Let $f \colon [0,2\pi ) \to \{ (x,y) \in \mathbb{R}^2 \mid x^2 + y^2 = 1\}$ be defined as $$f(t) = ( \cos t , \sin t) \ \ \mbox{ for all } \ t \in [0, 2\pi).$$ Then the map $f$ is bijective and continuous, as asserted by Rudin. But how do we rigorously show that the inverse map is not continuous at the point $(1,0) = f(0)$?
Intuitively, it is obvious since if we approach the point $(1,0)$ along that part of the unit circle which lies in the fourth quadrant, we can make the distance in $\mathbb{R}^2$ between $f(t)$ and $f(0)$ as small as we please, but the distance in $\mathbb{R}$ between $t= f^{-1}\left(f(t)\right)$ and $0= f^{-1}\left(f(0)\right) = f^{-1}\left( (1,0) \right)$ approaches $2\pi$.
But how to formulate this into a rigorous, $\epsilon$-$\delta$ argument?
Approach the point $(1,0)$ from above and below (in the circle). The limits of $f^{-1}$ are $0$ and $2\pi$ respectively.
Take $\epsilon=1$ then for every $\delta>0$ take a point $a=(x,\sqrt{1-x^2})$ and $b=(x,-\sqrt{1-x^2})$ such that $d(a,(1,0))<\delta$ and $d(b,(1,0))<\delta$.
But $f^{-1}(a)=\arctan(\frac{\sqrt{1-x^2}}{x})$ and $f^{-1}(b)=2\pi-\arctan(-\frac{\sqrt{1-x^2}}{x})=2\pi+\arctan{\frac{\sqrt{1-x^2}}{x}}$. Then $|f^{-1}(b)-f^{-1}(a)|=2\pi>\epsilon$.