Beefy Boxes and Bandwidth Generously Provided by pair Networks
No such thing as a small change

Re: Rounding off the scale value after decimal point

by rjt (Deacon)
on Sep 17, 2012 at 17:21 UTC ( #994064=note: print w/replies, xml ) Need Help??

in reply to Rounding off the scale value after decimal point

This is a common problem with floating point numbers. In general, you cannot rely on floating point numbers to be "rounded" to any precision unless the rounding is actually part of the program logic or underlying language/library. This happens with whole numbers used as floating points, too. For instance, 39 might become 38.999999999999, which can be a real problem if not handled correctly.

Small differences in rounding occur because floating point numbers are stored in a relatively small region of memory, and the mantissa and exponent are stored as powers of 2, rather than powers of 10, as you see on screen if you just print it out. As such, it's very improbable that a given number in base-10 (decimal) arithmetic will have a base-2 (binary) equivalent that gives the exact same number. The number will be close (yours broke on the seventeenth significant figure), but rarely exact. DBI and SQL Developer are likely clipping off spurious digit(s) from the result, while ASE is not, or it could be differences in machine/data architectures.

Do you actually need 17 significant figures, as your example indicates? Or do you just need the results to be the same across all tools? If it's the former, you will have to rethink your data design to include larger floating point numbers (larger means more precise with floating point). If it's the latter, I would just round the result off to whatever precision you need (and can safely expect to maintain).

  • Comment on Re: Rounding off the scale value after decimal point

Log In?

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://994064]
and all is quiet...

How do I use this? | Other CB clients
Other Users?
Others browsing the Monastery: (3)
As of 2018-06-23 04:46 GMT
Find Nodes?
    Voting Booth?
    Should cpanminus be part of the standard Perl release?

    Results (125 votes). Check out past polls.