B PL/SQL Name Resolution

This appendix explains PL/SQL name resolution; that is, how the PL/SQL compiler resolves ambiguous references to identifiers.

An unambiguous identifier reference can become ambiguous if you change identifiers in its compilation unit (that is, if you add, rename, or delete identifiers).

Note:

The AUTHID property of a stored PL/SQL unit affects the name resolution of SQL statements that the unit issues at run time. For more information, see "Invoker's Rights and Definer's Rights (AUTHID Property)".

Topics

B.1 Qualified Names and Dot Notation

When one named item belongs to another named item, you can (and sometimes must) qualify the name of the "child" item with the name of the "parent" item, using dot notation. For example:

When referencing ... You must qualify its name with ... Using this syntax ...

Field of a record

Name of the record

record_name.field_name

Method of a collection

Name of the collection

collection_name.method

Pseudocolumn CURRVAL

Name of a sequence

sequence_name.CURRVAL

Pseudocolumn NEXTVAL

Name of a sequence

sequence_name.NEXTVAL

If an identifier is declared in a named PL/SQL unit, you can qualify its simple name (the name in its declaration) with the name of the unit (block, subprogram, or package), using this syntax:

unit_name.simple_identifier_name

If the identifier is not visible, then you must qualify its name (see "Scope and Visibility of Identifiers").

If an identifier belongs to another schema, then you must qualify its name with the name of the schema, using this syntax:

schema_name.package_name

A simple name can be qualified with multiple names, as Example B-1 shows.

Some examples of possibly ambiguous qualified names are:

  • Field or attribute of a function return value, for example:

    func_name().field_name
    func_name().attribute_name
    
  • Schema object owned by another schema, for example:

    schema_name.table_name
    schema_name.procedure_name()
    schema_name.type_name.member_name()
    
  • Package object owned by another user, for example:

    schema_name.package_name.procedure_name()
    schema_name.package_name.record_name.field_name
    
  • Record containing an ADT, for example:

    record_name.field_name.attribute_name
    record_name.field_name.member_name()

Example B-1 Qualified Names

CREATE OR REPLACE PACKAGE pkg1 AUTHID DEFINER AS
  m NUMBER;
  TYPE t1 IS RECORD (a NUMBER);
  v1 t1;
  TYPE t2 IS TABLE OF t1 INDEX BY PLS_INTEGER;
  v2 t2; 
  FUNCTION f1 (p1 NUMBER) RETURN t1;
  FUNCTION f2 (q1 NUMBER) RETURN t2;
END pkg1;
/

CREATE OR REPLACE PACKAGE BODY pkg1 AS
  FUNCTION f1 (p1 NUMBER) RETURN t1 IS
    n NUMBER;
  BEGIN
     n := m;             -- Unqualified variable name
     n := pkg1.m;        -- Variable name qualified by package name
     n := pkg1.f1.p1;    -- Parameter name qualified by function name,
                         --  which is qualified by package name
     n := v1.a;          -- Variable name followed by component name
     n := pkg1.v1.a;     -- Variable name qualified by package name
                         --  and followed by component name
     n := v2(10).a;      -- Indexed name followed by component name
     n := f1(10).a;      -- Function invocation followed by component name
     n := f2(10)(10).a;  -- Function invocation followed by indexed name
                         --  and followed by component name
     n := hr.pkg1.f2(10)(10).a;  -- Schema name, package name,
                                 -- function invocation, index, component name
     v1.a := p1;
     RETURN v1;
   END f1;

   FUNCTION f2 (q1 NUMBER) RETURN t2 IS
     v_t1 t1;
     v_t2 t2;
   BEGIN
     v_t1.a := q1;
     v_t2(1) := v_t1;
     RETURN v_t2;
   END f2;
END pkg1;
/

B.2 Column Name Precedence

If a SQL statement references a name that belongs to both a column and either a local variable or formal parameter, then the column name takes precedence.

Caution:

When a variable or parameter name is interpreted as a column name, data can be deleted, changed, or inserted unintentionally.

In Example B-2, the name last_name belongs to both a local variable and a column (names are not case-sensitive). Therefore, in the WHERE clause, both references to last_name resolve to the column, and all rows are deleted.

Example B-3 solves the problem in Example B-2 by giving the variable a different name.

Example B-4 solves the problem in Example B-2 by labeling the block and qualifying the variable name with the block name.

