE
FALSE
TRUE
FALSE
TRUE
TRUE
FALSE
FALSE
FALSE
FALSE
TRUE
FALSE
NULL
FALSE
NULL
TRUE
NULL
TRUE
NULL
TRUE
NULL
NULL
FALSE
FALSE
NULL
NULL
NULL
NULL
NULL
NULL
NULL


As the truth table shows, AND returns TRUE only if both its operands are true. On the other hand, OR returns TRUE if either of its operands is true. NOT returns the opposite value (logical negation) of its operand. For example, NOT TRUE returns FALSE.
NOT NULL returns NULL because nulls are indeterminate. It follows that if you apply the NOT operator to a null, the result is also indeterminate. Be careful. Nulls can cause unexpected results; see "Handling Null Values in Comparisons and Conditional Statements".

Order of Evaluation

When you do not use parentheses to specify the order of evaluation, operator precedence determines the order. Compare the following expressions:
NOT (valid AND done)     |     NOT valid AND done

If the Boolean variables valid and done have the value FALSE, the first expression yields TRUE. However, the second expression yields FALSE because NOT has a higher precedence than AND. Therefore, the second expression is equivalent to:
(NOT valid) AND done

In the following example, notice that when valid has the value FALSE, the whole expression yields FALSE regardless of the value of done:
valid AND done

Likewise, in the next example, when valid has the value TRUE, the whole expression yields TRUE regardless of the value of done:
valid OR done

Short-Circuit Evaluation

When evaluating a logical expression, PL/SQL uses short-circuit evaluation. That is, PL/SQL stops evaluating the expression as soon as the result can be determined. This lets you write expressions that might otherwise cause an error. Consider the following OR expression:
DECLARE
   ...
   on_hand  INTEGER;
   on_order INTEGER;
BEGIN
   ..
   IF (on_hand = 0) OR ((on_order / on_hand) < 5) THEN 
      ...
   END IF;
END;

When the value of on_hand is zero, the left operand yields TRUE, so PL/SQL need not evaluate the right operand. If PL/SQL were to evaluate both operands before applying the OR operator, the right operand would cause a division by zero error. In any case, it is a poor programming practice to rely on short-circuit evaluation.

Comparison Operators

Comparison operators compare one expression to another. The result is always true, false, or null. Typically, you use comparison operators in conditional control statements and in the WHERE clause of SQL data manipulation statements. Here are a couple of examples:
IF quantity_on_hand > 0 THEN
   UPDATE inventory SET quantity = quantity - 1
      WHERE part_number = item_number;
ELSE
   ...
END IF;

Relational Operators

The relational operators allow you to compare arbitrarily complex expressions. The following list gives the meaning of each operator:
Operator Meaning
=
equal to
<>, !=, ~=, ^=
not equal to
<
less than
>
greater than
<=
less than or equal to
>=
greater than or equal to
IS NULL Operator
The IS NULL operator returns the Boolean value TRUE if its operand is null or FALSE if it is not null. Comparisons involving nulls always yield NULL. So, test for nullity (the state of being null), as follows:
IF variable IS NULL THEN ...
LIKE Operator
You use the LIKE operator to compare a character, string, or CLOB value to a pattern. Case is significant. LIKE returns the Boolean value TRUE if the patterns match or FALSE if they do not match.
The patterns matched by LIKE can include two special-purpose characters called wildcards. An underscore (_) matches exactly one character; a percent sign (%) matches zero or more characters. For example, if the value of ename is 'JOHNSON', the following expression is true:
ename LIKE 'J%SON'

BETWEEN Operator
The BETWEEN operator tests whether a value lies in a specified range. It means "greater than or equal to low value and less than or equal to high value." For example, the following expression is false:
45 BETWEEN 38 AND 44
IN Operator
The IN operator tests set membership. It means "equal to any member of." The set can contain nulls, but they are ignored. For example, the following statement does not delete rows in which the ename column is null:
DELETE FROM emp WHERE ename IN (NULL, 'KING', 'FORD');

Furthermore, expressions of the form
value NOT IN set

yield FALSE if the set contains a null. For example, instead of deleting rows in which the ename column is not null and not 'KING', the following statement deletes no rows:
DELETE FROM emp WHERE ename NOT IN (NULL, 'KING');

Concatenation Operator

Double vertical bars (||) serve as the concatenation operator, which appends one string (CHAR, VARCHAR2, CLOB, or the equivalent Unicode-enabled type) to another. For example, the expression
'suit' || 'case'

returns the following value:
'suitcase'

If both operands have datatype CHAR, the concatenation operator returns a CHAR value. If either operand is a CLOB value, the operator returns a temporary CLOB. Otherwise, it returns a VARCHAR2 value.

