Tài liệu Oracle PL/SQL Language Pocket Reference- P21 doc

What
database
objects are
dependent
on Foo_t?
USER_DEPENDENCIES
SELECT name, type
FROM user_dependencies
WHERE referenced_name = 'FOO_T';
18.6.2 SQL*Plus "Describe" Command
If you're like me and don't like to type any more than necessary, you'll appreciate a wonderful enhancement
that Oracle has provided for the describe command in SQL*Plus. It will report not only the attributes of an
object type, but also the methods and their arguments. To illustrate:
SQL> desc pet_t
Name Null? Type
------------------------------- -------- ----
TAG_NO NUMBER(38)
NAME VARCHAR2(60)
ANIMAL_TYPE VARCHAR2(30)
SEX VARCHAR2(1)
PHOTO BINARY FILE LOB
VACCINATIONS VACCINATION_LIST_T
OWNER REF OF PERSON_T
METHOD
------
MEMBER FUNCTION SET_TAG_NO RETURNS PET_T
Argument Name Type In/Out
Default?
------------------------------ ----------------------- ------
--------
NEW_TAG_NO NUMBER IN

METHOD
------
MEMBER FUNCTION SET_PHOTO RETURNS PET_T
Argument Name Type In/Out
Default?
------------------------------ ----------------------- ------
--------
FILE_LOCATION VARCHAR2 IN
MEMBER PROCEDURE PRINT_ME
Although the formatting could be improved, this is much easier than SELECTing the equivalent information
from the data dictionary.
18.6.3 Schema Evolution
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Let's say that you have created an object type and you need to make a change to its definition. What do you
do? The answer is that it depends -- on whether you have used the type, and on what type of change you want
to make. Precious few modifications are easy; the rest will probably age you prematurely. Consider the
implications of where you have used the type:
●
Type has no dependencies. Using CREATE OR REPLACE, you can change the object type to you
heart's content. Or drop and recreate it; who cares? Life is good.
●
Type is used only in PL/SQL modules. In this case, since you don't have to rebuild any dependent
tables, life is still easy. Oracle will automatically recompile dependent PL/SQL modules the next time
they are called.
●
Type is used in one or more tables. Consider what would be a simple change to a relational table:
adding a column. If you try to add a column to an object table, you get an "ORA-22856 cannot add
columns to object tables." The "Action" for this message says we need to "Create a new type with
additional attributes, and use the new type to create an object table. The new object table will have the
desired columns." Your frustrations are beginning.

OK, if you want to add an attribute, you're out of luck. What about methods? Oracle8.0 does include an
ALTER TYPE statement that allows you to recompile an object specification or body. It also allows you to
add new methods. It is extremely limited, however; it does not allow you to add or remove attributes, nor does
it allow you to modify the quantity or datatypes of existing method arguments. The basic syntax is:
Form I
ALTER TYPE [ BODY ] type_name COMPILE [ SPECIFICATION | BODY ];
which does not solve our problem, or:
Form II
ALTER TYPE [ BODY ] type_name REPLACE
<the entire new type or body definition>;
Using Form II, we can, in fact, add an entirely new method to an object type, even if there are dependencies on
the type.
In the case of changing a method's specification (or deleting a method) in object type Foo_t which is
implemented in table foo, you would think that export/import would work, using something like:
1. Export the foo table.
2. Drop the foo table.
3. CREATE OR REPLACE TYPE Foo_t with the new definition.
4. Import the foo table.
But alas, it doesn't work, because when you CREATE OR REPLACE the type, it actually assigns a new OID
to the type, and the import fails with IMP-00063 when it sees that the OID is different. Huh? What do you
mean, "assigns a new OID to the type?" For reasons apparently having to do with facilitating certain
operations in the Oracle Call Interface (OCI), object types themselves have an OID. See for yourself -- you can
easily retrieve them from the USER_TYPES data dictionary view.
Neither can you "CREATE new_object_table AS SELECT ... FROM old_object_table." Even if you could, the
REFs wouldn't match up to the OIDs of the new table.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
It's even worse if you want to make any serious modifications to an object type and you have a dependency on
the type from other types or tables. You cannot drop and recreate a parent object table unless you drop the
child object types and object tables first. So maybe you could:
1. Create new object types and tables.

