I worked on a summer project to add type annotations to Ruby. Didn't get very far since I ran into some challenges with the internals of the parser and the parser library, Ripper. I'm extremely interested in seeing how the Ruby team designs the type system. It'll be gradual, of course, but also it'll be interesting what adaptations they'll have to make to accommodate existing code. JavaScript relied on a lot of stringly typed code, so TypeScript added string literal types. Perhaps Ruby's dynamic, block oriented style could lead to some interesting decisions in the type system.

Not to mention, the types will most likely be reified as per Ruby's philosophy.

Super excited for this. Between the JIT and types, Ruby could definitely see a renaissance in the near future.

And +1 on the Ruby renaissance! Super excited about all the exciting things that are currently being built!

Speed is not the first and foremost benefit of types. Type checking is (and other stuff that comes with that, like better completions, self-documenting code, etc).

One of the reasons why Sorbet does both runtime checking[1] more than just static checking is so that we can know that signatures are accurate, even when a typed method is called from untyped code.

If the signatures are accurate, a future project could take advantage of method's signatures to make decisions about how the code should actually be run. If the signatures lie, then any runtime optimization made using the types would only be overhead, because the runtime would have to abort the optimization and fall back to just running the interpreter.

[1]: https://sorbet.org/docs/runtime

Dart had gradual types but didn't enforce them at runtime because of performance. The PyPy devs don't believe that type annotations help them for performance (http://doc.pypy.org/en/latest/faq.html#would-type-annotation...). Also there is no JS engine that uses TypeScript annotations so far to improve performance.

Types are usually on the wrong boundary: e.g. Integer doesn't state whether that value fits into a register or is a Bignum.

Also: Aren't some type checks quite expensive? So more expensive than a simple map/class/shape check? E.g. passing an array from untyped code to a signature with something like `Array<Integer>`. Wouldn't a runtime that verifies signatures have to check all elements in the array to be actual Integers?

To me, it feels that there is a very thick wall in between high level languages and something with raw data access like C, C++, and D. You either completely throw out every convenience feature, or go all in on them.

In C, a lot of data access turns into single digit number of load/store or register access instructions. It is easy to see that it is close to impossible to add fancy data access functionality on top of that without going from single cycles to kilocycles.

I was once told "when your try improving a programming language performance, it eventually turns into C"

P.S. on JIT - it is not given that a JIT language be automatically faster than a well written interpreter on a modern CPU. One of early tricks of making fast interpreters was to keep as much of interpreter in cache and data in registers as possible to benefit from more or less linear execution flow of unoptimised code in comparison to unpredictable flow of JIT made executable code. Today, with 16MB caches, I think the benefit of that will be even bigger.

Which is kind of ironic given how bad C compilers generated code during the mid-80s, versus other mainframe languages.

I doubt a competent JIT is ever slower than a competent interpreter, but it may not be that much faster or worth the workload.

It depends on the size of the primitives. An array language could be close to 1:1, while for a cpu-level instructions you will struggle to reach 1/6 of JITted perf.

now that ruby has an actual jit compiler, it could benefit from typing to optimize code further. And a gradual migration process will help people speed up parts of their code. Unless they mess it up like python where abstractions are costly.

In certain basic blocks typed ints or floats can be unboxed, if they will not escape. This is what php7 made 2x faster. the stack will get much leaner. simple arithmetic ops can be inlined, using native variants. ops with typed vars cannot be overridden.

I have said for awhile that "Ruby with types" would be my favorite language to work in. I recently returned to Ruby briefly and had to integrate with a poorly-documented API. I spent more time digging through third-party code trying to figure out what certain parameters were supposed to be than writing the program itself.

One of the major things that has kept me using Groovy over the last 10 years was the reluctance to leave optional / gradual typing behind. Now, nearly every major dynamic language has given in and introduced types, so it seems like this idea of hybrid dynamic/typed languages is now fully mainstream. The problem of course, is they are all built on a legacy of untyped code, not to mention giant communities of people with no culture or habit of tying their code. So it's not clear to me that any amount of added language features can actually compensate for that.

Some languages are better from a static POV, and offer some auto features. Some languages are better from a dynamic POV and offer some hinting feature.

You don't want to type your code to do data exploration and analysis, but you may want to extend the original project later to something bigger and move on to types.

There is no such thing as the perfect language for everything anyway. Plus, it's very good that some languages integrates unnatural features to them, for the case where you want to go beyond their initial best case scenario. It won't be perfect, but I don't need perfect, I need programmatic.

The world of programming is vast, the pool of programmers very heterogeneous, and the constraints are super diverse.

That could be the best of both worlds.

By now I think it is quite obvious that type annotations aren’t as helpful as initially expected and that a library approach seems more pragmatic and more powerful. See Clojure + Spec.

