To avoid numeric overflow conditions, the C or C++ function should define a decimal data type as big as it can handle.
If the assignment of the value of an input or output numeric argument would result in a loss of significant digits, a numeric overflow error is reported.
For example, consider a scalar function that takes a DECIMAL(2,0) argument:
CREATE FUNCTION UDF_SMLDEC( p1 DECIMAL(2,0) ) RETURNS DECIMAL(2,0) LANGUAGE C NO SQL PARAMETER STYLE TD_GENERAL EXTERNAL;
Passing a number with a maximum of two digits is successful:
An attempt to pass a number larger than 99 or smaller than -99 would result in a loss of significant digits.
SELECT UDF_SMLDEC(100); Failure 2616 Numeric overflow occurred during computation.
Any fractional numeric data that is passed or returned that does not fit as it is being assigned is rounded according to the Teradata rounding rules. For more information on rounding, see “Decimal/Numeric Data Types" in Teradata Vantage™ - Data Types and Literals, B035-1143.