T-SQL has the MIN and MAX aggregate functions. It would be *very* handy to have them for scalars as well.
- They should support all the common data types including uniqueidentifier and bit
- They should take a variable number of arguments
- They should automatically convert types implicitly according to some sensible rules
- They should propagate non-nullness if all arguments are non-null (in case of MAX it is enough if just one argument is guaranteed to be non-null because it will always be greater than NULL)
A scalar UDF is not a useful workaround because they don't inline and are known to be seriously slow.