I second this blog. The comparison of (10 different!) MIP modeling techniques for piecewise linear functions is more complete than any I’ve seen.