Comment by harveywi | Hacker News Reader

Comment by harveywi | Hacker News Reader

Parent Full thread

I dabbled with it about six months ago. I write most of my software in Scala, but I continue to hit the ceiling and run into places where type inference fails, compiler bugs add friction, or other oddities (22-tuple limit) get in the way. One of my experiments was to see if I could write some portions of my software in Frege, and some in Scala to get the best of both worlds.

I think the answer that to that question appears to be yes [1], but I stopped experimenting. I am considering resuming my experiment now that Frege appears to be crystallizing a bit more.

The java interop is nice, albeit it imposes discipline (as it should). Since java doesn't care about side effects, most of the values returned from a java function call will be in a Mutable monad. Here [2] is a good walkthrough for how it is done in case you want to try it out.

[1] https://groups.google.com/d/topic/frege-programming-language...

[2] http://mmhelloworld.github.io/blog/2013/07/10/frege-hello-ja...

Do you have some examples for "type inference fails"?

Do you have some examples for "compiler bug"?

Here is a quick shortlist of some type inference issues (some bugs, some enhancements, and some fundamental limitations of the language design) that I run into pretty much daily:

1. (Search for "asInstanceOf"): http://typelevel.org/cats/tut/freemonad.html

2. (Finally fixed and will hopefully be released soon - yay!): https://issues.scala-lang.org/browse/SI-2712

3. https://twitter.com/extempore2/status/430456002392391681

4. http://pchiusano.blogspot.com/2011/05/making-most-of-scalas-...

5. (Anecdotal): Various pattern matching peculiarities.

Compiler bugs: As long as I write code on the happy path and steer clear of scalaz/shapeless/cats, the compiler is pretty well-behaved and I can go months without getting a compiler crash. However, when I add those in and start writing pure functional code with complicated types that hits the intersections of various features (e.g., GADTs, package objects, implicits), I get a few crashes per week related to known bugs.

It works pretty well, and I am not trying to complain too much about all the mileage I get out of Scala and the excellent tooling! Just trying to get the most out of it.

> I get a few crashes per week related to known bugs

Interesting! Can you link to the issue?

Sure, you bet - I will start collecting them as they occur. However, I no longer find new bugs, it seems. I'll break something, google for the error message, and either find something on issues.scala-lang.org, IRC, or one of the Google Groups.

My current code base is not open-source just yet, so reporting self-contained reports may not be possible. (Currently I'm writing an interpreter for a fairly big declarative language using FreeApplicative from cats as the substrate, and I think the sheer size of it along with shapeless and scalaz usage is causing some undesirable feature interactions.

Where would be the best venue for reporting (known) bugs? Is there any value if I can't provide the actual code? Or is just reporting a "I experienced it too" message valuable?

I think this would help a lot. It would also help to check whether the code mentioned in an issue that triggers a crash is equivalent to the code you have.

Even if you can't publish your code, it might help to say "I'm seeing the same error, but I'm using X unlike the exmaple code that uses Y".

This will make sure that people who fix a bug will cover both causes with a test.

Here is a new one that I reported to scala-internals today: https://groups.google.com/forum/#!topic/scala-internals/w4HG...

Great, thanks!

I hit two bugs today. Both were produced during major global refactoring. I did not have time to catalogue the reproduction steps or anything, so I have no right to complain!

1. https://issues.scala-lang.org/browse/SI-9281

2. https://issues.scala-lang.org/browse/SI-8322

And another one today: https://issues.scala-lang.org/browse/SI-9222

Thanks! What the hell are you doing that you experience issues this often? :-)

I'm working on extending this ([1]) to contravariant functors while making the whole thing polytypic and typesafe (so I can abstract over type products and coproducts, among other things). My stuff will be open-sourced next year sometime at the latest. I think it's a combination of feature interactions and frequent refactoring during development that is a source for a lot of stuff that is not nice to the compiler.

edit: The extension to [1] is actually pretty benign and the compiler does well with it. The part that puts a big load on the compiler is that it is used as an interpreter for a popular, yet complicated, visualization DSL. The compiler, especially the typechecker, has to do a huge amount of work.

[1] https://github.com/jdegoes/scalaworld-2015/tree/master/src/m...

I just bumped into SI-1503 [1] and SI-7714 [2]. Fortunately I was able to work around those and keep plugging away.

[1] https://issues.scala-lang.org/browse/SI-1503

[2] https://issues.scala-lang.org/browse/SI-7714

Thanks!

SI-1503 should warn in 2.12, additional progress might happen in https://github.com/scala/scala/pull/5310.

Reply on news.ycombinator.com