In Example B-5, the function dept_name has a formal parameter and a local variable whose names are those of columns of the table DEPARTMENTS. The parameter and variable name are qualified with the function name to distinguish them from the column names.

Example B-2 Variable Name Interpreted as Column Name Causes Unintended Result

DROP TABLE employees2;
CREATE TABLE employees2 AS
  SELECT LAST_NAME FROM employees;
 
DECLARE
  last_name  VARCHAR2(10) := 'King';
BEGIN
  DELETE FROM employees2 WHERE LAST_NAME = last_name;
  DBMS_OUTPUT.PUT_LINE('Deleted ' || SQL%ROWCOUNT || ' rows.');
END;
/
 

Result:

Deleted 107 rows.

Example B-3 Fixing Example B-2 with Different Variable Name

DECLARE
  v_last_name  VARCHAR2(10) := 'King';
BEGIN
  DELETE FROM employees2 WHERE LAST_NAME = v_last_name;
  DBMS_OUTPUT.PUT_LINE('Deleted ' || SQL%ROWCOUNT || ' rows.');
END;
/
 

Result:

Deleted 2 rows.

Example B-4 Fixing Example B-2 with Block Label

<<main>>
DECLARE
  last_name  VARCHAR2(10) := 'King';
BEGIN
  DELETE FROM employees2 WHERE last_name = main.last_name;
  DBMS_OUTPUT.PUT_LINE('Deleted ' || SQL%ROWCOUNT || ' rows.');
END;
/
 

Result:

Deleted 2 rows.

Example B-5 Subprogram Name for Name Resolution

DECLARE
  FUNCTION dept_name (department_id IN NUMBER)
    RETURN departments.department_name%TYPE
  IS
    department_name  departments.department_name%TYPE;
  BEGIN
    SELECT department_name INTO dept_name.department_name
      --    ^column               ^local variable
    FROM departments
    WHERE department_id = dept_name.department_id;
    --     ^column          ^formal parameter
    RETURN department_name;
  END dept_name;
BEGIN
  FOR item IN (
    SELECT department_id
    FROM departments
    ORDER BY department_name) LOOP
 
      DBMS_OUTPUT.PUT_LINE ('Department: ' || dept_name(item.department_id));
  END LOOP;
END;
/
 

Result:

Department: Accounting
Department: Administration
Department: Benefits
Department: Construction
Department: Contracting
Department: Control And Credit
Department: Corporate Tax
Department: Executive
Department: Finance
Department: Government Sales
Department: Human Resources
Department: IT
Department: IT Helpdesk
Department: IT Support
Department: Manufacturing
Department: Marketing
Department: NOC
Department: Operations
Department: Payroll
Department: Public Relations
Department: Purchasing
Department: Recruiting
Department: Retail Sales
Department: Sales
Department: Shareholder Services
Department: Shipping
Department: Treasury

B.3 Differences Between PL/SQL and SQL Name Resolution Rules

PL/SQL and SQL name resolution rules are very similar. However:

  • PL/SQL rules are less permissive than SQL rules.

    Because most SQL rules are context-sensitive, they recognize as legal more situations than PL/SQL rules do.

  • PL/SQL and SQL resolve qualified names differently.

    For example, when resolving the table name HR.JOBS:

    • PL/SQL searches first for packages, types, tables, and views named HR in the current schema, then for public synonyms, and finally for objects named JOBS in the HR schema.

    • SQL searches first for objects named JOBS in the HR schema, and then for packages, types, tables, and views named HR in the current schema.

To avoid problems caused by the few differences between PL/SQL and SQL name resolution rules, follow the recommendations in "Avoiding Inner Capture in SELECT and DML Statements".

Note:

When the PL/SQL compiler processes a static SQL statement, it sends that statement to the SQL subsystem, which uses SQL rules to resolve names in the statement. For details, see "Resolution of Names in Static SQL Statements".

B.4 Resolution of Names in Static SQL Statements

Static SQL is described in PL/SQL Static SQL.

When the PL/SQL compiler finds a static SQL statement:

  1. If the statement is a SELECT statement, the PL/SQL compiler removes the INTO clause.

  2. The PL/SQL compiler sends the statement to the SQL subsystem.

  3. The SQL subsystem checks the syntax of the statement.

    If the syntax is incorrect, the compilation of the PL/SQL unit fails. If the syntax is correct, the SQL subsystem determines the names of the tables and tries to resolve the other names in the scope of the SQL statement.

