3 Expressions

Expression specification

An expression specifies a value to be used in a SQL operation.

An expression can consist of a primary or several primaries connected by arithmetic operators, comparison operators, string or binary operators, bit operators or any of the functions described in Chapter 4, "Functions". A primary is a signed or unsigned value derived from one of the items listed in the SQL syntax.

SQL syntax

{ColumnName | ROWID | {? | :DynamicParameter} |
  Function | Constant | (Expression)}

or

[[+ |-] {ColumnName | SYSDATE | TT_SYSDATE|GETDATE() |
{? | :DynamicParameter} | Function |
Constant | {~ | + | -} Expression}]
 [...]

or

Expression1 [& | | | ^ | + | / | * | - ] Expression2

or

Expression1 | | Expression2

or

Expression

Description

• Arithmetic operators can be used between numeric values. See "Numeric data types".

• Arithmetic operators can also be used between datetime values and interval types. The result of a datetime expression is either a datetime data type or an interval data type.

• Arithmetic operators cannot be applied to string values.

• Elements in an expression are evaluated in the following order:

  • Functions and expressions in parentheses

  • Unary pluses and minuses

  • The * and / operations

  • The + and – operations

  • Elements of equal precedence are evaluated in left-to-right order

• You can enclose expressions in parentheses to control the order of their evaluation. For example:

  10 * 2 – 1 = 19 but 10 * (2 – 1) = 10
  

• Type conversion, truncation, underflow, or overflow can occur when some expressions are evaluated. See Chapter 1, "Data Types".

• If either operand in a numeric expression is NULL, the result is NULL.

• Since NVL takes two parameters, both designated as an "expression", TimesTen does not permit NULL in either position. If there is a NULL value in an expression, comparison operators and other predicates evaluate to NULL. See Chapter 5, "Search Conditions" for more information on evaluation of comparison operators and predicates containing NULL values. TimesTen permits inserting NULL, but in general INSERT takes only specific values, and not general expressions.

• The query optimizer and execution engine permit multiple rowid lookups when a predicate specifies a disjunct of rowid equalities or uses IN. For example, multiple fast rowid lookups are executed for:

  WHERE ROWID = :v1 OR ROWID = :v2
  

• or equivalently:

  WHERE ROWID IN (:v1, :v2)
  

• The ? or :DynamicParameter can be used as a dynamic parameter in an expression.

Examples

This example shows a dynamic parameter in the WHERE clause of any SELECT statement:

SELECT * FROM purchasing.orders
  WHERE partnumber = ? AND ordernumber > ?
  ORDER BY ordernumber;

This example shows a dynamic parameter in the WHERE and SET clauses of an UPDATE statement:

UPDATE purchasing.parts
  SET salesprice = :dynamicparameter1
  WHERE partnumber = :dynamicparameter2;

This example shows a dynamic parameter in the WHERE clause of a DELETE statement:

DELETE FROM purchasing.orderitems
  WHERE itemnumber BETWEEN ? AND ?;

This example shows a dynamic parameter in the VALUES clause of an INSERT statement. In this example, both ? and :dynamicparameter are used where :dynamicparameter1 corresponds to both the second and fourth columns of the purchasing.orderitems table. Therefore, only four distinct dynamic parameters need to be passed to this expression with the second parameter used for both the second and fourth columns.

INSERT INTO purchasing.orderitems VALUES  
   (?,:dynamicparameter1,
      :dynamicparameter2,
      :dynamicparameter1,?);

This example demonstrates that both ? and :dynamicparameter can be used in the same SQL statement and shows the semantic difference between repeating both types of dynamic parameters.

Examples of bitwise operators:

Command> SELECT 0x183D & 0x00FF FROM dual;
< 003D >
1 row found.
Command> SELECT ~255 FROM dual;
< -256 >
1 row found.
Command> SELECT 0x08 | 0x0F FROM dual;
< 0F >
1 row found.

Subqueries

TimesTen supports subqueries in INSERT...SELECT, CREATE VIEW or UPDATE statements and in the SET clause of an UPDATE statement, in a search condition and as a derived table. TimesTen supports table subqueries and scalar subqueries. TimesTen does not support row subqueries. A subquery can specify an aggregate with a HAVING clause or joined table. It can also be correlated.

SQL syntax

[NOT] EXISTS | [NOT] IN (Subquery)
Expression {= | <> | > | >= | < | <= } [ANY | ALL] (Subquery)
Expression [NOT] IN (ValueList | Subquery)

Where ValueList is a list of constant expressions. Each constant expression specifies a constant value or an expression that evaluates to a constant value (such as a number, character string, or date). This includes support for bound values (? or :DynamicParameter), NULL, and calls to functions that return constant values.

