3 ms·

In my experience: Far less up-front specification, far fewer names/concepts, fewer lines of code, far fewer of those "oh, this should work, but how do I wrestle

by olau 8y ago

In my experience: Far less up-front specification, far fewer names/concepts, fewer lines of code, far fewer of those "oh, this should work, but how do I wrestle the type system into understanding it?". Now, there's a continuum here. When designing the main data store for a project, you want to iterate on that and get it just right. Or if you're working on a library to be used by lots of people, it's definitely nice to be able to specify things that lets the compiler help people get their code right. On the other hand, in almost all the applications I've worked on, there's also a much larger fraction of absolutely necessary and often complex code where each little module is working on some details. There the data structure doesn't really matter so what JSON provides (untyped arrays + simple objects/dictionaries) results in much less cognitive overhead than a fully specified and typed thing. A few black-box tests both documents and checks functionality, and also ensures that the thing keeps working after a refactor. Of course, YMMV.

rstuart4133 8y ago
Surely the most common argument was that dynamically typed languages need less lines of code to do the same thing, combine that with the time it takes to write code is roughly proportional to the size in lines, and the size of the unit tests similarly proportional, and it becomes a pretty hard argument to refute. For a long while you were comparing languages line Python to something like Java, and it was pretty much a no brainer - the argument was right. But nowadays if you compare something like Javascript to Typescript, it's not near so clear. Type inference is really biting into the size advantage line dynamic typing had over static typing. The other thing is the claim 1 line of statically typed code of takes as long to write as 1 line of dynamically typed code. It's true as stated, but change it to "correct line of code" and it's not true. You don't have to write tests for type declarations, and statically typed code has less bugs after it's accepted by the compiler / interpreter, so while the writing takes the same amount of time the testing and fixing takes less for statically typed languages. The final advantage of dynamically typed languages is statically typed languages just need more head space - just ask someone who has used Rust. Granted Rust with it's explosion of types caused by lifetimes is a bit of an extreme case, but it's still true for most languages. But it's not so true for the Typescript vs Javascript comparison, and I suspect really careful standard library design might make this mostly go away too.

by olau 8y ago

rstuart4133 8y ago

Surely the most common argument was that dynamically typed languages need less lines of code to do the same thing, combine that with the time it takes to write code is roughly proportional to the size in lines, and the size of the unit tests similarly proportional, and it becomes a pretty hard argument to refute. For a long while you were comparing languages line Python to something like Java, and it was pretty much a no brainer - the argument was right. But nowadays if you compare something like Javascript to Typescript, it's not near so clear. Type inference is really biting into the size advantage line dynamic typing had over static typing. The other thing is the claim 1 line of statically typed code of takes as long to write as 1 line of dynamically typed code. It's true as stated, but change it to "correct line of code" and it's not true. You don't have to write tests for type declarations, and statically typed code has less bugs after it's accepted by the compiler / interpreter, so while the writing takes the same amount of time the testing and fixing takes less for statically typed languages. The final advantage of dynamically typed languages is statically typed languages just need more head space - just ask someone who has used Rust. Granted Rust with it's explosion of types caused by lifetimes is a bit of an extreme case, but it's still true for most languages. But it's not so true for the Typescript vs Javascript comparison, and I suspect really careful standard library design might make this mostly go away too.