>>INDEX ON (a * 10^order of magnitude of max value of B+1)+ b +...
>>
>>will result in a number adjusted for relative magnitude; the first has the advantage of being magnitude/precision independent at the cost of not working correctly when negative values are possibly an issue. With very large numbers, rounding inherent in floating point representations can step in and make things mis-sort since the decimal precision of a computation has limits as we've discuissed before.
>>
>>>Thanks
>
>If we were dealing with integers ( instead of numerics ) here, could we use BINTOC() to create the index and would it make it more compact/efficient?
>
BINTOC() would work; you want to expressly name a size for the BINTOC conversion, since a small initial value would result in too small a string, possible with the consequence of truncating large value keys and missorting.