Description

TimesTen supports queries with the characteristics listed in each section.

Table subqueries

• A subquery can appear in the WHERE clause or HAVING clause of any statement except one that creates a materialized view. Only one table subquery can be specified in a predicate. These predicates can be specified in a WHERE or HAVING clause, an OR expression within a WHERE or HAVING clause, or an ON clause of a joined table. They cannot be specified in a CASE expression, a materialized view, or a HAVING clause that uses the + operator for outer joins.

• A subquery can be specified in an EXISTS or NOT EXISTS predicate, a quantified predicate with ANY or ALL, or a comparison predicate. The allowed operators for both comparison and quantified predicates are: =, <, >, <=, >=, <>. The subquery cannot be connected to the outer query through a UNIQUE or NOT UNIQUE operator.

• Only one subquery can be specified in a quantified or comparison predicate. Specify the subquery as either the right operand or the left operand of the predicate, but not both.

• The subquery should not have an ORDER BY clause.

• FIRST NumRows is not supported in subquery statements.

• In a query specified in a quantified or comparison predicate, the underlying SELECT must have a single expression in the select list. In a query specified in a comparison predicate, if the underlying select returns a single row, the return value is the select result. If the underlying select returns no row, the return value is NULL. It is an error if the subquery returns multiple rows.

Scalar subqueries

A scalar subquery returns a single value.

• A nonverifiable scalar subquery has a predicate such that the optimizer cannot detect at compile time that the subquery returns at most one row for each row of the outer query. The subquery cannot be specified in an OR expression.

Examples

Examples of supported subqueries for a list of customers having at least one unshipped order:

SELECT customers.name FROM customers
  WHERE EXISTS (SELECT 1 FROM orders
    WHERE customers.id = orders.custid
    AND orders.status = 'unshipped');

SELECT customers.name FROM customers
  WHERE customers.id = ANY 
    (SELECT orders.custid FROM orders
  WHERE orders.status = 'unshipped');

SELECT customers.name FROM customers 
  WHERE customers.id IN 
    (SELECT orders.custid FROM orders 
  WHERE orders.status = 'unshipped');

In this example, list items are shipped on the same date as when they are ordered:

SELECT line_items.id FROM line_items 
  WHERE line_items.ship_date = 
    (SELECT orders.order_date FROM orders 
  WHERE orders.id = line_items.order_id);

Constants

A constant is a literal value.

SQL syntax

{IntegerValue | FloatValue |FloatingPointLiteral|
    FixedPointValue | 'CharacterString'|
    'NationalCharacterString' | HexadecimalLiteral |
    'DateString' | DateLiteral |'TimeString' | 
    TimeLiteral | 'TimestampString' | TimestampLiteral |
    IntervalLiteral | BINARY_FLOAT_INFINITY |
    BINARY_DOUBLE_INFINITY | -BINARY_FLOAT_INFINITY |
    -BINARY_DOUBLE_INFINITY | BINARY_FLOAT_NAN |
     BINARY_DOUBLE_NAN
}

N'Here''s how!'

N'This is an \u0061'

N'This is an a'

0xFFFAB0880088343330FFAA7

0x000A001231

Command> insert into tabvb values (0x1234);
1 row inserted.
Command> insert into tabvb values ('1234');
1 row inserted.
Command> select colbin from tabvb;
< 1234 >
< 31323334 >
2 rows found.

'2007-01-27 12:00:00'

'2007-01-27'

DATE '2007-01-27' or DATE '2007-01-27 12:00:00'

TT_DATE '2007-01-27'. 

{d '2007-01-27'}. 

'20:25:30'

TIME '20:25:30'

INSERT INTO timetable VALUES (TIME '10:00:00');

SELECT * FROM timetable WHERE col1 < TIME '10:00:00';

{t '12:00:00'}

SELECT * FROM testable WHERE C1 = TIME '12:00:00'

TIMESTAMP '2007-01-27 11:00:00.000000'

{ts '9999-12-31 12:00:00'}

Format models

A format model is a character literal that describes the format of datetime and numeric data stored in a character string. When you convert a character string into a date or number, a format model determines how TimesTen interprets the string.

This section covers the following format models:

• Number format models

• Datetime format models

• Format model for ROUND and TRUNC date functions

• Format model for TO_CHAR of TimesTen datetime data types

Number format models

Use number format models in the following functions:

• In the TO_CHAR function to translate a value of NUMBER, BINARY_FLOAT, or BINARY_DOUBLE data type to VARCHAR2 data type.

• In the TO_NUMBER function to translate a value of CHAR or VARCHAR2 data type to NUMBER data type.

The default american_america NLS language and territory setting is used.

