I think this is an interesting question:
In the 2-dimensional real vector space, we are given a linear transformation $f$. Suppose we already know the images of the standard bases, say $f(e_1),f(e_2)$. Of course, in some cases the eigenspace might not exist. But assuming the eigenspaces exist, how can we use ruler and compasses to determine them ?
Ok,thas was normal way but corcret! what is your idea about this: $f(e_1)=(f(e_{1}).(1,0)) (1,0)+(f({e_1}).(0,1) )(0,1)=f_{1}(e_1)(1,o)+f_{2}(e_1)(0,1)$ $f(e_2)=(f(e_{2}).(1,0)) (1,0)+(f({e_2}).(0,1) )(0,1)=f_{1}(e_2)(1,o)+f_{2}(e_2)(0,1)$ so matrix of f at standara basis will be : $A= \left[\begin{matrix} f_1 ({e_1}) & f_1 (e_2)\\ f_2 (e_1) & f_2 (e_2) \end{matrix}\right]$ and so we can find eigen value from this relation: det ($A= \left[\begin{matrix} f_1 ({e_1}) & f_1 (e_2)\\ f_2 (e_1) & f_2 (e_2) \end{matrix}\right]$ -$\lambda I$)=0 for this we must solve this relation : $\lambda^{2}+\lambda(-f_{1}(e_1)-f_2(e_2))+f_{1}(e_1)f_2(e_{2})-f_{2}(e_1)f_{1}(e_2)=0$ now if we can find a good formula for $\lambda$ then we can find a good formula for it's eigen vector too.