mapminmax bug with zeroes?

I am encountering a strange issue with mapminmax in R2017b (and previous versions), am I doing this correctly?

This should map the vector [0, 1, 2] to the range of -1 to +1

[a, b] = mapminmax([0, 1, 2], -1, 1);

As expected, a =

    -1     0     1

Now if I try and use apply, this is what I get:

>> mapminmax('apply', 0, b)

ans =

       0

>> mapminmax('apply', 2, b)

ans =

       2

Essentially it fails at mapping the original vector [0, 1, 2]. Now if I make a small change to the vector such that it is [1, 2, 3], it works as expected.

>> [a, b] = mapminmax([1, 2, 3], -1, 1);
>> mapminmax('apply', 1, b)

ans =

      -1

>> mapminmax('apply', 3, b)

ans =

       1

What is going on? Why am I unable to use mapminmax with the original vector [0, 1, 2]?