2. Somehow populate new from the old.
3. Drop the old object tables and types.
4. Rename the new types and object tables to the old names.
It is not obvious to me how to do the second step in a way that will preserve REFs to the type. The only way I
see to do it in a guaranteed fashion is to rely on relational primary and foreign keys for tuple identification.
That is, your schema will include not only REFs but also equivalent foreign keys. Then, when your OIDs
change because you have rebuilt an object table, you can update all the REFs to that object table using foreign
key values. Not a pretty picture.
Also, you cannot rename object types (number 4 above); attempting to do so fails with "ORA-03001:
unimplemented feature."
WARNING: Requiring the dropping of all dependent types and objects before altering a type is
not going to endear the Oracle objects option to the average database administrator (or to anyone
else, for that matter). Object schema evolution is a significant area where Oracle could make a
lot of improvements.
Previous: 18.5 Modifying
Persistent Objects
Oracle PL/SQL
Programming, 2nd Edition
Next: 18.7 Making the
Objects Option Work
18.5 Modifying Persistent
Objects
Book Index
18.7 Making the Objects
Option Work
The Oracle Library
Navigation

Copyright (c) 2000 O'Reilly & Associates. All rights reserved.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.

Previous: 18.6 Object
Housekeeping
Chapter 18
Object Types
Next: 19. Nested Tables
and VARRAYs

18.7 Making the Objects Option Work
This stuff isn't designed to be easy for the beginner, and the complexities are more than syntax-deep.
In addition to the operational limitations we have discussed, the act of "thinking objects" is not a trait
that comes naturally to programmers schooled in database or structured approaches. But if you feel
intimidated, take heart from this advice: "There may be an OO revolution, but that does not mean you
have to make the change all at once. Instead, you can incorporate what you know worked before, and
bring in the best that OO has to offer, a little at a time as you understand it."[16]
[16] See Rick Mercer and A. Michael Berman, "Object-Oriented Technology and C++
in the First Year: Ten Lessons Learned." Presented at the Northeastern Small College
Computing Conference, April 18- 20, 1996, and on the web at />~berman/tenLessons/paper.htm.
If object technology is such a challenge, what is it that drives many organizations to consider object
approaches in the first place? The overriding interest of managers seems to be their desire to reuse
rather than reinvent the software needed to run their businesses.[
17] In industries whose automation
needs are not satisfied by off-the-shelf solutions, IS managers are continuously squeezed by the need
to deliver more and more solutions while maintaining their legacy code, all while attempting to keep
costs under control.
[17] See Ivar Jacobson, "Reuse in Reality: The Reuse-Driven Software-Engineering
Business." Presented at Object Expo Paris, available at />support/techpapers/objex_ivar.pdf.
It may not be obvious from our examples just how the objects option is going to facilitate reuse,
particularly given Oracle8.0's lack of inheritance and difficulties with schema evolution. Indeed, the
benefits of an object approach do not automatically accrue to the practitioner; large systems, in
particular, must exhibit other characteristics.[18] Achieving reuse requires careful planning and