Boolean Expressions

PL/SQL lets you compare variables and constants in both SQL and procedural statements. These comparisons, called Boolean expressions, consist of simple or complex expressions separated by relational operators. Often, Boolean expressions are connected by the logical operators AND, OR, and NOT. A Boolean expression always yields TRUE, FALSE, or NULL.
In a SQL statement, Boolean expressions let you specify the rows in a table that are affected by the statement. In a procedural statement, Boolean expressions are the basis for conditional control. There are three kinds of Boolean expressions: arithmetic, character, and date.

Boolean Arithmetic Expressions

You can use the relational operators to compare numbers for equality or inequality. Comparisons are quantitative; that is, one number is greater than another if it represents a larger quantity. For example, given the assignments
number1 := 75;
number2 := 70;

the following expression is true:
number1 > number2

Boolean Character Expressions

You can compare character values for equality or inequality. By default, comparisons are based on the binary values of each byte in the string.
For example, given the assignments
string1 := 'Kathy';
string2 := 'Kathleen';

the following expression is true:
string1 > string2

By setting the initialization parameter NLS_COMP=ANSI, you can make comparisons use use the collating sequence identified by the NLS_SORT initialization parameter. A collating sequence is an internal ordering of the character set in which a range of numeric codes represents the individual characters. One character value is greater than another if its internal numeric value is larger. Each language might have different rules about where such characters occur in the collating sequence. For example, an accented letter might be sorted differently depending on the database character set, even though the binary value is the same in each case.
There are semantic differences between the CHAR and VARCHAR2 base types that come into play when you compare character values. For more information, see Appendix B.
Many types can be converted to character types. For example, you can compare, assign, and do other character operations using CLOB variables. For details on the possible conversions, see "Character Types".

Boolean Date Expressions

You can also compare dates. Comparisons are chronological; that is, one date is greater than another if it is more recent. For example, given the assignments
date1 := '01-JAN-91';
date2 := '31-DEC-90';

the following expression is true:
date1 > date2
Guidelines for PL/SQL Boolean Expressions
  • In general, do not compare real numbers for exact equality or inequality. Real numbers are stored as approximate values. So, for example, the following IF condition might not yield TRUE:
    count := 1;
    IF count = 1.0 THEN
       ...
    END IF;
    
    
  • It is a good idea to use parentheses when doing comparisons. For example, the following expression is not allowed because 100 < tax yields a Boolean value, which cannot be compared with the number 500:
    100 < tax < 500  -- not allowed
    
    
    The debugged version follows:
    (100 < tax) AND (tax < 500)
    
    
  • A Boolean variable is itself either true or false. So, comparisons with the Boolean values TRUE and FALSE are redundant. For example, assuming the variable done is of type BOOLEAN, the WHILE statement
    WHILE NOT (done = TRUE) LOOP
       ...
    END LOOP;
    
    
    can be simplified as follows:
    WHILE NOT done LOOP
       ...
    END LOOP;
    
    
  • Using CLOB values with comparison operators, or functions such as LIKE and BETWEEN, can result in creation of temporary LOBs. You might need to make sure your temporary tablespace is large enough to handle these temporary LOBs. For details, see the "Modeling and Design" chapter in Oracle9i Application Developer's Guide - Large Objects (LOBs).

