Assume $f$ is a function from $\mathcal{R}^n$ to $\mathcal{R}$, $f$ is a convex and smooth function, can we express divergence theorem in this way?
$$\nabla\int\limits_{B_\delta}f(x+v)dv = \int\limits_{S_\delta}f(x+u)\frac{u}{\|u\|}du$$
where $B_\delta = \{y \mid \|y\|\leq \delta\}$ and $S_\delta = \{y \mid \|y\|= \delta\}$.
This is taken from equation (6.5) from page 109 of the book http://ocobook.cs.princeton.edu/OCObook.pdf Although it states that it is from Stokes theorem, but I believe it is actually divergence theorem.
Divergence theorem is usually used on the vector field, i.e., $\vec{f}$, can we generalize to the case of $f$ defined here? What about the dot product of $\langle\vec{f}, u\rangle$ in the original divergence theorem, can we simply state it in the above way? Why the gradient operator in the above equation is moved outside?