Re:Steve McConnell
by Ologn on Tuesday December 08, 2020 @05:27PM
(#25352307)
I like Clean Code and Code Complete.
Clean Code has a narrative and is opinionated. It explains the experiences that formed those opinions. It is a better read.
Code Complete is dry, academic, and comprehensive for its scope. It references many studies and is well documented. It has opinions too of a sort. It is a better reference.
Thinking of function length - Clean Code says short functions are good. Code Complete references studies that say it is not long functions that are bad, it is functions using too many objects/variables that are bad. This of course has an indirect effect on function length. So I keep this in mind, and both books are helpful.
I read a bunch of Uncle Bob's books and usually don't mind him being opinionated, although I did think the opinions in Clean Coder got to be too much. Although not all of it. Recently I have been mulling the idea from that book that there should be zero regressions. I think there is something to that, if you start to accept there can be 2-3 regressions per sprint, then why not 5-7? Why not 10? I agree with him that acceptance of regressions is a problem.
Clean Code has a narrative and is opinionated. It explains the experiences that formed those opinions. It is a better read.
Code Complete is dry, academic, and comprehensive for its scope. It references many studies and is well documented. It has opinions too of a sort. It is a better reference.
Thinking of function length - Clean Code says short functions are good. Code Complete references studies that say it is not long functions that are bad, it is functions using too many objects/variables that are bad. This of course has an indirect effect on function length. So I keep this in mind, and both books are helpful.
I read a bunch of Uncle Bob's books and usually don't mind him being opinionated, although I did think the opinions in Clean Coder got to be too much. Although not all of it. Recently I have been mulling the idea from that book that there should be zero regressions. I think there is something to that, if you start to accept there can be 2-3 regressions per sprint, then why not 5-7? Why not 10? I agree with him that acceptance of regressions is a problem.