A number format model is composed of one or more number format elements. The table lists the elements of a number format model. Negative return values automatically contain a leading negative sign and positive values automatically contain a leading space unless the format model contains the MI, S, or PR format element.

Datetime format models

Use datetime format models in the following functions:

• In the TO_CHAR or TO_DATE functions to translate a character value that is in a format other than the default format for a datetime value.

• In the TO_CHAR function to translate a datetime value that is in a format other than the default format into a string.

The total length of a datetime format model cannot exceed 22 characters.

The default american_america NLS language and territory setting is used.

A datetime format model is composed of one or more datetime format elements, which are shown in Table 3-2.

Format model for ROUND and TRUNC date functions

The table lists the format models you can use with the ROUND and TRUNC date functions and the units to which they round and truncate dates. The default model DD returns the date rounded or truncated to the day with a time of midnight.

Format model for TO_CHAR of TimesTen datetime data types

Use this format model when invoking the TO_CHAR function to convert a datetime value of TT_TIMESTAMP or TT_DATE. In addition, use this format model when invoking the TO_CHAR function to convert any numeric value other than NUMBER or ORA_FLOAT.

• If a numeric value does not fit in the specified format, TimesTen truncates the value.

• The format string cannot exceed 50 characters.

• D always results in a decimal point. Its value cannot be changed with an NLS parameter.

• If a float with an absolute value less than 1e-126 or greater than 1e126 is specified as input to the TO_CHAR function, TimesTen returns an error.

CASE expressions

Specifies a conditional value. Both simple and searched case expressions are supported. The CASE expression can be specified anywhere an expression can be specified and can be used as often as needed.

Instead of using a series of IF statements, the CASE expression enables you to use a series of conditions that return the appropriate values when the conditions are met. With CASE, you can simplify queries and write more efficient code.

SQL syntax

The syntax for a searched CASE expression is:

CASE
  {WHEN SearchCondition THEN Expression1}[…] 
   [ELSE Expression2]
END

The syntax for a simple CASE expression is:

CASE Expression 
  {WHEN CompExpression THEN Expression1}[…]
   [ELSE Expression2] 
END

Parameters

CASE has the parameters:

Description

You cannot specify the CASE expression in the value clause of an INSERT statement.

Examples

To specify a searched CASE statement that specifies the value of a color, use:

SELECT CASE 
  WHEN color=1 THEN 'red' 
  WHEN color=2 THEN 'blue' 
  ELSE 'yellow' 
END FROM cars;

To specify a simple CASE statement that specifies the value of a color, use the following.

SELECT CASE color 
  WHEN 1 THEN 'red' 
  WHEN 2 THEN 'blue' 
  ELSE 'yellow' 
END FROM cars;

ROWID

TimesTen assigns a unique ID called a rowid to each row stored in a table. The rowid has data type ROWID. You can examine a rowid by querying the ROWID pseudocolumn.

Because the ROWID pseudocolumn is not a real column, it does not require database space and cannot be updated, indexed or dropped.

The rowid value persists throughout the life of the table row, but the system can reassign the rowid to a different row after the original row is deleted. Zero is not a valid value for a rowid.

Rowids persists through recovery, backup and restore operations. They do not persist through replication, ttMigrate or ttBulkCp operations.

See "Expression specification" for more information on rowids. See "ROWID data type" for more information about the ROWID data type.

ROWNUM psuedocolumn

For each row returned by a query, the ROWNUM pseudocolumn returns a number indicating the order in which the row was selected. The first row selected has a ROWNUM of 1, the second a ROWNUM of 2, and so on.

Use ROWNUM to limit the number of rows returned by a query as in this example:

SELECT * FROM employees WHERE ROWNUM < 10;

The order in which rows are selected depends on the index used and the join order. If you specify an ORDER BY clause, ROWNUM is assigned before sorting. However, the presence of the ORDER BY clause may change the index used and the join order. If the order of selected rows changes, the ROWNUM value associated with each selected row could also change.

For example, the following query may return a different set of employees than the preceding query if a different index is used:

SELECT * FROM employees WHERE ROWNUM < 10 ORDER BY last_name;

Conditions testing for ROWNUM values greater than a positive integer are always false. For example, the following query returns no rows:

SELECT * FROM employees WHERE ROWNUM > 1;

Use ROWNUM to assign unique values to each row of a table. For example:

UPDATE my_table SET column1 = ROWNUM;

If your query contains either FIRST NumRows or ROWS m TO n, do not use ROWNUM to restrict the number of rows returned. For example, the following query results in an error message:

SELECT FIRST 2 * FROM employees WHERE ROWNUM <1 ORDER BY employee_id;
2974: Using rownum to restrict number of rows returned cannot be combined with 
first N or rows M to N