  4. If the SQL subsystem cannot resolve a name in the scope of the SQL statement, then it sends the name back to the PL/SQL compiler. The name is called an escaped identifier.

  5. The PL/SQL compiler tries to resolve the escaped identifier.

    First, the compiler tries to resolve the identifier in the scope of the PL/SQL unit. If that fails, the compiler tries to resolve the identifier in the scope of the schema. If that fails, the compilation of the PL/SQL unit fails.

  6. If the compilation of the PL/SQL unit succeeds, the PL/SQL compiler generates the text of the regular SQL statement that is equivalent to the static SQL statement and stores that text with the generated computer code.

  7. At run time, the PL/SQL runtime system invokes routines that parse, bind, and run the regular SQL statement.

    The bind variables are the escaped identifiers (see step 4).

  8. If the statement is a SELECT statement, the PL/SQL runtime system stores the results in the PL/SQL targets specified in the INTO clause that the PL/SQL compiler removed in step 1.

Note:

Bind variables can be evaluated in any order. If a program determines order of evaluation, then at the point where the program does so, its behavior is undefined.

B.5 What is Capture?

When a declaration or definition prevents the compiler from correctly resolving a reference in another scope, the declaration or definition is said to capture the reference. Capture is usually the result of migration or schema evolution.

Topics

Note:

Same-scope and inner capture occur only in SQL scope.

B.5.1 Outer Capture

Outer capture occurs when a name in an inner scope, which had resolved to an item in an inner scope, now resolves to an item in an outer scope. Both PL/SQL and SQL are designed to prevent outer capture; you need not be careful to avoid it.

B.5.2 Same-Scope Capture

Same-scope capture occurs when a column is added to one of two tables used in a join, and the new column has the same name as a column in the other table. When only one table had a column with that name, the name could appear in the join unqualified. Now, to avoid same-scope capture, you must qualify the column name with the appropriate table name, everywhere that the column name appears in the join.

B.5.3 Inner Capture

Inner capture occurs when a name in an inner scope, which had resolved to an item in an outer scope, now either resolves to an item in an inner scope or cannot be resolved. In the first case, the result might change. In the second case, an error occurs.

In Example B-6, a new column captures a reference to an old column with the same name. Before new column col2 is added to table tab2, col2 resolves to tab1.col2; afterward, it resolves to tab2.col2.

To avoid inner capture, follow the rules in "Avoiding Inner Capture in SELECT and DML Statements".

Example B-6 Inner Capture of Column Reference

Table tab1 has a column named col2, but table tab2 does not:

DROP TABLE tab1;
CREATE TABLE tab1 (col1 NUMBER, col2 NUMBER);
INSERT INTO tab1 (col1, col2) VALUES (100, 10);

DROP TABLE tab2;
CREATE TABLE tab2 (col1 NUMBER);
INSERT INTO tab2 (col1) VALUES (100);

Therefore, in the inner SELECT statement, the reference to col2 resolves to column tab1.col2:

CREATE OR REPLACE PROCEDURE proc AUTHID DEFINER AS
  CURSOR c1 IS
    SELECT * FROM tab1
    WHERE EXISTS (SELECT * FROM tab2 WHERE col2 = 10);
BEGIN
  OPEN c1;
  CLOSE c1;
END;
/

Add a column named col2 to table tab2:

ALTER TABLE tab2 ADD (col2 NUMBER);

Now procedure proc is invalid. At its next invocation, the database automatically recompiles it, and the reference to col2 in the inner SELECT statement resolves to column tab2.col2.

B.6 Avoiding Inner Capture in SELECT and DML Statements

Avoid inner capture of references in SELECT, SELECT INTO, and DML statements by following these recommendations:

  • Specify a unique alias for each table in the statement.

  • Do not specify a table alias that is the name of a schema that owns an item referenced in the statement.

  • Qualify each column reference in the statement with the appropriate table alias.

In Example B-7, schema hr owns tables tab1 and tab2. Table tab1 has a column named tab2, whose Abstract Data Type (ADT) has attribute a. Table tab2 does not have a column named a. Against recommendation, the query specifies alias hr for table tab1 and references table tab2. Therefore, in the query, the reference hr.tab2.a resolves to table tab1, column tab2, attribute a. Then the example adds column a to table tab2. Now the reference hr.tab2.a in the query resolves to schema hr, table tab2, column a. Column a of table tab2 captures the reference to attribute a in column tab2 of table tab1.

