Re: Storage of decimal

On Mon, 12 Jan 1998 17:32:04 GMT, manel@semic.es (Manel Falc') wrote: >Hi, >Why the storage of decimal(14,4) is 8 bytes and decimal(14,3) or >decimal(14,5) is 9 ?? >( 8 for even decimals and 9 for odd decimals, it seems) >according to the storage calculation: 14/2+1 -> 8 bytes > > >Ciao, > > Manel > >----------------------------------------------------- >SEMIC, S.A. >Lleida (Catalunya) >Spain >----------------------------------------------------- >Informix version: Informix SE 5.07 / RDS 4.x >Operating system: Unix SCO V 5.04 > Your calculation is too simple, unfortunately. DECIMAL values are stored in base-100 units, so for each pair (or part-pair) of digits before or after the decimal point requires 1 byte, plus 1 byte for exponent and sign information. For DECIMAL(14,4), the are 2 base-100 digits after the dp, 5 before, and 1 extra, for 8 bytes overall. For DECIMAL(14,5), there are 3 base-100 digits after the dp, and 5 before, plus the sign/exponent byte for a total of 9. The last digit after the dp is always a multiple of 10, of course. I think the current generation of manuals has this formula correct. For example, the 9.1 Informix Guide to SQL: Reference specifies: If the scale is odd: N = (precision + 4) / 2 If the scale is even: N = (precision + 3) / 2 Hence: For odd scales and precision = 14, we get (14 + 4) / 2 = 9 For even scales and precision = 14, we get (14 + 3) / 2 = 8 It also points out that you can't have a DECIMAL(32,31) type. Yours, Jonathan Leffler (johnl@informix.com) #include <witticism.h> Ciao, Manel ----------------------------------------------------- SEMIC, S.A. Lleida (Catalunya) Spain ----------------------------------------------------- Informix version: Informix SE 5.07 / RDS 4.x Operating system: Unix SCO V 5.04