My question refers to following conclusion in Liu's "Algebraic Geometry" at page 290 (Thm 4.16):
Why does $e'_x= length_{\hat{B}}(W_{\hat{B}/\hat{A}}/\hat{B})$ hold?
Here $e'_x$ is introduced as $e'_x = length_{\mathcal{O}_{X,x}}(\omega_f/ \mathcal{O}_X)_x +1$.
As the excerps explains we get $(\omega_f/ \mathcal{O}_X)_x = Hom_{\hat{A}}(\hat{B},\hat{A})/\hat{B}$ and Exercise 6.4.8 provides the identification: $Hom_{\hat{A}}(\hat{B},\hat{A}) \cong W_{\hat{B}/\hat{A}}$, as see below:
Combine this both results we get:
$e'_x = length_{\mathcal{O}_{X,x}}(W_{\hat{B}/\hat{A}}/\hat{B}) +1$.
But how to concude the final step $length_{\mathcal{O}_{X,x}}(W_{\hat{B}/\hat{A}}/\hat{B}) +1= length_{\hat{B}}(W_{\hat{B}/\hat{A}}/\hat{B})$; especially how to cope the "change" of the "base modules" from $\mathcal{O}_{X,x}$ to $\hat{B}$ under contolling the $length$ of modules?

