3.1.2 Precision and Normalization The built-in library uses its own internal precision for whole numbers, which may be the range from -999999999 to +999999999. That is probably far more than you will ever need in the built-in functions. For most functions, neither parameters nor return values will be effected by any setting of NUMERIC. In the few cases where this does not hold, it is explicitly stated in the description of the function. In general, only parameters that are required to be whole numbers are used in the internal precision, while numbers not required to be whole numbers are normalized according to the setting of NUMERIC before use. But of course, if a parameter is a numeric expression, that expression will be calculated and normalized under the settings of NUMERIC before it is given to the function as a parameter.