A collection is a homogeneous single dimensional structure, which constitutes an ordered set of elements of a similar type. Being a homogeneous structure, all elements are of the same data type. The structure of the element contains cells with a subscript. The elements reside in these cells to make the index as their location information. The subscript or cell index becomes identification of an element and is used for its access.
Structure of a collection type, SPORT, is shown in the following diagram. Note the subscript and elements into it. A new element, GOLF, enters at the last empty location and is represented as SPORT :
A collection element can be of any valid SQL data type or a user-defined type. An element of the SQL primitive data type is a scalar value while an element of the user-defined type is an object type instance. A collection can be used within a PL/SQL program by declaring a PL/SQL variable of collection type. The local PL/SQL variable can hold the instances of its collection type. Besides, a database column in a table can also be of the schema collection type.
The collections in Oracle are strictly one dimensional. They cannot be realized on two-dimensional coordinates. However, multidimensional arrays can be realized when the collection has an object type or collection type attribute.
A collection can be bounded or unbounded. Bounded collections can accommodate a limited number of elements while unbounded collections have no upper limit for subscripts.
Collections provide an efficient way to organize the data in an array or set format while making the use of object-oriented features. An instance of a nested table or varray collection type is accessed as an object while the data is still stored in database columns. Collections can be used to avail data caching in programs and boost up the performance of SQL operations. On dedicated server connections, a session always uses User Global Area (UGA), a component of PGA, for collection operations. On the other hand, for shared server mode, the collection operations are still carried out in UGA; but UGA is now a part of System Global Area (SGA), thus indirectly in SGA. This is because in shared server connections, multiple server processes can affect a session, thus UGA must be allocated out of the SGA.
Collections are of two types—persistent and non-persistent. A collection is persistent if it stores the collection structure and elements physically in the database. Contrarily, a non-persistent collection is active for a program only that is, maximum up to a session.
Apart from the preceding categories, a collection can be realized in three formats namely, associative array, nested table or varray. This categorization is purely based on their objective and behavioral properties in a PL/SQL program. The following diagram combines the abstract and physical classification of collections:
We will take a quick tour of these collection types now and discuss them in detail in the coming sections:
- Associative array (index-by table): This is the simplest form of non- persistent unbounded collections. As a non-persistent collection, it cannot be stored in the database, but they are available within a PL/SQL block only. The collection structure and data of associative array cannot be retained once the program is completed. Initially, during the days of Oracle 7, it was known as PL/SQL tables. Later, Oracle 8 version released it as index-by tables as they used an index to identify an element.
- Nested table: This is a persistent form of unbounded collections which can be created in the database as well as in PL/SQL block.
- Varray (variable-size array): This is a persistent but bounded form of collection which can be created in the database as well as in PL/SQL. Similar to a nested table, a varray is also a unidimensional homogeneous collection. The collection size and storage scheme are the factors which differentiate varrays from nested tables. Unlike a nested table, a varray can accommodate only a defined (fixed) number of elements.
Selecting an appropriate collection type
Here are a few guidelines to decide upon the appropriate usage of collection types in programs:
Use of associative arrays is required when:
- You have to temporarily cache the program data in an array format for lookup purpose.
- You need string subscripts for the collection elements. Note that it supports negative subscripts, too.
- Map hash tables from the client to the database.
Use of nested tables is preferred when:
- You have to stores data as sets in the database. Database columns of nested table type can be declared to hold the data persistently.
- Perform major array operations such as insertion and deletion, on a large volume of data.
Use of varrays is preferred when:
- You have to store calculated or predefined volume of data in the database. Varray offers limited and defined storage of rows in a collection.
- Order of the elements has to be preserved.
Associative arrays are analogous to conventional arrays or lists which can be defined within a PL/SQL program only. Neither the array structure nor the data can be stored in the database. It can hold the elements of a similar type in a key-value structure without any upper bound to the array. Each cell of the array is distinguished by its subscript, index, or cell number. The index can be a number or a string.
