Recent searches
Search options
What is your opinion on (conditional) branching in your code?
Conditional branching includes things like (non-trivial) if statements, for statements, etc.
Also, for those of you who haven't heard about this, even though conditional branching is a programming fundamental, it's not great for CPU performance (though a lot of work has gone into improving this to the point where you typically don't have to worry about this with modern CPU hardware and compilers) and there are programming styles which reduce or eschew branching, either for style reasons or to help with performance.
For instance, if you've heard that you should use a switch statement instead of if statements, that's to cut down on the number of branches for performance reasons. I think the switch is itself a conditional branch, but it's a single one regardless of how many conditions you have.
@urusan Where it's possible I like writing "straight line" code that only jumps forward. When it does this, the program can't enter an infinite loop, and it has bounded execution time.
@urusan
do you mean things like APL/numpy-like array programming? or something else entirely?
@niplav There are a lot of different problem domains and approaches under this umbrella.
I don't know enough about APL to comment on it. As for numpy, you could definitely reframe certain problems that would normally involve lots of branching into a bunch of matrix calculations without branching.
@urusan if you can read your own code 6 months later then it's probably fine
@urusan "Premature optimization is the root of all evil" -- Donald Knuth
Worrying about overhead from branching is almost always wasted effort, especially if you haven't profiled the code to find what the true bottlenecks are.
Obviously, there are exceptions to this ("it depends" is always the correct answer in CS) but you should never worry about something like that until you have to.
@urusan I used to be okay with branching, but then I discovered pattern matching. Oh how I wish ECMAScript had pattern matching standard…