Wednesday, April 9, 2008

On defence of implicit arguments and the lack of type inference power

I found two quite recent comments by Martin Odersky (main developer of Scala) that give insight to two commonly arised questions. They are 1) why implicit arguments exists, and aren't they too dangerous? And 2) Why on earth is this so cool type inference system behaving like an idiot, and isn't inferring some obvious cases.

Here are the quotes and links to the original source.

1)

"In practice, of course, you want to find the right mixture between expressiveness and safety. -- -- and you'll accept some implicit parameters at places where they are really useful. And they are amazingly useful when used as a tool in the right hands.

For instance ScalaCheck lets you write properties of your programs and then constructs unit tests for these properties automatically. Several bugs in the standard Scala library have been found this way. Without implicit parameters, the plumbing you'd have to do to set up the properties would make them much uglier, so people most likely would not write any. So that's an example where implicits are really useful. Expect to see many more libraries that do amazing things like ScalaCheck in the future.

-- --

P.S. I think this discussion highlights a bit the design philosophies behind Java and Scala. They are not the same. Scala puts more trust in its users that they will do the right thing. Maybe this assumption is wrong, and users will make a mess of it, but that's what Scala is."

Martin Odersky, posted on March 9, 2008 at 6:16 am

Original post

2)

"The reason Scala does not have Hindley/Milner type inference is that it is very difficult to combine with features such as overloading (the ad-hoc variant, not type classes), record selection, and subtyping. I’m not saying impossible — there exist a number of extensions that incorporate these features; in fact I have been guitly of some of them myself. I’m just saying it’s very difficult to make this work well in practice, where one needs to have small type expressions, and good error messages. It’s not a shut case either — many researchers are working on pushing the boundaries here (look for instance at Remy’s MLF). But right now it is a tradeoff of better type inferencing vs better support for these features. You can make the tradeoff both ways. The fact that we wanted to integrate with Java tipped the scales in favor of subtyping and away from Hindley/Milner."

Martin Odersky, posted on April 9, 2008 at 2:08 am

Original post

I hope this is not violation of any copyright ;)

Another news regarding Scala are that the contractiviness requirement for implicits was removed by David MacIver's suggestion. Moreover in one month Scala is likely to have some kind of version of virtual classes, so it's possible to extend class hierarchies in one more powerful way. I hope there will be a good documentation available when it's out *cough* existential types *cough*. :)

2 comments:

B Jax said...

"Moreover in one month Scala is likely to have some kind of version of virtual classes, so it's possible to extend class hierarchies in one more powerful way."

I'm just wondering if you could point to a link where I could read more about this, it sounds pretty cool and I hadn't heard about it yet.

Henrik said...

Sorry, it has been only discussed slightly on mailing list, and you can find some information searching from Nabble or Gname. Also there was a draft document about it, but I can't find it anymore.

Here's the ticket for the addition btw. http://lampsvn.epfl.ch/trac/scala/ticket/634

I encourage you to ask in the mailing list!