CASE Expressions

A CASE expression selects a result from one or more alternatives, and returns the result. The CASE expression uses a selector, an expression whose value determines which alternative to return. A CASE expression has the following form:
CASE selector
   WHEN expression1 THEN result1
   WHEN expression2 THEN result2
   ...
   WHEN expressionN THEN resultN
  [ELSE resultN+1]
END;

The selector is followed by one or more WHEN clauses, which are checked sequentially. The value of the selector determines which clause is executed. The first WHEN clause that matches the value of the selector determines the result value, and subsequent WHEN clauses are not evaluated. An example follows:
DECLARE
   grade CHAR(1) := 'B';
   appraisal VARCHAR2(20);
BEGIN
   appraisal := 
      CASE grade
         WHEN 'A' THEN 'Excellent'
         WHEN 'B' THEN 'Very Good'
         WHEN 'C' THEN 'Good'
         WHEN 'D' THEN 'Fair'
         WHEN 'F' THEN 'Poor'
         ELSE 'No such grade'
      END;
END;

The optional ELSE clause works similarly to the ELSE clause in an IF statement. If the value of the selector is not one of the choices covered by a WHEN clause, the ELSE clause is executed. If no ELSE clause is provided and none of the WHEN clauses are matched, the expression returns NULL.
An alternative to the CASE expression is the CASE statement, where each WHEN clause can be an entire PL/SQL block. For details, see "CASE Statement".

Searched CASE Expression

PL/SQL also provides a searched CASE expression, which has the form:
CASE
   WHEN search_condition1 THEN result1
   WHEN search_condition2 THEN result2
   ...
   WHEN search_conditionN THEN resultN
  [ELSE resultN+1]
END;

A searched CASE expression has no selector. Each WHEN clause contains a search condition that yields a Boolean value, which lets you test different variables or multiple conditions in a single WHEN clause. An example follows:
DECLARE
   grade CHAR(1);
   appraisal VARCHAR2(20);
BEGIN
   ...
   appraisal := 
      CASE
         WHEN grade = 'A' THEN 'Excellent'
         WHEN grade = 'B' THEN 'Very Good'
         WHEN grade = 'C' THEN 'Good'
         WHEN grade = 'D' THEN 'Fair'
         WHEN grade = 'F' THEN 'Poor'
         ELSE 'No such grade'
      END;
   ...
END;

The search conditions are evaluated sequentially. The Boolean value of each search condition determines which WHEN clause is executed. If a search condition yields TRUE, its WHEN clause is executed. After any WHEN clause is executed, subsequent search conditions are not evaluated. If none of the search conditions yields TRUE, the optional ELSE clause is executed. If no WHEN clause is executed and no ELSE clause is supplied, the value of the expression is NULL.

Handling Null Values in Comparisons and Conditional Statements

When working with nulls, you can avoid some common mistakes by keeping in mind the following rules:
  • Comparisons involving nulls always yield NULL
  • Applying the logical operator NOT to a null yields NULL
  • In conditional control statements, if the condition yields NULL, its associated sequence of statements is not executed
  • If the expression in a simple CASE statement or CASE expression yields NULL, it cannot be matched by using WHEN NULL. In this case, you would need to use the searched case syntax and test WHEN expression IS NULL.
In the example below, you might expect the sequence of statements to execute because x and y seem unequal. But, nulls are indeterminate. Whether or not x is equal to y is unknown. Therefore, the IF condition yields NULL and the sequence of statements is bypassed.
x := 5;
y := NULL;
...
IF x != y THEN  -- yields NULL, not TRUE
   sequence_of_statements;  -- not executed
END IF;

In the next example, you might expect the sequence of statements to execute because a and b seem equal. But, again, that is unknown, so the IF condition yields NULL and the sequence of statements is bypassed.
a := NULL;
b := NULL;
...
IF a = b THEN  -- yields NULL, not TRUE
   sequence_of_statements;  -- not executed
END IF;

NOT Operator

Recall that applying the logical operator NOT to a null yields NULL. Thus, the following two statements are not always equivalent:
IF x > y THEN     |     IF NOT x > y THEN
   high := x;     |        high := y;
ELSE              |     ELSE
   high := y;     |        high := x;
END IF;           |     END IF;

The sequence of statements in the ELSE clause is executed when the IF condition yields FALSE or NULL. If neither x nor y is null, both IF statements assign the same value to high. However, if either x or y is null, the first IF statement assigns the value of y to high, but the second IF statement assigns the value of x to high.

Zero-Length Strings

PL/SQL treats any zero-length string like a null. This includes values returned by character functions and Boolean expressions. For example, the following statements assign nulls to the target variables:
null_string := TO_CHAR('');
zip_code := SUBSTR(address, 25, 0);
valid := (name != '');

So, use the IS NULL operator to test for null strings, as follows:
IF my_string IS NULL THEN ...

