Table of Contents


  1. Preface
  2. Using the Designer
  3. Working with Sources
  4. Working with Flat Files
  5. Working with Targets
  6. Mappings
  7. Mapplets
  8. Mapping Parameters and Variables
  9. Working with User-Defined Functions
  10. Using the Debugger
  11. Viewing Data Lineage
  12. Comparing Objects
  13. Managing Business Components
  14. Creating Cubes and Dimensions
  15. Using the Mapping Wizards
  16. Datatype Reference
  17. Configure the Web Browser

Rounding Methods for Double Values

Rounding Methods for Double Values

Due to differences in system run-time libraries and the computer system where the database processes double datatype calculations, the results may not be as expected. The double datatype conforms to the IEEE 794 standard. Changes to database client library, different versions of a database or changes to a system run-time library affect the binary representation of mathematically equivalent values. Also, many system run-time libraries implement the round-to-even or the symmetric arithmetic method. The round-to-even method states that if a number falls midway between the next higher or lower number it round to the nearest value with an even least significant bit. For example, with the round-to-even method, 0.125 is rounded to 0.12. The symmetric arithmetic method rounds the number to next higher digit when the last digit is 5 or greater. For example, with the symmetric arithmetic method 0.125 is rounded to 0.13 and 0.124 is rounded to 0.12.
To provide calculation results that are less susceptible to platform differences, the Integration Service stores the 15 significant digits of double datatype values. For example, if a calculation on Windows returns the number 1234567890.1234567890, and the same calculation on UNIX returns 1234567890.1234569999, the Integration Service converts this number to 1234567890.1234600000.

Updated July 02, 2020