The thing is dynamically typed languages with type annotations tend to no longer feel like dynamically typed languages as the annotations and the tooling spreads and spreads and spreads. Not easy to put up boundaries.

There seems to be a never ending cycle of new languages that are dynamically typed because it is easy for small codebases, which then become popular, get large codebases and then realise that static types are actually a really good idea.

Python, Dart, Ruby, JavaScript (via Typescript), etc...

    “foo”.class # => String

This might be misleading. That is, jump to around the 29 minute mark where he talks about the type profiler and .rbi file stuff.

also for some reason homebrew really likes to updates its index all the time (I think it got tamed in the newest version), but setting HOMEBREW_NO_AUTO_UPDATE to 1 helps a lot.

For example, Java is pretty terrible at type inference (still) and you have to annotate types almost everywhere (Java 8 had a very tepid improvement on that front.)

But languages like Haskell and Rust are very good at type inference, and you almost never actually need to specify the types.

It's still good Haskell style to always annotate the type sigs of top-level functions. Why? Because they serve as more than just hints to the compiler: they are part (and a very important part!) of the documentation. That is why they're in-line. Because A function like

    zipWith:: [a] -> [b] -> (a -> b -> c) -> [c]

  sig {params(name: String).returns(Integer)}

  sig {name: String, returns: Integer}

In this case, your example is not valid syntax, which violates this rule. Not that I personally could tell you why the parser makes a distinction here, but it's at least part of the reason :)

  irb(main):010:0> foo {a: "b"}
  SyntaxError: (irb):10: syntax error, unexpected ':', expecting '}'
  foo {a: "b"}
       ^
  (irb):10: syntax error, unexpected '}', expecting end-of- input
  foo {a: "b"}
            ^
   from /Users/bhuga/.rbenv/versions/2.4/bin/irb:11:in `<main>'
  irb(main):011:0> foo {params(a: "b")}
  NoMethodError: undefined method `foo' for main:Object
   from (irb):11
   from /Users/bhuga/.rbenv/versions/2.4/bin/irb:11:in `<main>'
  irb(main):012:0> 

We had them as hashes for a while, but it meant that code in all sigs was loaded as the code was loaded, even if runtime typechecking was disabled. We were forced to load all constants in any signature in a file, an effect which cascades quickly. It had a big impact on the dev-edit-test loop.

For example, if we're testing `method1` on `Foo`, but `method2` has a sig that references `Bar`, we'd have to load `Bar` to run a test against `method1`.

Now sigs are blocks and lazy, and we pay that load penalty the first time the method is called and a typecheck is performed.

   sig {params(name: String).returns(Integer)}

   sig [String]=>[Integer]

It was a design decision that all type annotation arguments be named as opposed to positional. As one example why, it makes the error messages better. You can always say "You're missing a type declaration for parameter 'foo'" as opposed to "You have four positional arguments and 3 types".

We could probably still bikeshed our annotations inside the `sig { ... }`. I'm not sure we'd make constants with unicode like BigMap‹Address→Account› for generics, though, how do you even type that? :)

Having said that as far as I understand, type support in Ruby 3 will not prescribe which type checker is used and what limitations exist. Some of the mentioned projects are structural and I think even Sorbet might add support for it at some point.

Classes are types by default, but you can define non-class types as well: https://sorbet.org/docs/abstract

0: https://crystal-lang.org

- https://sorbet.run/talks/RubyKaigi2019/#/53

- https://sorbet.run/talks/RubyKaigi2019/#/55

So I don’t think that the divide will be at Rails. And more than that, I think there will be very little divide at all. Sorbet is designed to be gradual, so it works 100% fine with untyped code:

https://sorbet.org/docs/gradual

https://sorbet.run/talks/RubyKaigi2019/#/14

But we'd like to catch more than just errors related to constants, like those related to missing methods, or calling a method with the wrong arguments. Errors like these are only reported in files marked `typed: true`:

https://sorbet.run/talks/RubyKaigi2019/#/19

Sorbet doesn't need to have method signatures to know whether a method exists at all, or what arity that method has.

But more than that, Sorbet ships with type definitions for the standard library. So you don't even need to start annotating your methods to type check the body of your methods, because most of the body of your methods are calling standard library things (arrays, hashes, map, each, etc.).

The statistics in those slides are sharing "out of the box, what's the highest strictness level that could be added to a file without introducing errors?" So ideally an entire project be made `typed: true`, but Sorbet can be adopted gradually, so a project can exist in a partial state of typedness. We wanted to see how painful it would be to adopt Sorbet in a handful of large, open source rails projects, and it turned out to be not that bad.

[1]: https://sorbet.org/docs/static#file-level-granularity-strict...

If not then why are you using a dynamically typed language? If you're not using it's abilities then it doesn't sound like the right tool for the job.