Concatenation Operator

The concatenation operator ignores null operands. For example, the expression
'apple' || NULL || NULL || 'sauce'

returns the following value:
'applesauce'

Functions

If a null argument is passed to a built-in function, a null is returned except in the following cases.
The function DECODE compares its first argument to one or more search expressions, which are paired with result expressions. Any search or result expression can be null. If a search is successful, the corresponding result is returned. In the following example, if the column rating is null, DECODE returns the value 1000:
SELECT DECODE(rating, NULL, 1000, 'C', 2000, 'B', 4000, 'A', 5000)
   INTO credit_limit FROM accts WHERE acctno = my_acctno;

The function NVL returns the value of its second argument if its first argument is null. In the example below, if hire_date is null, NVL returns the value of SYSDATE. Otherwise, NVL returns the value of hire_date:
start_date := NVL(hire_date, SYSDATE);

The function REPLACE returns the value of its first argument if its second argument is null, whether the optional third argument is present or not. For instance, after the assignment
new_string := REPLACE(old_string, NULL, my_string);

the values of old_string and new_string are the same.
If its third argument is null, REPLACE returns its first argument with every occurrence of its second argument removed. For example, after the assignments
syllabified_name := 'Gold-i-locks';
name := REPLACE(syllabified_name, '-', NULL);

the value of name is 'goldilocks'
If its second and third arguments are null, REPLACE simply returns its first argument.

Built-In Functions

PL/SQL provides many powerful functions to help you manipulate data. These built-in functions fall into the following categories:
error reporting
number
character
datatype conversion
date
object reference
miscellaneous
Table 2-3 shows the functions in each category. For descriptions of the error-reporting functions, see Chapter 13. For descriptions of the other functions, see Oracle9i SQL Reference.
Except for the error-reporting functions SQLCODE and SQLERRM, you can use all the functions in SQL statements. Also, except for the object-reference functions DEREF, REF, and VALUE and the miscellaneous functions DECODE, DUMP, and VSIZE, you can use all the functions in procedural statements.
Although the SQL aggregate functions (such as AVG and COUNT) and the SQL analytic functions (such as CORR and LAG) are not built into PL/SQL, you can use them in SQL statements (but not in procedural statements).
Table 2-3 Built-In Functions
Error Number Character Conversion Date Obj Ref Misc
SQLCODE
SQLERRM
ABS
ACOS
ASIN
ATAN
ATAN2
BITAND
CEIL
COS
COSH
EXP
FLOOR
LN
LOG
MOD
POWER
ROUND
SIGN
SIN
SINH
SQRT
TAN
TANH
TRUNC
ASCII
CHR
CONCAT
INITCAP
INSTR
INSTRB
LENGTH
LENGTHB
LOWER
LPAD
LTRIM
NLS_INITCAP
NLS_LOWER
NLSSORT
NLS_UPPER
REPLACE
RPAD
RTRIM
SOUNDEX
SUBSTR
SUBSTRB
TRANSLATE
TRIM
UPPER
CHARTOROWID
CONVERT
HEXTORAW
RAWTOHEX
ROWIDTOCHAR
TO_BLOB
TO_CHAR
TO_CLOB
TO_DATE
TO_MULTI_BYTE
TO_NCLOB
TO_NUMBER
TO_SINGLE_BYTE
ADD_MONTHS
CURRENT_DATE
CURRENT_TIMESTAMP
DBTIMEZONE
EXTRACT
FROM_TZ
LAST_DAY
LOCALTIMESTAMP
MONTHS_BETWEEN
NEW_TIME
NEXT_DAY
NUMTODSINTERVAL
NUMTOYMINTERVAL
ROUND
SESSIONTIMEZONE
SYSDATE
SYSTIMESTAMP
TO_DSINTERVAL
TO_TIMESTAMP
TO_TIMESTAMP_LTZ
TO_TIMESTAMP_TZ
TO_YMINTERVAL
TZ_OFFSET
TRUNC
DEREF
REF
VALUE
TREAT
BFILENAME
DECODE
DUMP
EMPTY_BLOB
EMPTY_CLOB
GREATEST
LEAST
NLS_CHARSET_DECL_LEN
NLS_CHARSET_ID
NLS_CHARSET_NAME
NVL
SYS_CONTEXT
SYS_GUID
UID
USER
USERENV
VSIZE

No comments:

Post a Comment