On Jul 6, 2013, at 2:20 PM, Luis Carvalho <firstname.lastname@example.org> wrote:
Interesting* NaN would actually work as a potential 'empty' then for my
Yes, this went through my mind when I saw the suggestion. SQL support for NaN is all over the place, but in my particular case I would regard a NaN in a database as illegal, and the fact that it would be interpreted as NULL is ok, though not perfect.
The R NA/NULL thing is certainly a confusing concept for many, and undoubtedly 'empty' carries similar dangers, though my original intent was actually much simpler (for example, no ternary logic, can be used as a table key etc.)