Type system updates: moving from research into development

Awesome news!

Personally I’m really excited about this. I know it’s super early but I’d love a good type system on top of Elixir.

I understand the first 2 milestones, but can somebody say what’s a difference between the @spec attribute and 3rd milestone?

The third milestone is to introduce the (most likely) $-prefixed type annotations for functions, with no or very limited type reconstruction: users can annotate their code with types, but any untyped parameter will be assumed to be of the dynamic() type. If successful, then we will effectively have introduced a type system into the language.

Can’t Elixir compiler use already existing specs instead of adding yet another annotation? We have any() and term() which should be more than enough as replacement for dynamic().

Also is adding a type system like in the post means replacing part or even all features of tools like dialyzer by Elixir compiler? If so I hope the compilation would not be longer because of creating plts …

Typespecs as they exist right now don’t cover negations only unions. E.g. there’s no way to have something like any atom, but not :ok.

Really happy with the direction the type system is taking so far.

As a wise man once said, my body is ready. It will be glorious (hopefully!).

There are a few details I don’t appreciate about how we implemented typespecs (for example, I find them too verbose) but the deal breaker here is that typespecs are not capable of expressing the proposed type system, so we will need something new.

This is great to hear. Huge congrats to everyone involved. I’m sure I’m not the only one here eager to see these ideas translated to code.

I’d like to especially extend congratulations to Guillaume Duboc – not every PhD candidate has the opportunity for their work to make such a direct and significant impact. It must be extremely exciting!

is there any information if the new type system will support literals as types?
one of the limiting factors for me to use dialyzer types is that it only support atoms for specific values when building types.

Edit: I misunderstood the question. I’m not sure that it will support arbitrary literals beyond atoms.

Sweet Jesus, Halelujah!

Working with Elixir/Phoenix daily, you slowly get used to some level of pain and suffering, typo here, wrong return tuple here, wrong atom elsewhere. You start see it as part of the job.

But then, you switch back to ASP.NET for another project, like we do in my team, and you find out the pain is gone, there is no more suffering, you write more but you can relax, you can trust the system, if it compiles you are good, if it doesn’t it tells you right away where you made a mistake. With ReSharper and continuous compilation I can see all compilation errors right away, after every change.

As much, as I prefer Elixir/Phoenix for production systems and we introduced some into DHL, this is an enormous problem for us, the trust in the code is simply not there.

Does this mean that the type system you’re exploring will ultimately replace typespecs entirely?