Re:The Next Step for Generics
by kelnos on Tuesday June 16, 2020 @07:30PM
(#23545637)
> I spent more time playing type system golf trying to come up with the optimal type for whatever usecase. In Go I might just "do the work", but in Rust I've turned 5 minute functions into 30 minute api design thought exercises. The jury is out if thats is a good thing or a bad thing.
The way I tend to look at this is the 30 minutes I spend up front making my types work will usually save me hours of time later (sometimes much later, and spread over different coding sessions), because if I'm using the type system to enforce constraints, I'm not going to have to worry about writing buggy constraint code.
When I was learning Scala (after working in Java most of the time), I noticed that I would sometimes spend quite a bit of time getting short snippets of code correct so the types would line up. The result would be shorter than the equivalent Java code. But I had much more confidence in that code actually working, without needing to write tests (or at least as many tests), because I was leaning hard on the type system to prove the code correct. You can do that to some extent in Java, but the type system there is a bit weaker, and the standard library and common idioms make it harder to do. Ultimately if I have a chunk of Java code and a chunk of Scala code that does the same thing, I'm much more confident in the Scala code because I know scalac is providing me stronger guarantees than javac, assuming I've written the code to make good use of the compiler's abilities.
The way I tend to look at this is the 30 minutes I spend up front making my types work will usually save me hours of time later (sometimes much later, and spread over different coding sessions), because if I'm using the type system to enforce constraints, I'm not going to have to worry about writing buggy constraint code.
When I was learning Scala (after working in Java most of the time), I noticed that I would sometimes spend quite a bit of time getting short snippets of code correct so the types would line up. The result would be shorter than the equivalent Java code. But I had much more confidence in that code actually working, without needing to write tests (or at least as many tests), because I was leaning hard on the type system to prove the code correct. You can do that to some extent in Java, but the type system there is a bit weaker, and the standard library and common idioms make it harder to do. Ultimately if I have a chunk of Java code and a chunk of Scala code that does the same thing, I'm much more confident in the Scala code because I know scalac is providing me stronger guarantees than javac, assuming I've written the code to make good use of the compiler's abilities.