Associative arrays were first introduced in Oracle 7 release as PL/SQL tables to signify its usage within the scope of a PL/SQL block. Oracle 8 release identified the PL/SQL table as Index by table due to its structure as an index-value pair. Oracle 10g release recognized the behavior of index by tables as arrays so as to rename it as associative arrays due to association of an index with an array.
The following diagram explains the physical lookup structure of an associative array:
Associative arrays follow the following syntax for declaration in a PL/SQL declare block:
TYPE [COLL NAME] IS TABLE OF [ELEMENT DATA TYPE] NOT NULL INDEX BY [INDEX DATA TYPE]
In the preceding syntax, the index type signifies the data type of the array subscript. RAW, NUMBER, LONG-RAW, ROWID, and CHAR are the unsupported index data types. The suited index types are BINARY_INTEGER, PLS_INTEGER, POSITIVE, NATURAL, SIGNTYPE, or VARCHAR2.
The element’s data type can be one of the following:
- PL/SQL scalar data type: NUMBER (along with its subtypes), VARCHAR2 (and its subtypes), DATE, BLOB, CLOB, or BOOLEAN
- Inferred data: The data type inherited from a table column, cursor expression or predefined package variable
- User-defined type: A user defined object type or collection type
For illustration, the following are the valid conditions of the associative array in a PL/SQL block:
/*Array of CLOB data*/ TYPE clob_t IS TABLE OF CLOB INDEX BY PLS_INTEGER; /*Array of employee ids indexed by the employee names*/ TYPE empno_t IS TABLE OF employees.empno%TYPE NOT NULL INDEX BY employees.ename%type;
The following PL/SQL program declares an associative array type in a PL/ SQL block. Note that the subscript of the array is of a string type and it stores the number of days in a quarter. This code demonstrates the declaration of an array and assignment of the element in each cell and printing them. Note that the program uses the FIRST and NEXT collection methods to display the array elements. The collection methods would be covered in detail in the PL/SQL collection methods section:
/*Enable the SERVEROUTPUT on to display the output*/ SET SERVEROUTPUT ON /*Start the PL/SQL block*/ DECLARE /*Declare a collection type associative array and its variable*/ TYPE string_asc_arr_t IS TABLE OF NUMBER INDEX BY VARCHAR2(10); l_str string_asc_arr_t; l_idx VARCHAR2(50); BEGIN /*Assign the total count of days in each quarter against each cell*/ l_str ('JAN-MAR') := 90; l_str ('APR-JUN') := 91; l_str ('JUL-SEP') := 92; l_str ('OCT-DEC') := 93; l_idx := l_str.FIRST; WHILE (l_idx IS NOT NULL) LOOP DBMS_OUTPUT.PUT_LINE('Value at index '||l_idx||' is '||l_str(l_ idx)); l_idx := l_str.NEXT(l_idx); END LOOP; END; / Value at index APR-JUN is 91 Value at index JAN-MAR is 90 Value at index JUL-SEP is 92 Value at index OCT-DEC is 93 PL/SQL procedure successfully completed.
In the preceding block, note the string indexed array. A string indexed array considerably improves the performance by using indexed organization of array values. In the last block, we noticed the explicit assignment of data.
In the following program, we will try to populate the array automatically in the program. The following PL/SQL block declares an associative array to hold the ASCII values of number 1 to 100:
/*Enable the SERVEROUTPUT on to display the output*/ SET SERVEROUTPUT ON /*Start the PL/SQL Block*/ DECLARE /*Declare an array of string indexed by numeric subscripts*/ TYPE ASCII_VALUE_T IS TABLE OF VARCHAR2(12) INDEX BY PLS_INTEGER; L_GET_ASCII ASCII_VALUE_T; BEGIN /*Insert the values through a FOR loop*/ FOR I IN 1..100 LOOP L_GET_ASCII(I) := ASCII(I); END LOOP; /*Display the values randomly*/ DBMS_OUTPUT.PUT_LINE(L_GET_ASCII(5)); DBMS_OUTPUT.PUT_LINE(L_GET_ASCII(15)); DBMS_OUTPUT.PUT_LINE(L_GET_ASCII(75)); END; / 53 49 55 PL/SQL procedure successfully completed.
