Prove without induction that $$\sum_{r=1}^n {n\choose r}(-1)^{r+1}\dfrac{1}{r}=\sum_{r=1}^n \dfrac{1}{r}$$ for $n\geq1$.
I have shown the above using induction. However, I am interested in knowing if we can show it directly without using induction.
This is not entirely trivial, I think, if one is armed only with the very basic ideas of combinatorial identities. This can be proved in at least two different ways: one using induction and the other is by evaluating an integral. However, regarding the second approach, I am quite certain that if not known beforehand which integral to compute to get this equality, it is a hard exercise.
EDIT: In the link claiming to be a duplicate of my question, induction was allowed. I am looking for approaches that use no induction or calculus. So this question is different.
Start by
$$\sum_{r=0}^n {n\choose r}x^r=(1+x)^n$$
Which can be converted to integration
$$\sum_{r=1}^n {n\choose r}\frac{(-1)^{r}}{r}=\int^{-1}_0 \frac{(x+1)^n-1}{x} dx$$
By substitution we have
$$\sum_{r=1}^n {n\choose r}\frac{(-1)^{r+1}}{r}=\int^{1}_0 \frac{t^n-1}{t-1} dt = H_n$$
Note the last step by expanding $(1-t)^{-1}$.