Null inference

[NicoLaval]

As null as no type in the current version, I didn't find in specification that we are allowed to infer null type.

I think a := 1 + null; is not null but raised en error and we have to define a := 1 + cast(null, integer);

How can we improve?

• defining a null type?
• enable null type inference (would it be possible in each cases?)?

[linardian]

Please refer to the paragraph "Scalar types" page 52 the "default" assumptions ("All the scalar types are assumed by default to contain the conventional value “NULL”, which means “no value”, or “absence of known value” or “missing value” (in other words, the scalar types by default are “nullable”). Note that the “NULL” value, therefore, is the only value of multiple different types (i.e., all the nullable scalar types).definition of "Null" value and the behaviour of the different operators.") and the paragraph "Behaviour for Missing data", page 91 of the VTL User Manual - Version 2.1 for the behaviour of the different operators dealing with null values. We have already discussed about he "reshuffled" asset of the future versions of documentation since some chapters/paragraph of the User Manual could be inserted in the Reference Manual and viceversa.

[vpinna80]

I'd like to point out that "null" is not a valid VTL literal, for good reasons IMHO.
Allowing "null" literals with the new implicit castings would open the way to any sort of shenanigans.

[linardian]

If the corresponding #455 is going to closed, I propose to close this one too. @NicoLaval please let me know.