Re:The Next Step for Generics
by eru on Wednesday June 17, 2020 @03:00AM
(#23548590)
That sounds like a problem with Java more than a problem with generics.
In general, I agree with your sentiment. Though in some cases, an abstraction can be justified even when used only once.
Take the simple case of a function. Generally a function only communication with its surroundings via parameters and returned values. A function that respects those conventions makes reading code much easier, and can be worth it even if used only once.
In contrast, a code block inside a much larger stretch of code doesn't give the reader any explicit guidance one how it intends to interact with the rest of the system. Even if it's exactly the same code as in the function.
Similarly, a for-loop can do almost anything. But using map or filter immediately tells you that only a few things are possible here. Even a fold is more restricted than a general for-loop.
In general, I agree with your sentiment. Though in some cases, an abstraction can be justified even when used only once.
Take the simple case of a function. Generally a function only communication with its surroundings via parameters and returned values. A function that respects those conventions makes reading code much easier, and can be worth it even if used only once.
In contrast, a code block inside a much larger stretch of code doesn't give the reader any explicit guidance one how it intends to interact with the rest of the system. Even if it's exactly the same code as in the function.
Similarly, a for-loop can do almost anything. But using map or filter immediately tells you that only a few things are possible here. Even a fold is more restricted than a general for-loop.
'Theorems for free' makes a very similar argument for generics. https://people.mpi-sws.org/~dreyer/tor/papers/wadler.pdf