deliberate execution.
[18] See Grady Booch, Object-Oriented Analysis and Design with Applications,
Addison-Wesley 1994.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Experts recommend not attempting object approaches just because someone says they are cool or
because everyone else is doing it. Without a financial and time commitment to understanding, and
without taking advantage of a different programming model, you are not likely to get much benefit,
and yours will join the landscape of projects that didn't deliver.
Yes, object approaches can be a way to do more with less. In fact, computer industry pundits assert
that "componentware" is becoming the dominant form of software, and that application development
is evolving into a process of wiring together components -- whether built in-house or procured --
rather than developing software from scratch. These components are typically built using object
design (to specify the component's interfaces, and what it will and won't do) and object-oriented
programming languages. The game isn't over, though; we need look only as close as the nearest
desktop computer to see both the benefits and the perils of componentware. Windows DLLs, for
example, allow module sharing and dynamic loading, but lack a superstructure for managing multiple
versions. Other component models exist (CORBA, ActiveX, COM, JavaBeans) in varying states of
industry acceptance.
Almost certainly, Oracle Corporation will be adding needed features such as inheritance and schema
evolution tools to their objects option. One day, objects may even be a standard part of the server.
Until the technology matures, early adopters will enjoy the pleasures of finding workarounds, and
will gain a deeper appreciation of features that appear later in the product.
Previous: 18.6 Object
Housekeeping
Oracle PL/SQL
Programming, 2nd Edition
Next: 19. Nested Tables
and VARRAYs
18.6 Object Housekeeping
Book Index

19. Nested Tables and
VARRAYs
The Oracle Library
Navigation

Copyright (c) 2000 O'Reilly & Associates. All rights reserved.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Previous: 18.7 Making the
Objects Option Work
Chapter 19
Next: 19.2 Creating the
New Collections

19. Nested Tables and VARRAYs
Contents:
Types of Collections
Creating the New Collections
Syntax for Declaring Collection Datatypes
Using Collections
Collection Pseudo-Functions
Collection Built-Ins
Example: PL/SQL-to-Server Integration
Collections Housekeeping
Which Collection Type Should I Use?
In PL/SQL Version 2, Oracle introduced the TABLE datatype as a way of storing singly dimensioned
sparse arrays in PL/SQL. Known as the "PL/SQL table," this structure is thoroughly documented in
Chapter 10, PL/SQL Tables. PL/SQL8 introduces two new collection structures that have a wide
range of new uses. These structures are nested tables and variable-size arrays (VARRAYs). Like PL/
SQL tables, the new structures can also be used in PL/SQL programs. But what is dramatic and new
is the ability to use the new collections as the datatypes of fields in conventional tables and attributes

of objects. While not an exhaustive implementation of user-defined datatypes, collections offer rich
new physical (and, by extension, logical) design opportunities for Oracle practitioners.
In this chapter we'll include brief examples showing how to create and use collection types both in
the database and in PL/SQL programs. We'll also show the syntax for creating collection types. We'll
present the three different initialization techniques with additional examples, and we'll discuss the
new built-in "methods," EXTEND, TRIM, and DELETE, for managing collection content. This
chapter also contains an introduction to the new "collection pseudo-functions" that Oracle8 provides
to deal with nonatomic values in table columns. Although we can't cover every aspect of SQL usage,
the examples will give you a sense of how important -- and useful -- these new devices can be,
despite their complexity. We also include a reference section that details all of the built-in methods
for collections: for each we'll show its specification, an example, and some programming
considerations. The chapter concludes with a brief discussion of which type of collection is most
appropriate for some common situations.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
19.1 Types of Collections
Oracle now supports three types of collections:
●
PL/SQL tables are singly dimensioned, unbounded, sparse collections of homogeneous
elements and are available only in PL/SQL (see Chapter 10). These are now called index-by
tables.
●
Nested tables are also singly dimensioned, unbounded collections of homogeneous elements.
They are initially dense but can become sparse through deletions. Nested tables are available
in both PL/SQL and the database (for example, as a column in a table).
●
VARRAYs, like the other two collection types, are also singly dimensioned collections of
homogeneous elements. However, they are always bounded and never sparse. Like nested
tables, they can be used in PL/SQL and in the database. Unlike nested tables, when you store
and retrieve a VARRAY, its element order is preserved.
Using a nested table or VARRAY, you can store and retrieve nonatomic data in a single column. For

