
Archives of the TeradataForumMessage Posted: Thu, 10 Apr 2003 @ 10:49:45 GMT
The answer to your question depends on the purpose of such a large number. If you're performing arithmetic operations, you could use FLOAT  but keep in mind that floating point numbers have a varying degree of accuracy, depending on the scale of the number. It's probably better not to think of a FLOAT as a large number, but rather an approximation of a number. For example, you might say you're going to add a number to a number. If you're using FLOAT, you should be saying that you're adding an approximation of a number to the approximation of another number with a result of unknown accuracy (it depends on the accuracy of the approximations). Keep in mind that the lost of precision is accumulative as you perform arithmetic operations. In scientific/engineering applications, the numbers are so large that the lost of some accuracy just isn't important. You can't tolerate this kind of inaccuracy in something like a financial application. The use of FLOAT data types is generally prohibited in applications which must be auditable. If accuracy is an issue, then don't use FLOAT. Some basic information can be found at: If you aren't performing arithmetic operations, then you might consider storing it as a CHAR. If you want to perform arithmetic operations on such a large number and need accuracy, then you can't do it on the Teradata. That said, I would question the reality of specifying such a large number in the first place  it's been my experience that when somebody specifies such a large number it's because they don't know the metrics of their data and are afraid of overflows, so they replace design with overkill. Then again, there's always the exception... If you can give some more detail on your situation, maybe a better answer could be found. Admin Comment: When last tried, the URL in this Post no longer functioned. Use the following URL instead: www2.hursley.ibm.com...
 
 
Copyright 2016  All Rights Reserved  
Last Modified: 27 Dec 2016  