It's like a cost/benefit analysis where none of the benefits you're using, and the cost is the total inability to validate, refactor, and navigate your code base.

Yes, yes, and no. I do most of my work in languages that prevent the first two, but when I do have access to this kind of runtime trickery I do use it when useful.

The issue is not "Do you purposefully do those things?", but rather "Do you have a call stack where you can't guarantee it won't happen by accident?" Type checking is not relevant when you know what will happen and want the dynamic/ducktyping behaviour.

Another issue is: I'd use a different framework which doesn't use Ruby, but this was the most productive framework at the time the codebase was started, and nobody will port that many lines of code to a non-dynamic language now. So the best course of action is to validate the current code is not overly-dynamic.

I guess some companies started fast with Ruby/Python and similar and instead of rewriting to static/typed languages they pushed forward features that would allow them to just continue where they left off at the expense of having a more concise problem oriented programming language that's good for solving specific problems.

Will be interesting to see how ruby handles types vs duck typing etc 10 years from now, when the new best practices have been figured out.

But it is always an incomplete solution because a) old code needs to be retrofitted, see TypeScript's way of defining type maps for vanilla JS or b) more commonly you keep the code around that's using unsafe types, effectively passing void*|Object|"choose your poison" around.

I'm personally tired of staring at variables trying to figure out what they're supposed to be, then having to dive into source to see how its used. C/C++/C# solved that problem, why are we still dealing with it?

How does nobody(?) support this yet?

Python supports some parts: you can subclass int, and you get all of the int methods like addition and subtraction for free, but "distanceInKm + distanceInKm" gives you an int instead of a distanceInKm; and "distanceInKm + distanceInMiles" gives you an int instead of an error.

Rust also has partial support but from the other end: distanceInMiles and distanceInKm can be two distinct subclasses of int, and adding them together is a compile time error. But also adding distanceInMiles with distanceInMiles is a compiler error, because these are basically "completely new classes" rather than "subclasses of int", and so you have to implement add / subtract / stringify / etc for yourself for every type D: (I'm fairly new to Rust so if there is a shortcut there that I'm missing please do point it out)

Depending on your type system, a well-typed program can eg run faster, because the compiler / interpreter can elide certain runtime safety checks that would be necessary in untyped code.

If your type system is crazy enough, you can even track the runtime complexity of your program at the type level, including whether your program runs in finite time. See eg Dhall (https://dhall-lang.org/) whose type systems only allows programs running in finite time.

Generally, there is this huge disconnect between how code is expressed as text and how it is handled in as a graph structure inside the tooling. It is soon time to move beyond simple text files for code, I believe.

So, please, be brave and evaluate a 100kloc codebase+deps that may contain a top level `rm -rf ~/`

But honestly, if you're asking your IDE to do it, that means you're asking your IDE to do static analysis of your code - and type-checking in a lot of ways is just another static analysis technique. And for a lot of us (myself included) we prefer to catch as many of these bugs as possible using static analysis, instead of waiting for someone to get paged when it causes an outage.

Yes, there's a trade-off, and types can be obnoxious (Java imports being probably one of the worst offenders, C++11 introduced `auto` for a reason), but that's the cost we pay.

Now the JS fanboys discovered and moved (from Coffeescript to Babel ESxx) to Typescript they apparently think that they can write beautiful and bugfree software just because of static type checking! Let them please move to C++ or whatever statically typed language and shoot themselves in the foot by making all those mistakes that have nothing to do with static type checking at all! Oh, and of course hitting the wall because they are missing their precious 'any' keyword!

I totally agree that type checking for dynamic languages should be done in the IDE, tooling. But static typing in the dynamic language world is a hype at the moment, so we'll have to go through a wave static type checking frenzy. For my work I look at horrible code bases, perfectly typed and strictly formatted by tslint..

> For my work I look at horrible code bases, perfectly typed and strictly formatted by tslint.

There is no language that can stop people from producing horrible code.

Javascript and ruby, the underlying types can change depending on where the code is in execution - a variable holding a 1 can turn into a "1" and back (implicit type conversion - try 3 * "3"). This leads to a whole class of bugs not possible in a statically typed codebase where explicit conversion needs to happen - I have no hard data, but I remember debugging this type of stuff far too often and far too many times when I could've spend my time better elsewhere. (but I actually like ruby a lot!)

Type checking is not the same as being statically vs. dynamically typed!

Meanwhile, I've gotten myself more and more into Clojure. Which now that other dynamic languages seems to move closer to types, seems to be in a niche in that Clojure is moving further away from types.

It'll be interesting to see what happens at both extremes and in the happy middles.

What about clojure.spec?

The earliest possible date (and somewhere I read it's a probable one, but I can't find it right now) is Christmas 2020.