example, the employee table used by the HR department could store the date of birth for each
employee's dependents in a single column, as shown in
Table 19.1.
Table 19.1: Storing a Nonatomic Column of Dependents in a Table of Employees
Id (NUMBER) Name (VARCHAR2) Dependents_ages (Dependent_birthdate_t)
10010 Zaphod Beeblebrox 12-JAN-1763
4-JUL-1977
22-MAR-2021
10020 Molly Squiggly 15-NOV-1968
15-NOV-1968
10030 Joseph Josephs
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
10040 Cepheus Usrbin 27-JUN-1995
9-AUG-1996
19-JUN-1997
10050 Deirdre Quattlebaum 21-SEP-1997
It's not terribly difficult to create such a table. First we define the collection type:
CREATE TYPE Dependent_birthdate_t AS VARRAY(10) OF DATE;
Now we can use it in the table definition:
CREATE TABLE employees (
id NUMBER,
name VARCHAR2(50),
...other columns...,
Dependents_ages Dependent_birthdate_t
);
We can populate this table using the following INSERT syntax, which relies on the type's default
constructor to transform a list of dates into values of the proper datatype:
INSERT INTO employees VALUES (42, 'Zaphod
Beeblebrox', ...,
Dependent_birthdate_t( '12-JAN-1765', '4-JUL-1977',

'22-MAR-2021'));
One slight problem: most of us have been trained to view nonatomic data as a design flaw. So why
would we actually want to do this? In some situations (for those in which you don't need to scan the
contents of all the values in all the rows), theoreticians and practitioners alike consider nonatomic
data to be perfectly acceptable. Even the conscience of the relational model, Chris Date, suggests that
relational domains could contain complex values, including lists.[
1] Some database designers have
believed for years that the large percentage of nonatomic data inherent in their applications demands
a nonrelational solution.
[1] See Hugh Darwen and C. J. Date, "The Third Manifesto," SIGMOD Record,
Volume 24 Number 1, March 1995.
Setting aside theoretical arguments about "natural" data representations, Oracle collections provide a
dramatic advantage from an application programmer's perspective: you can pass an entire collection
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
between the database and PL/SQL using a single fetch. This feature alone could have significant
positive impact on application performance.
As we've mentioned, within PL/SQL both nested tables and VARRAYs are ordered collections of
homogeneous elements. Both bear some resemblance to the PL/SQL Version 2 table datatype, the
elder member of the "collection" family. The new types are also singly dimensioned arrays, but they
differ in areas such as sparseness (not exactly), how they're initialized (via a constructor), and
whether they can be null (yes).
One chief difference between nested tables and VARRAYs surfaces when we use them as column
datatypes. Although using a VARRAY as a column's datatype can achieve much the same result as a
nested table, VARRAY data must be predeclared to be of a maximum size, and is actually stored
"inline" with the rest of the table's data.
Nested tables, by contrast, are stored in special auxiliary tables called store tables, and there is no pre-
set limit on how large they can grow. For this reason, Oracle Corporation says that VARRAY
columns are intended for "small" arrays, and that nested tables are appropriate for "large" arrays.
As we've mentioned, the old Version 2 table datatype is now called an index-by table , in honor of the
INDEX BY BINARY_INTEGER syntax required when declaring such a type. Despite the many

benefits of the new collection types, index-by tables have one important unique feature: initial
sparseness.
Table 19.2 illustrates many of the additional differences among index-by tables and the
new collection types.
Table 19.2: Comparing Oracle Collection Types
Characteristic Index-By Table Nested Table VARRAY
Dimensionality Single Single Single
Usable in SQL? No Yes Yes
Usable as column
datatype in a table?
No Yes; data stored
"out of line" (in
separate table)
Yes; data stored "in
line" (in same table)
Uninitialized state Empty (cannot be null);
elements undefined
Atomically null;
illegal to reference
elements
Atomically null;
illegal to reference
elements
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Initialization Automatic, when declared Via constructor,
fetch, assignment
Via constructor,
fetch, assignment
In PL/SQL, elements
referenced via

