Thanks for your answer. However, your approximation is not useful since it still involves one of the original variable, i.e. y. I need an approximation where, for x->0 and y->0, the original function becomes f(t).
As you have a function f(x,y)=(1-x)/(1-y) of 2 variables, it is not possible in this case to express the function only using t=x/y. What you want it is not possible in the general case. Indeed, the knowledge of t=x/y is not sufficient to know f. Suppose t=10, so x=10y. If y=0.1, x=1, so f=0. If y=0.01, x=0.1, so f=0.9/0.99=90/99=10/11=0.90909... If y decreases, f becomes closer to 1.