Arithmetic Operators

For arithmetic operators, there are binary operators for addition, subtraction, multiplication, or division, and unary operators to represent whether the number is positive or negative. The unary operators to represent the numbers’ positive/negative status have higher priority over the binary operators.

<expression>  <mathematical_operator>  <expression>

    <expression> ::=
        bit_string |
        character_string |
        numeric_value |
        date-time_value |
        collection_value |
        NULL

    <mathematical_operator> ::=
        <set_arithmetic_operator> |
        <arithmetic_operator>

            <arithmetic_operator> ::=
                + |
                - |
                * |
                { / | DIV } |
                { % | MOD }

            <set_arithmetic_operator> ::=
                UNION |
                DIFFERENCE |
                { INTERSECT | INTERSECTION }

• <expression>: Declares the mathematical operation to be calculated.

• <mathematical_operator>: A operator that performs an operation the arithmetic and the set operators are applicable.

  • <set_arithmetic_operator>: A set arithmetic operator that performs operations such as union, difference and intersection on collection type operands.

  • <arithmetic_operator>: An operator to perform the four fundamental arithmetic operations.

The following table shows the arithmetic operators supported by CUBRID and their return values.

Arithmetic Operators

Arithmetic Operator | Description			Operator	Return Value
+ | Addition			1+2	3
-		Subtraction	1-2	-1
*		Multiplication	1*2	2
/		Division. Returns quotient.	1/2.0	0.500000000
DIV		Division. Returns quotient.	1 DIV 2	0
% , MOD		Division. Returns quotient. An operator must be an integer type, and it always returns integer. If an operand is real number, the MOD function can be used.	1 % 2 1 MOD 2	1

Arithmetic Operations and Type Casting of Numeric Data Types

All numeric data types can be used for arithmetic operations. The result type of the operation differs depending on the data types of the operands and the type of the operation. The following table shows the result data types of addition/subtraction/multiplication for each operand type.

Result Data Type by Operand Type

	INT	NUMERIC	FLOAT	DOUBLE
INT	INT or BIGINT	NUMERIC	FLOAT	DOUBLE
NUMERIC	NUMERIC	NUMERIC (p and s are also converted)	DOUBLE	DOUBLE
FLOAT	FLOAT	DOUBLE	FLOAT	DOUBLE
DOUBLE	DOUBLE	DOUBLE	DOUBLE	DOUBLE

Note that the result type of the operation does not change if all operands are of the same data type but type casting occurs exceptionally in division operations. An error occurs when a denominator, i.e. a divisor, is 0.

The following table shows the total number of digits (p) and the number of digits after the decimal point (s) of the operation results when all operands are of the NUMERIC type.

Result of NUMERIC Type Operation

Operation	Maximum Precision	Maximum Scale
N(p1, s1) + N(p2, s2)	max(p1 - s1, p2 - s2) + max(s1, s2) + 1	max(s1, s2)
N(p1, s1) - N(p2, s2)	max(p1 - s1, p2 - s2) + max(s1, s2)	max(s1, s2)
N(p1, s1) * N(p2, s2)	p1 + p2 + 1	s1 + s2
N(p1, s1) / N(p2, s2)	Pt = (p1 - s1) + s2 + max(9, max(s1, s2)) Pt = (Pt > 38) ? 38 : Pt	St = max(9, max(s1, s2)) St = (Pt > 38) ? min(9, St - (Pt - 38)) : St

Example

--int * int
SELECT 123*123;

      123*123
=============
        15129

-- int * int returns overflow error
SELECT (1234567890123*1234567890123);

ERROR: Data overflow on data type bigint.