BINARY_INTEGER
(-2,147,483,647 ..
2,147,483,647)
Positive integer
between 1 and
2,147,483,647
Positive integer
between 1 and
2,147,483,647
Sparse? Yes Initially, no; after
deletions, yes
No
Bounded? No Can be extended Yes
Can assign value to
any element at any
time?
Yes No; may need to
EXTEND first
No; may need to
EXTEND first, and
cannot EXTEND
past upper bound
Means of extending Assign value to element with
a new subscript
Use built-in
EXTEND
procedure (or
TRIM to
condense), with no
predefined

maximum
EXTEND (or
TRIM), but only up
to declared
maximum size
Can be compared for
equality?
No No No
Retains ordering and
subscripts when
stored in and
retrieved from
database?
N/A No Yes
The inevitable question is: Which construct should I use? This chapter reviews some examples of the
new collections and offers some suggestions in this area. The short answer:
●
Nested tables are more flexible than VARRAYs for table columns.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
●
VARRAYs are best when you need bounded arrays that preserve element order.
●
Index-by tables are the only option that allows initial sparseness.
●
If your code must run in both Oracle7 and Oracle8, you can use only index-by tables.
We'll revisit these suggestions in more detail at the end of the chapter. Before diving in, though, let's
review a few of the new terms:
Collection
A term which can have several different meanings:
❍

A nested table, index-by table, or VARRAY datatype
❍
A PL/SQL variable of type nested table, index-by table, or VARRAY
❍
A table column of type nested table or VARRAY
Outer table
A term referring to the "enclosing" table in which you have used a nested table or VARRAY
as a column's datatype
Inner table
The "enclosed" collection that is implemented as a column in a table; also known as a "nested
table column"
Store table
The physical table which Oracle creates to hold values of the inner table
Unfortunately, the term "nested table" can be a bit misleading. A nested table, when declared and
used in PL/SQL, is not nested at all! It is instead fairly similar to an array. Even when you use a
nested table as a table column, in Oracle 8.0 you can only nest these structures to a single level. That
is, your column cannot consist of a nested table of nested tables.
"Variable-size array" is also a deceptive name; one might assume, based on the fact that it is
supposed to be "variable size," that it can be arbitrarily extended; quite the opposite is true. Although
a VARRAY can have a variable number of elements, this number can never exceed the limit that you
define when you create the type.
Previous: 18.7 Making the
Objects Option Work
Oracle PL/SQL
Programming, 2nd Edition
Next: 19.2 Creating the
New Collections
18.7 Making the Objects
Option Work
Book Index

19.2 Creating the New
Collections
The Oracle Library
Navigation

Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Copyright (c) 2000 O'Reilly & Associates. All rights reserved.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Previous: 19.1 Types of
Collections
Chapter 19
Nested Tables and
VARRAYs
Next: 19.3 Syntax for
Declaring Collection
Datatypes

19.2 Creating the New Collections
There are two different ways of creating the new user-defined collection types:
1. You can define a nested table type or VARRAY type "in the database" using the CREATE
TYPE command, which makes the datatype available to use for a variety of purposes:
columns in database tables, variables in PL/SQL programs, and attributes of object types.
2. You can declare the collection type within a PL/SQL program using TYPE ... IS ... syntax.
This collection type will then be available only for use within PL/SQL.
Let's look at a few examples that illustrate how to create collections.
19.2.1 Collections "In the Database"
Before you can define a database table containing a nested table or VARRAY, you must first create
the collection's datatype in the database using the CREATE TYPE statement. There is no good
analogy for this command in Oracle7; it represents new functionality in the server. If we wanted to
create a nested table datatype for variables that will hold lists of color names, we'll specify:

