From Nicholas C. Zakas’s regularly interesting newsletter:
The best way to work with legacy code is to approach it patiently—understand small parts, write tests, and document what you learn.
My favorite line about legacy code, which I’ve probably shared too much, is when David Khourshid called it “Legendary Code”. Legacy code feels a smidge mean for code that’s been around a long time, doing exactly what it should.
I do wonder, in this age of AI, how it affects our desire to rid our code bases of “legacy” code. Will we be too quick to do it? Or is AI just what we need to make sure the new code is rigorously tested in a way that many of us are too lazy to do manually?
