[R] tests for NULL objects

[Charilaos Skiadas]

On Nov 29, 2006, at 7:58 PM, Benilton Carvalho wrote:

> But that's okay, just a matter of adding an extra test (already done).
>
I'm guessing it would be something like: all(is.na(v) | v==2)
> When I asked for clarification about the reasons for this, I assumed
> that:
> if all(v) is TRUE ==> any(v) is TRUE; for all " (logical) v"...
> When actually:
> if all(v) is TRUE ==> any(v) is TRUE; for all "(logical) v of length
>> = 1".

I like to think of it as: in order for the statement "all elements of  
v are TRUE" to be false, one would need to provide an element of v  
which is false. Since that is not possible if v has no elements, then  
the statement "all elements of v are TRUE" can't be false, hence has  
to be true.

Similarly, the statement "any(v) is TRUE" translates to "there is an  
element of v that is true".  Since there are no elements in v, this  
can't possibly happen. So this statement is false.

As for the sums and products, the explanation that Martin linked to  
sums it up nicely. Basically, suppose we have two vectors, v,w, and v  
is actually numeric(0). Then sum(c(v,w)) ought to be sum(v) + sum(w)  
(since that would be the case for any other "honest" vectors). But  
since c(v,w) is really w, the only way this will happen is if sum(v)  
= 0. Same for products, since in that case prod(c(v,w))=prod(v) times  
prod(w).

> Thank you,
>
> benilton

Haris