CREATE TYPE Color_tab_t AS TABLE OF VARCHAR2(30);
This command stores the type definition for the Color_tab_t nested table in the data dictionary. Once
created, it can serve as the datatype for items in at least two different categories of database object:
●
A "column" in a conventional table
●
An attribute in an object type
Defining a VARRAY datatype is similar to defining a nested table, but you must also specify an
upper bound on the number of elements collections of this type may contain. For example:
CREATE TYPE Color_array_t AS VARRAY (16) OF VARCHAR2(30);
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
Type Color_array_t has an upper limit of 16 elements regardless of where it is used.
While these examples use VARCHAR2, collections can also consist of other primitive datatypes,
object types, references to object types, or (in PL/SQL only) PL/SQL record types. To show
something other than a table of scalars, let's look at an example of a VARRAY of objects. Here we
define an object type that will contain information about documents:
CREATE TYPE Doc_t AS OBJECT (
doc_id INTEGER,
name VARCHAR2(512),
author VARCHAR2(60),
url VARCHAR2(2000)
);
We can then define a collection type to hold a list of these objects:
CREATE TYPE Doc_array_t AS VARRAY(10) OF Doc_t;
In this case, we've chosen to make it a variable-size array type with a maximum of ten elements.
Another useful application of collections is in their ability to have elements which are REFs
(reference pointers) to objects in the database. That is, your collection may have a number of pointers
to various persistent objects (see
Chapter 18, Object Types, for more discussion of REFs). Consider
this example:

CREATE TYPE Doc_ref_array_t AS TABLE OF REF Doc_t;
This statement says "create a user-defined type to hold lists of pointers to document objects." You
can use a nested table of REFs as you would any other nested table: as a column, as an attribute in an
object type, or as the type of a PL/SQL variable.
NOTE: While Oracle 8.0.3 allows you to create homogeneous collections, in some
cases we might want to build heterogeneous collections. It would be useful to be able
to define a type like the following:
CREATE TYPE Generic_ref_t AS TABLE OF REF ANY;
-- not in 8.0.3
This could allow you to make collections that hold references to more than one type of
object in your database ... or, if OID's are globally unique, each REF could point to any
object in any database on your entire network!1
19.2.1.1 Collection as a "column" in a conventional table
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
In the following case, we are using a nested table datatype as a column. When we create the outer
table personality_inventory, we must tell Oracle what we want to call the "out of line" store table:
CREATE TABLE personality_inventory (
person_id NUMBER,
favorite_colors Color_tab_t,
date_tested DATE,
test_results BLOB)
NESTED TABLE favorite_colors STORE AS favorite_colors_st;
The NESTED TABLE ... STORE AS clause tells Oracle that we want the store table for the
favorite_colors column to be called favorite_colors_st.
You cannot directly manipulate data in the store table, and any attempt to retrieve or store data
directly into favorite_colors_st will generate an error. The only path by which you can read or write
its attributes is via the outer table. (See
Section 19.5, "Collection Pseudo-Functions" for a few
examples of doing so.) You cannot even specify storage parameters for the store table; it inherits the
physical attributes of its outermost table.

As you would expect, if you use a VARRAY as a column rather than as a nested table, no store table
is required. Here, the colors collection is stored "in line" with the rest of the table:
CREATE TABLE birds (
genus VARCHAR2(128),
species VARCHAR2(128),
colors Color_array_t
);
19.2.1.2 Collection as an attribute of an object type
In this example, we are modeling automobile specifications, and each Auto_spec_t object will
include a list of manufacturer's colors in which you can purchase the vehicle. (See
Chapter 18 for
more information about Oracle object types.)
CREATE TYPE Auto_spec_t AS OBJECT (
make VARCHAR2(30),
model VARCHAR2(30),
available_colors Color_tab_t
);
Because there is no data storage required for the object type, it is not necessary to designate a name
for the companion table at the time we issue the CREATE TYPE ... AS OBJECT statement.
When the time comes to implement the type as, say, an object table, you could do this:
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
CREATE TABLE auto_specs OF Auto_spec_t
NESTED TABLE available_colors STORE AS
available_colors_st;
This statement requires a bit of explanation. When you create a "table of objects," Oracle looks at the
object type definition to determine what columns you want. When it discovers that one of the object
type's attributes, available_colors, is in fact a nested table, Oracle treats this table in a way similar to
the examples above; in other words, it wants to know what to name the store table. So the phrase
...NESTED TABLE available_colors STORE AS
available_colors_st

