Open RickMoynihan opened 5 years ago
This was suggested by @RicSwirrl ... cc @Robsteranium to sanity check this is a good idea.
We certainly don't want two observations with the same dimensions but different attributes. The cube IC-12 will take care of that.
Having consistent units per measure seems reasonable but I wonder if there are scenarios where the unit might depend on another dimension (than the measure-type). I can't seem to find it now but IIRC the ABI used to include counts of enterprises and employees. You could imagine modelling this with a single :count
measure and a :statistical-population
dimension which the measurement unit (and sdmxa:statUnit
) would depend upon (instead of the measure).
That said, you could imagine this being a warning (for stuff like sdmxa:unitMult
too as per the original example).
Strictly this might not be a requirement but it's probably worth us adding a validation to ensure that users don't mix units in a dataset within a measureType.
e.g. in a dataset with one measure (counting people); we probably want to check that all
counts
arePeople
(units) and that we don't mixPeople
with something likePeople (thousands)
.