-- int * numeric returns numeric type
SELECT (1234567890123*CAST(1234567890123 AS NUMERIC(15,2)));

 (1234567890123* cast(1234567890123 as numeric(15,2)))
======================
  1524157875322755800955129.00

-- int * float returns float type
SELECT (1234567890123*CAST(1234567890123 AS FLOAT));

 (1234567890123* cast(1234567890123 as float))
===============================================
                                  1.524158e+024

-- int * double returns double type
SELECT (1234567890123*CAST(1234567890123 AS DOUBLE));

 (1234567890123* cast(1234567890123 as double))
================================================
                          1.524157875322756e+024

-- numeric * numeric returns numeric type
SELECT (CAST(1234567890123 AS NUMERIC(15,2))*CAST(1234567890123 AS NUMERIC(15,2)));

 ( cast(1234567890123 as numeric(15,2))* cast(1234567890123 as numeric(15,2)))
======================
  1524157875322755800955129.0000

-- numeric * float returns double type
SELECT (CAST(1234567890123 AS NUMERIC(15,2))*CAST(1234567890123 AS FLOAT));

 ( cast(1234567890123 as numeric(15,2))* cast(1234567890123 as float))
=======================================================================
                                                 1.524157954716582e+024

-- numeric * double returns double type
SELECT (CAST(1234567890123 AS NUMERIC(15,2))*CAST(1234567890123 AS DOUBLE));

 ( cast(1234567890123 as numeric(15,2))* cast(1234567890123 as double))
========================================================================
                                                  1.524157875322756e+024

-- float * float returns float type
SELECT (CAST(1234567890123 AS FLOAT)*CAST(1234567890123 AS FLOAT));

 ( cast(1234567890123 as float)* cast(1234567890123 as float))
===============================================================
                                                  1.524158e+024

-- float * double returns float type
SELECT (CAST(1234567890123 AS FLOAT)*CAST(1234567890123 AS DOUBLE));

 ( cast(1234567890123 as float)* cast(1234567890123 as double))
================================================================
                                          1.524157954716582e+024

-- double * double returns float type
SELECT (CAST(1234567890123 AS DOUBLE)*CAST(1234567890123 AS DOUBLE));

 ( cast(1234567890123 as double)* cast(1234567890123 as double))
=================================================================
                                           1.524157875322756e+024

-- int / int returns int type without type conversion or rounding
SELECT 100100/100000;

  100100/100000
===============
              1

-- int / int returns int type without type conversion or rounding
SELECT 100100/200200;

  100100/200200
===============
              0

-- int / zero returns error
SELECT 100100/(100100-100100);

ERROR: Attempt to divide by zero.

Arithmetic Operations and Type Casting of DATE/TIME Data Types

If all operands are date/time type, only a subtraction operation is allowed and its return value is BIGINT . Note that the unit of the operation differs depending on the types of the operands. Both addition and subtraction operations are allowed in case of date/time and integer types In this case, operation units and return values are date/time data type.

The following table shows operations allowed for each operand type, and their result types.

Allowable Operation and Result Data Type by Operand Type

	TIME (in seconds)	DATE (in day)	TIMESTAMP (in seconds)	DATETIME (in milliseconds)	INT
TIME	A subtraction is allowed. BIGINT	X	X	X	An addition and a subtraction are allowed. TIME
DATE	X	A subtraction is allowed. BIGINT	A subtraction is allowed. BIGINT	A subtraction is allowed. BIGINT	An addition and a subtraction are allowed. DATE
TIMESTAMP	X	A subtraction is allowed. BIGINT	A subtraction is allowed. BIGINT	A subtraction is allowed. BIGINT	An addition and a subtraction are allowed. TIMESTAMP
DATETIME	X	A subtraction is allowed. BIGINT	A subtraction is allowed. BIGINT	A subtraction is allowed. BIGINT	An addition and a subtraction are allowed. DATETIME
INT	An addition and a subtraction are allowed. TIME	An addition and a subtraction are allowed. DATE	An addition and a subtraction are allowed. TIMESTAMP	An addition and a subtraction are allowed. DATETIME	All operations are allowed.

Note

If any of the date/time arguments contains NULL, NULL is returned.