says that you want the available_colors column to have a store table named available_colors_st.
19.2.2 Collections in PL/SQL
Whether you use a predefined collection type or declare one in your program, using it requires that
you declare a variable in a separate step. This declare-type-then-declare-variable motif should be
familiar to you if you have ever used an index-by table or a RECORD type in a PL/SQL program.
19.2.2.1 Collection variables
Using the collection types we've declared above, the following shows some legal declarations of PL/
SQL variables:
DECLARE
-- A variable that will hold a list of available font
colors
font_colors Color_tab_t;
/* The next variable will later hold a temporary copy
of
|| font_colors. Note that we can use %TYPE to refer to
the
|| datatype of font_colors. This illustrates two
different
|| ways of declaring variables of the Color_tab_t type.
*/
font_colors_save font_colors%TYPE;
-- Variable to hold a list of paint colors
paint_mixture Color_array_t;
But there is no reason you must use only types you have created in the database. You can declare
them locally, or mix and match from both sources:
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
DECLARE
/* As with Oracle7 index-by tables, you can define
|| a table datatype here within a declaration
section...

*/
TYPE Number_t IS TABLE OF NUMBER;
/* ...and then you can use your new type in the
declaration
|| of a local variable. The next line declares and
initializes
|| in a single statement. Notice the use of the
constructor,
|| Number_t(value, value, ...), to the right of the ":
="
*/
my_favorite_numbers Number_t := Number_t(42, 65536);
/* Or you can just refer to the Color_tab_t datatype
in the
|| data dictionary. This next line declares a local
variable
|| my_favorite_colors to be a "nested" table and
initializes it
|| with two initial elements using the default
constructor.
*/
my_favorite_colors Color_tab_t := Color_tab_t
('PURPLE', 'GREEN');
BEGIN
/* Once the local variables exist, usage is
independent of whether
|| they were declared from local types or from types
that live in
|| the data dictionary.
*/

my_favorite_colors(2) := 'BLUE'; -- changes 2nd
element to BLUE
my_favorite_numbers(1) := 3.14159; -- changes first
element to pi
END;
This code also illustrates default constructors, which are special functions Oracle provides whenever
you create a type, that serve to initialize and/or populate their respective types. A constructor has the
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
same name as the type, and accepts as arguments a comma-separated list of elements.
19.2.2.2 Collections as components of a record
Using a collection type in a record is very similar to using any other type. You can use VARRAYs,
nested tables, or index-by tables (or any combination thereof) in RECORD datatypes. For example:
DECLARE
TYPE toy_rec_t IS RECORD (
manufacturer INTEGER,
shipping_weight_kg NUMBER,
domestic_colors Color_array_t,
international_colors Color_tab_t
);
RECORD types cannot live in the database; they are only available within PL/SQL programs.
Logically, however, you can achieve a similar result with object types. Briefly, object types can have
a variety of attributes, and you can include the two new collection types as attributes within objects;
or you can define a collection whose elements are themselves objects.
19.2.2.3 Collections as module parameters
Collections can also serve as module parameters. In this case, you cannot return a user-defined type
that is declared in the module itself. You will instead use types that you have built outside the scope
of the module, either via CREATE TYPE or via public declaration in a package.
/* This function provides a pseudo "UNION ALL" operation
on
|| two input parameters of type Color_tab_t. That is, it

