Hacker News new | ask | show | jobs
by film42 4438 days ago
Probably the best change in my opinion was resolving the arity limit for case classes: https://issues.scala-lang.org/browse/SI-7296

I've bumped up against that limit pretty bad while trying to deserialize json. Shapeless did a lot to solve the problem, but I'm sure glad that limit has been removed.
2 comments
neverminder 4438 days ago
I wish the same was done for tuples, but it wasn't and it's unclear whether that's even in plans.
link
DannoHung 4438 days ago
I'm not sure I understand the situations where having a tuple of that arity is beneficial. Some sort of code generation thing, perhaps?
link
joostdevries 4438 days ago
Shouldn't that go hand in hand? The generated unapply method will involve a tuple of same arity I believe.
link
gourlaysama 4438 days ago
For that very reason, there is no generated unapply method for case classes with >22 parameters. This doesn't prevent pattern matching on the case class, because the patmat knows it is a case class and extracts the parameters directly, without calling unapply.
link
neverminder 4438 days ago
As the matter of fact I'm not sure now. I was looking for a confirmation that tuple limit is removed in Scala 2.11, but couldn't find it whereas statements that case class 22 parameter limit is removed are all over the place.
link
gourlaysama 4438 days ago
AbstractFunction, Function, Product and Tuple are all still limited to 22.

Increasing the limit would create too much bytecode, and in the current design, lifting the limit is just impossible (without runtime code generation or custom classloader, etc.).

link
film42 4438 days ago
So why did they choose to allow case classes to have arbitrary arity? Not that I'm complaining, I just don't understand the underworking of scala that well.
link
frowaway001 4438 days ago
They just don't emit all the boilerplate which would require unlimited-arity Tuples/Functions/... for larger case classes.
link
acjohnson55 4438 days ago
film42: I think it was driven by Slick and the need to accommodate database tables with more than 22 columns
link
neverminder 4438 days ago
Indeed, because the fallback is HList and when I have tables up to 200 columns it's killing the CPU and crashing Scala IDE up to the point where you just can't work with it.
link
soq_4 4438 days ago
Another really big change is the AnyRefMap class which provides a very fast mutable hashmap. There's also a specialized map for keys of type long.
link