So along came Oracle 8, and I had to design a data model. All the documentation says that for any integer, you should use Number(38). According to the same documentation, this is allows for the storage of up to 38 digits, with decimal point anywhere. Now, I know that this will probably work, but it doesn't sound like a terribly optimal way of going about storing a 32-bit Long Integer.
Does anyone know anything about how Oracle
handles Number fields internally? Is there anything to be gained by using,
say Number(10) over Number(38), and ignoring the documentation?
Number(10) can't provide better performance
than number(38). Rather it's a ristriction
on the column.
As for internal format, for example, number
1234 will be store in a format like 1.234x10^3. You can refer Oracle server
concepts for detail info.