creates an
|| OUT parameter which is the superset of the colors of
the two
|| input parameters.
*/
CREATE PROCEDURE make_colors_superset (first_colors IN
Color_tab_t,
second_colors IN Color_tab_t, superset OUT Color_tab_t)
AS
working_colors Color_tab_t := Color_tab_t();
element INTEGER := 1;
which INTEGER;
BEGIN
/* Invoke the EXTEND method to allocate enough storage
|| to the nested table working_colors.
*/
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
working_colors.EXTEND (first_colors.COUNT +
second_colors.COUNT);
/* Loop through each of the input parameters, reading
their
|| contents, and assigning each element to an element
of
|| working_colors. Input collections may be sparse.
*/
which := first_colors.FIRST;
LOOP
EXIT WHEN which IS NULL;
working_colors(element) := first_colors(which);
element := element + 1;

which := first_colors.NEXT(which);
END LOOP;
which := second_colors.FIRST;
LOOP
EXIT WHEN which IS NULL;
working_colors(element) := second_colors(which);
element := element + 1;
which := second_colors.NEXT(which);
END LOOP;
superset := working_colors;
END;
As a bit of an aside, let's take a look at the loops used in the code. The general form you can use to
iterate over the elements of a collection is as follows:
1 which := collection_name.FIRST;
2 LOOP
3 EXIT WHEN which IS NULL;
4 -- do something useful with the current element...
5 which := collection_name.NEXT(which);
6 END LOOP;
This works for both dense and sparse collections. The first assignment statement, at line 1, gets the
subscript of the FIRST element in the collection; if it's NULL, that means there are no elements, and
we would therefore exit immediately at line 3.
But if there are elements in the collection, we reach line 4, where the program will do "something
useful" with the value, such as assign, change, or test its value for some purpose.
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.
The most interesting line of this example is line 5, where we use the NEXT method on the collection
to retrieve the next-higher subscript above "which" on the right-hand side. In the event that a
particular subscript has been DELETEd, the NEXT operator simply skips over it until it finds a non-
deleted element. Also in line 5, if NEXT returns a NULL, that is our cue that we have iterated over
all of the collection's elements, and it's time to exit the loop when we get back to line 3.

You might also ask why we should use the local variable working_colors in the example above? Why
not simply use the superset parameter as the working variable in the program? As it turns out, when
we EXTEND a nested table, it must also read the table. So we would have to make superset an IN
OUT variable, because OUT variables cannot be read within the program. It's better programming
style to avoid using an IN OUT variable when OUT would suffice -- -and more efficient, especially
for remote procedure calls.
19.2.2.4 Collections as the datatype of a function's return value
In the next example, the programmer has defined Color_tab_t as the type of a function return value,
and it is also used as the datatype of a local variable. The same restriction about datatype scope
applies to this usage; types must be declared outside the module's scope.
CREATE FUNCTION true_colors (whose_id IN NUMBER) RETURN
Color_tab_t
AS
l_colors Color_tab_t;
BEGIN
SELECT favorite_colors INTO l_colors
FROM personality_inventory
WHERE person_id = whose_id;
RETURN l_colors;
EXCEPTION
WHEN NO_DATA_FOUND
THEN
RETURN NULL;
END;
This example also illustrates a long-awaited feature: the retrieval of a complex data item in a single
fetch. This is so cool that it bears repeating, so we'll talk more about it later in this chapter.
How would you use this function in a PL/SQL program? Since it acts in the place of a variable of
type Color_tab_t, you can do one of two things with the returned data:
1. Assign the entire result to a collection variable
2. Assign a single element of the result to a variable (as long as the variable is of a type

compatible with the collection's elements)
The first option is easy. Notice, by the way, that this is another circumstance where you don't have to
Please purchase PDF Split-Merge on www.verypdf.com to remove this watermark.