Hi, There was some discussion earlier about setting TypeDescriptorImpl's collationDerivation field to StringDataValue.COLLATION_DERIVATION_NONE(which in SQL Spec language means none). I can dig that mail out if anyone needs it for the background information.
While working on language based ordering, what I am finding is all the TypeDescriptors related to character data types should have the collation derivation of implicit with the exception of TypeDescriptors corresponding to aggregate methods (as mentioned on the wiki page http://wiki.apache.org/db-derby/BuiltInLanguageBasedOrderingDERBY-1478 Section Collation Determination point 11.) So, my question is wouldn't it be better for us to start out with collationDerivation of implicit and make it the job of aggregate methods to set the collation to none if they are dealing with operands of different collation types? Pros of setting collation derivation to implicit 1)Less code change will be required because majority of the places where the TypeDescriptorImpl is created for character types, we want the collation derivation to be implicit. Cons of setting collation derivation to implicit 1)We shouldn't miss places where the collation derivation should be none rather than the default of implicit. Based on this, does anyone have objections to me changing the default of TypeDescriptorImpl's collationDerivation field to implicit? Thanks, Mamta
