Numeric cubes only in double


A long time ago I received the following advice: only use "double" cubes. Loading values between int and double, or single and double types could be buggy and fail unexpectedly.

Is this still advice we should be following? I have been creating all my numeric cubes as double types, even the matrix cubes that are always only a 1 or a 0. Is it good practice to always use double type?





  • Hello @Krisztina Zappert

    for me this is an horrible advise.

    Please consider the data you have, when you choose the cube type. Sometimes double will be the best fit (cause you need the precission), sometimes it might be single (if 6-9 digits precission is fine), or integer might be the best fit.

    Please consider that this choice has not only an impact on precission, but also on size of DB/Cubes in RAM and on disk



  • Thank you Björn!

    I have a small follow-up question - if let's say I load from a double cube to an integer cube, does it round the values, or just "cut off" anything after the decimal? What does 3,5 or 3,999 become in Board when loaded from double to integer? Apologies if this is in the documentation.



  • Hello Krisztina,

    according to my knowledge board is working like other tools and will round those values



  • Thank you for your time!