The salient features of associative arrays are as follows:
- An associative array can exist as a sparse or empty collection
- Being a non-persistent collection, it cannot participate in DML transactions
- It can be passed as arguments to other local subprograms within the same block
- Sorting of an associative array depends on the NLS_SORT parameter
- An associative array declared in package specification behaves as a session-persistent array
Nested tables are a persistent form of collections which can be created in the database as well as PL/SQL. It is an unbounded collection where the index or subscript is implicitly maintained by the Oracle server during data retrieval. Oracle automatically marks the minimum subscript as 1 and relatively handles others. As there is no upper limit defined for a nested table, its size can grow dynamically. Though not an index-value pair structure, a nested table can be accessed like an array in a PL/SQL block.
A nested table is initially a dense collection but it might become sparse due to delete operations on the collection cells.
Dense collection is the one which is tightly populated. That means, there exists no empty cells between the lower and upper indexes of the collection. Sparse collections can have empty cells between the first and the last cell of the collection. A dense collection may get sparse by performing the “delete” operations.
When a nested table is declared in a PL/SQL program, they behave as a one-dimensional array without any index type or upper limit specification.
A nested table defined in a database exists as a valid schema object type. It can be either used in a PL/SQL block to declare a PL/SQL variable for temporarily holding program data or a database column of particular nested table type can be included in a table, which can persistently store the data in the database. A nested table type column in a table resembles a table within a table, but Oracle draws an out- of-line storage table to hold the nested table data. This scenario is illustrated in the following diagram:
Whenever a database column of nested table type is created in a table (referred to as parent table), Oracle creates a storage table with the same storage options as that of the parent table. The storage table created by Oracle in the same segment carries the name as specified in the NESTED TABLE STORE AS clause during creation of the parent table. Whenever a row is created in the parent table, the following actions are performed by the Oracle server:
- A unique identifier is generated to distinguish the nested table instances of different parent rows, for the parent row
- The instance of the nested table is created in the storage table alongside the unique identifier of the parent row
The Oracle server takes care of these nested table operations. For the programmer or user, the whole process is hidden and appears as a normal “insert” operation.
A nested table definition in PL/SQL follows the following syntax:
DECLARE TYPE type_name IS TABLE OF element_type [NOT NULL];
In the preceding syntax, element_type is a primitive data type or a user-defined type, but not as a REF CURSOR type.
In a database, a nested table can be defined using the following syntax:
CREATE [OR REPLACE] TYPE type_name IS TABLE OF [element_type] [NOT NULL]; /
In the preceding syntax, [element_type] can be a SQL supported scalar data type, a database object type, or a REF object type. Unsupported element types are BOOLEAN, LONG, LONG-RAW, NATURAL, NATURALN, POSITIVE, POSITIVEN, REF CURSOR, SIGNTYPE, STRING, PLS_INTEGER, SIMPLE_INTEGER, BINARY_INTEGER and all other non-SQL supported data types.
If the size of the element type of a database collection type has to be increased, follow this syntax:
ALTER TYPE [type name] MODIFY ELEMENT TYPE [modified element type] [CASCADE | INVALIDATE];
The keywords, CASCADE or INVALIDATE, decide whether the collection modification has to invalidate the dependents or the changes that have to be cascaded across the dependents.
The nested table from the database can be dropped using the DROP command, as shown in the following syntax (note that the FORCE keyword drops the type irrespective of its dependents):
DROP TYPE [collection name] [FORCE]
Nested table collection type as the database object
We will go through the following illustration to understand the behavior of a nested table, when created as a database collection type:
/*Create the nested table in the database*/ SQL> CREATE TYPE NUM_NEST_T AS TABLE OF NUMBER; / Type created.
