When inserted values into float columns in Teradata it is often found the values are approximated after a certain length.Is there any particular pattern followed by
Teradata db for float values when they are stored in database.When I try to export the float columns using an export utility like TPT also I can see this behaviour.
In there any way to avoid this approximation?
Teradata FLOATs use Intel binary floating point format, so they are only as accurate as conversion between decimal and binary fractions. A simple decimal fraction can be a repeating sequence in binary, so the binary float is an approximation. Teradata NUMBER on the other hand stores the exponent and mantissa as decimal numbers. Number also supports more digits. If you need precise values, consider using NUMBER instead of FLOAT.