Example

-- initial systimestamp value
SELECT SYSDATETIME;

  SYSDATETIME
===============================
  07:09:52.115 PM 01/14/2010

-- time type + 10(seconds) returns time type
SELECT (CAST (SYSDATETIME AS TIME) + 10);

 ( cast( SYS_DATETIME  as time)+10)
====================================
  07:10:02 PM

-- date type + 10 (days) returns date type
SELECT (CAST (SYSDATETIME AS DATE) + 10);

 ( cast( SYS_DATETIME  as date)+10)
====================================
  01/24/2010

-- timestamp type + 10(seconds) returns timestamp type
SELECT (CAST (SYSDATETIME AS TIMESTAMP) + 10);

 ( cast( SYS_DATETIME  as timestamp)+10)
=========================================
  07:10:02 PM 01/14/2010

-- systimestamp type + 10(milliseconds) returns systimestamp type
SELECT (SYSDATETIME  + 10);

 ( SYS_DATETIME +10)
===============================
  07:09:52.125 PM 01/14/2010

SELECT DATETIME '09/01/2009 03:30:30.001 pm'- TIMESTAMP '08/31/2009 03:30:30 pm';

 datetime '09/01/2009 03:30:30.001 pm'-timestamp '08/31/2009 03:30:30 pm'
=======================================
  86400001

SELECT TIMESTAMP '09/01/2009 03:30:30 pm'- TIMESTAMP '08/31/2009 03:30:30 pm';

 timestamp '09/01/2009 03:30:30 pm'-timestamp '08/31/2009 03:30:30 pm'
=======================================
  86400

Behavior related to timezone parameters

TIMESTAMP and TIMESTAMP WITH LOCAL TIME ZONE data types stores internally UNIX epoch values (number of secons elapsed from 1970). When leap second is used (tz_leap_second_support is set to yes, see Timezone Parameter), they may contain virtual date-time values.

Virtual date-time       Unix timestamp
2008-12-31 23:59:58  -> 79399951
2008-12-31 23:59:59  -> 79399952
2008-12-31 23:59:60  -> 79399953    -> not real date (introduced by leap second)
2009-01-01 00:00:00  -> 79399954
2009-01-01 00:00:01  -> 79399955

Arithmetic operations with TIMESTAMP and TIMESTAMPLTZ values are performed directly on Unix epoch values. Unix epoch values coresponding to non-exising date/time values are allowed. For this reason, the comparison:

SELECT TIMESTAMPLTZ'2008-12-31 23:59:59 UTC'=TIMESTAMPLTZ'2008-12-31 23:59:59 UTC'+1;

timestampltz '2008-12-31 23:59:59 UTC'=timestampltz '2008-12-31 23:59:59 UTC'+1
=================================================================================
                                                                            0

is equivalent to comparing the Unix timestamps : 79399952 and 79399953. But when same values are used as TIMESTAMPTZ, there is equality:

SELECT TIMESTAMPTZ'2008-12-31 23:59:59 UTC'=TIMESTAMPTZ'2008-12-31 23:59:59 UTC'+1;

timestamptz '2008-12-31 23:59:59 UTC'=timestamptz '2008-12-31 23:59:59 UTC'+1
===============================================================================
                                                                            1

The inconsistency arise at display :

SELECT TIMESTAMPLTZ'2008-12-31 23:59:59 UTC'+1;

timestampltz '2008-12-31 23:59:59 UTC'+1
=============================================
11:59:59 PM 12/31/2008 Etc/UTC UTC

Since ‘2008-12-31 23:59:60 UTC’ corresponding to Unix timestamp value 79399953 is not a real date, the immediately preceding value is used. Internally, it is equivalent to the value (‘2008-12-31 23:59:60 UTC’).

TIMESTAMP WITH TIME ZONE data type contains both a UNIX timestamp and a timezone identifier. Arithmetic on TIMESTAMPTZ is also performed on UNIX timestamp part value, but is followed by an automatic adjusting operation. The presence of timezone identifier (which includes region, offset and daylight saving), requires the TIMESTAMPTZ object to be a valid date-time. The operation timestamptz’2008-12-31 23:59:59 UTC’+1 implies an automatic validation-conversion: instead of (79399953, UTC) which is not a valid date-time the value is automatically converted to (79399952,UTC) which coresponds to ‘2008-12-31 23:59:59 UTC’.

After each arithmetic operation implying DATETIMETZ and TIMESTAMPTZ, CUBRID performs an automatic adjustment of result value which involves:

• adjusting the timezone identifier : adding a number of seconds to a date with timezone may lead to change of internally stored offset rule, daylight saving rule, hence the timezone identifier must be updated

• adjusting the Unix timestamp (only for TIMESTAMPTZ): virtual date-time values (when leap-second is enabled) are always converted to the immediately preceding Unix timestamp value.