The nested table type, NUM_NEST_T, is now created in the database. Its metadata information can be queried from the USER_TYPES and USER_COLL_TYPES dictionary views:
SELECT type_name, typecode, type_oid FROM USER_TYPES WHERE type_name = 'NUM_NEST_T'; TYPE_NAME TYPECODE TYPE_OID --------------- --------------- -------------------------------- NUM_NEST_T COLLECTION 96DE421E47114638A9F5617CE735731A
Note that the TYPECODE value shows the type of the object in the database and differentiates collection types from user-defined object types:
SELECT type_name, coll_type, elem_type_name FROM user_coll_types WHERE type_name = 'NUM_NEST_T'; TYPE_NAME COLL_TYPE ELEM_TYPE_NAME --------------- ---------- -------------------- NUM_NEST_T TABLE NUMBER
Once the collection type has been successfully created in the database, it can be used to specify the type for a database column in a table. The CREATE TABLE statement in the following code snippet declares a column of the NUM_NEST_T nested table type in the parent table, TAB_USE_NT_COL. The NESTED TABLE [Column] STORE AS [Storage table] clause specifies the storage table for the nested table type column. A separate table for the nested table column, NUM, ensures its out-of-line storage.
SQL> CREATE TABLE TAB_USE_NT_COL (ID NUMBER, NUM NUM_NEST_T) NESTED TABLE NUM STORE AS NESTED_NUM_ID; Table created.
A nested table collection type in PL/SQL
n PL/SQL, a nested table can be declared and defined in the declaration section of the block as a local collection type. As a nested table follows object orientation, the PL/SQL variable of the nested table type has to be necessarily initialized. The Oracle server raises the exception ORA-06531: Reference to uninitialized collection if an uninitialized nested table type variable is encountered during block execution.
As the nested table collection type has been declared within the PL/SQL block, its scope, visibility, and life is the execution of the PL/SQL block only.
The following PL/SQL block declares a nested table. Observe the scope and visibility of the collection variable. Note that the COUNT method has been used to display the array elements.
/*Enable the SERVEROUTPUT to display the results*/ SET SERVEROUTPUT ON /*Start the PL/SQL block*/ DECLARE /*Declare a local nested table collection type*/ TYPE LOC_NUM_NEST_T IS TABLE OF NUMBER; L_LOCAL_NT LOC_NUM_NEST_T := LOC_NUM_NEST_T (10,20,30); BEGIN /*Use FOR loop to parse the array and print the elements*/ FOR I IN 1..L_LOCAL_NT.COUNT LOOP DBMS_OUTPUT.PUT_LINE('Printing '||i||' element: '||L_LOCAL_ NT(I)); END LOOP; END; / Printing 1 element: 10 Printing 2 element: 20 Printing 3 element: 30 PL/SQL procedure successfully completed.
Additional features of a nested table
In the earlier sections, we saw the operational methodology of a nested table. We will now focus on the nested table’s metadata. Furthermore, we will demonstrate a peculiar behavior of the nested table for the “delete” operations.
Oracle’s USER_NESTED_TABLES and USER_NESTED_TABLE_COLS data dictionary views maintain the relationship information of the parent and the nested tables. These dictionary views are populated only when a database of a nested table collection type is included in a table.
The USER_NESTED_TABLES static view maintains the information about the mapping of a nested table collection type with its parent table.
The structure of the dictionary view is as follows:
SQL> desc USER_NESTED_TABLES Name Null? Type ----------------------- -------- --------------- TABLE_NAME VARCHAR2(30) TABLE_TYPE_OWNER VARCHAR2(30) TABLE_TYPE_NAME VARCHAR2(30) PARENT_TABLE_NAME VARCHAR2(30) PARENT_TABLE_COLUMN VARCHAR2(4000) STORAGE_SPEC VARCHAR2(30) RETURN_TYPE VARCHAR2(20) ELEMENT_SUBSTITUTABLE VARCHAR2(25)
Let us query the nested table relationship properties for the TAB_USE_NT_COL table from the preceding view:
SELECT parent_table_column, table_name, return_type, storage_spec FROM user_nested_tables WHERE parent_table_name='TAB_USE_NT_COL' / PARENT_TAB TABLE_NAME RETURN_TYPE STORAGE_SPEC ---------------------------------------------------------------------- NUM NESTED_NUM_ID VALUE DEFAULT
In the preceding view query, RETURN_TYPE specifies the return type of the collection. It can be VALUE (in this case) or LOCATOR. Another column, STORAGE_SPEC, signifies the storage scheme used for the storage of a nested table which can be either USER_SPECIFIED or DEFAULT (in this case).
