Note, we don't know what h, s, and g are but I'm guessing they are numeric vectors of the same size as v in which case, either of these two solutions should work (obviously not tested). If they are not vectors or not the same size as v, it would probably be easy to adapt either of these solutions to their size and class.
Loop method
If you're using a loop, you should only access a single element of v at a time by indexing v(i).
v = -1 :0.1: 1;
for i= 1:numel(v) % replaced length() with numel()
if v(i) >= -0.9 || v(i)<=0.9
d = sqrt((h^2)+(s^2));
elseif v(i)<-0.9 || v(i)>0.9 % this line can be replace with just 'else'
g= sqrt(s^2 + (v(i)-0.9)^2);
d = sqrt(g^2 + h^2);
end
end
disp(d);
Vectorization method
However, you don't need a loop. The method below is more efficient and easier to read. Vectorization is kind of the point of using Matlab over competitors.
v = -1 :0.1: 1;
idx = v <= -0.9 | v >= 0.9;
d = nan(size(v));
d(idx) = (sqrt(s(idx).^2 + (v(idx)-0.9).^2) + h(idx).^2).^2;
d(~idx) = sqrt((h(~idx).^2)+(s(~idx).^2));