Topics

Example B-7 Inner Capture of Attribute Reference

CREATE OR REPLACE TYPE type1 AS OBJECT (a NUMBER);
/
DROP TABLE tab1;
CREATE TABLE tab1 (tab2 type1);
INSERT INTO tab1 (tab2) VALUES (type1(10));

DROP TABLE tab2;
CREATE TABLE tab2 (x NUMBER);
INSERT INTO tab2 (x) VALUES (10);

/* Alias tab1 with same name as schema name,
   a bad practice used here for illustration purpose.
   Note lack of alias in second SELECT statement. */

SELECT * FROM tab1 hr
WHERE EXISTS (SELECT * FROM hr.tab2 WHERE x = hr.tab2.a);

Result:

TAB2(A)
---------------
 
TYPE1(10)
 
1 row selected.

Add a column named a to table tab2 (which belongs to schema hr):

ALTER TABLE tab2 ADD (a NUMBER);

Now, when the query runs, hr.tab2.a resolves to schema hr, table tab2, column a. To avoid this inner capture, apply the recommendations to the query:

SELECT * FROM hr.tab1 p1
WHERE EXISTS (SELECT * FROM hr.tab2 p2 WHERE p2.x = p1.tab2.a);

B.6.1 Qualifying References to Attributes and Methods

To reference an attribute or method of a table element, you must give the table an alias and use the alias to qualify the reference to the attribute or method.

In Example B-8, table tbl1 has column col1 of data type t1, an ADT with attribute x. The example shows several correct and incorrect references to tbl1.col1.x.

Example B-8 Qualifying ADT Attribute References

CREATE OR REPLACE TYPE t1 AS OBJECT (x NUMBER);
/
DROP TABLE tb1;
CREATE TABLE tb1 (col1 t1); 

The references in the following INSERT statements do not need aliases, because they have no column lists:

BEGIN
  INSERT INTO tb1 VALUES ( t1(10) );
  INSERT INTO tb1 VALUES ( t1(20) );
  INSERT INTO tb1 VALUES ( t1(30) );
END;
/

The following references to the attribute x cause error ORA-00904:

UPDATE tb1 SET col1.x = 10 WHERE col1.x = 20;

UPDATE tb1 SET tb1.col1.x = 10 WHERE tb1.col1.x = 20;

UPDATE hr.tb1 SET hr.tb1.col1.x = 10 WHERE hr.tb1.col1.x = 20;

DELETE FROM tb1 WHERE tb1.col1.x = 10;

The following references to the attribute x, with table aliases, are correct:

UPDATE hr.tb1 t SET t.col1.x = 10 WHERE t.col1.x = 20;

DECLARE
  y NUMBER;
BEGIN
  SELECT t.col1.x INTO y FROM tb1 t WHERE t.col1.x = 30;
END;
/

DELETE FROM tb1 t WHERE t.col1.x = 10;

B.6.2 Qualifying References to Row Expressions

Row expressions must resolve as references to table aliases. A row expression can appear in the SET clause of an UPDATE statement or be the parameter of the SQL function REF or VALUE.

In Example B-9, table ot1 is a standalone nested table of elements of data type t1, an ADT with attribute x. The example shows several correct and incorrect references to row expressions.

Example B-9 Qualifying References to Row Expressions

CREATE OR REPLACE TYPE t1 AS OBJECT (x number);
/
DROP TABLE ot1;
CREATE TABLE ot1 OF t1;

BEGIN
  INSERT INTO ot1 VALUES (t1(10));
  INSERT INTO ot1 VALUES (20);
  INSERT INTO ot1 VALUES (30);
END;
/

The following references cause error ORA-00904:

UPDATE ot1 SET VALUE(ot1.x) = t1(20) WHERE VALUE(ot1.x) = t1(10);

DELETE FROM ot1 WHERE VALUE(ot1) = (t1(10));

The following references, with table aliases, are correct:

UPDATE ot1 o SET o = (t1(20)) WHERE o.x = 10;

DECLARE
  n_ref  REF t1;
BEGIN
  SELECT REF(o) INTO n_ref FROM ot1 o WHERE VALUE(o) = t1(30);
END;
/

DECLARE
  n t1;
BEGIN
  SELECT VALUE(o) INTO n FROM ot1 o WHERE VALUE(o) = t1(30);
END;
/

DECLARE
  n NUMBER;
BEGIN
  SELECT o.x INTO n FROM ot1 o WHERE o.x = 30;
END;
/

DELETE FROM ot1 o WHERE VALUE(o) = (t1(20));