The USER_NESTED_TABLE_COLS view maintains the information about the collection attributes contained in the nested tables:
SQL> desc USER_NESTED_TABLE_COLS Name Null? Type ----------------------- -------- --------------- TABLE_NAME NOT NULL VARCHAR2(30) COLUMN_NAME NOT NULL VARCHAR2(30) DATA_TYPE VARCHAR2(106) DATA_TYPE_MOD VARCHAR2(3) DATA_TYPE_OWNER VARCHAR2(30) DATA_LENGTH NOT NULL NUMBER DATA_PRECISION NUMBER DATA_SCALE NUMBER NULLABLE VARCHAR2(1) COLUMN_ID NUMBER DEFAULT_LENGTH NUMBER DATA_DEFAULT LONG NUM_DISTINCT NUMBER LOW_VALUE RAW(32) HIGH_VALUE RAW(32) DENSITY NUMBER NUM_NULLS NUMBER NUM_BUCKETS NUMBER LAST_ANALYZED DATE SAMPLE_SIZE NUMBER CHARACTER_SET_NAME VARCHAR2(44) CHAR_COL_DECL_LENGTH NUMBER GLOBAL_STATS VARCHAR2(3) USER_STATS VARCHAR2(3) AVG_COL_LEN NUMBER CHAR_LENGTH NUMBER CHAR_USED VARCHAR2(1) V80_FMT_IMAGE VARCHAR2(3) DATA_UPGRADED VARCHAR2(3) HIDDEN_COLUMN VARCHAR2(3) VIRTUAL_COLUMN VARCHAR2(3) SEGMENT_COLUMN_ID NUMBER INTERNAL_COLUMN_ID NOT NULL NUMBER HISTOGRAM VARCHAR2(15) QUALIFIED_COL_NAME VARCHAR2(4000)
We will now query the nested storage table in the preceding dictionary view to list all its attributes:
SELECT COLUMN_NAME, DATA_TYPE, DATA_LENGTH, HIDDEN_COLUMN FROM user_nested_table_cols where table_name='NESTED_NUM_ID' / COLUMN_NAME DATA_TYP DATA_LENGTH HID ------------------------------ ---------- ----------- ----- NESTED_TABLE_ID RAW 16 YES COLUMN_VALUE NUMBER 22 NO
We observe that though the nested table had only number elements, there is two- columned information in the view. The COLUMN_VALUE attribute is the default pseudo column of the nested table as there are no “named” attributes in the collection structure. The other attribute, NESTED_TABLE_ID, is a hidden unique 16-byte system generated raw hash code which latently stores the parent row identifier alongside the nested table instance to distinguish the parent row association.
If an element is deleted from the nested table, it is rendered as parse. This implies that once an index is deleted from the collection structure, the collection doesn’t restructure itself by shifting the cells in a forward direction. Let us check out the sparse behavior in the following example.
The following PL/SQL block declares a local nested table and initializes it with a constructor. We will delete the first element and print it again. The system raises the NO_DATA_FOUND exception when we query the element at the index 1 in the collection:
/*Enable the SERVEROUTPUT to display the block messages*/ SQL> SET SERVEROUTPUT ON /*Start the PL/SQL block*/ SQL> DECLARE /*Declare the local nested table collection*/ TYPE coll_method_demo_t IS TABLE OF NUMBER; /*Declare a collection variable and initialize it*/ L_ARRAY coll_method_demo_t := coll_method_demo_t (10,20,30,40,50); BEGIN /*Display element at index 1*/ DBMS_OUTPUT.PUT_LINE('Element at index 1 before deletion:'||l_ array(1)); /*Delete the 1st element from the collection*/ L_ARRAY.DELETE(1); /*Display element at index 1*/ DBMS_OUTPUT.PUT_LINE('Element at index 1 after deletion:'||l_ array(1)); END; / Element at index 1 before deletion:10 DECLARE * ERROR at line 1: ORA-01403: no data found ORA-06512: at line 15