Announcing TypeScript 1.4 | Hacker News Reader

Announcing TypeScript 1.4

blogs.msdn.com

227 pointsnumo1646 comments

TypeScript is pretty amazing. In a few weeks, it turned our node project from an unruly mess (with tests) to a well organized and compact codebase with nice interfaces that formalize the various data types that flow around.

We just changed the extensions of all of our files to `ts` and added some type definition files (most from DefinitelyTyped) to our project; then started gradually adding types to the codebase. The compiler complains about type errors, but always generates JS from code that is still valid JS, so the project was fully operational during the migration.

Once that was done, moving code around, renaming things and changing entire interfaces became really easy.

The structural type system (basically formalized duck types) is a great fit for JS, and features such as generics ensure that the typesystem is expressive enough to model common functional code (no higher kinded types though, look into purescript for that).

Its not perfect (all types nullable by default; function arguments are bivariant wrt input/output params and some other weirdness here and there) but its a remarkable improvement over plain JS.

I just recently migrated as well and it was very straightforward and pleasing.

Little oddnesses here and there in DefinitelyTyped d.ts choices.

Overall though typescript feels like first class javascript+a much better future javascript.

Agreed. We migrated most of the ES6 codebase to Typescript, and it feels great.

We have an angular.js app that makes REST calls. The Java backend has DTOs (data transfer objects) that are mapped by Jackson to json, and after migrating to Typescript, we also run the DTOs through compile-phase that generates corresponding typescript interfaces.

i.e.

in Java:

  public class PersonInfo { // the dto for person
     public final String name;
     public final PersonId id; // we have a class for each id type
     public PersonInfo(...) { ... }
  }

  @RestController
  public PersonController {
    
    @RequestMapping("/get-person-info")
    public PersonInfo getPersonInfo(@RequestParam int id) {
      // dummy example
      return new PersonInfo("Peter Jackson", id);
    }  
  }

In the java->typescript generator we add the controllers for which we need the rest API definitions to be generated as typescript interfaces:

   generator.addClassesFromController(PersonController.class);

And when the generator is run, a typescript file is generated with following definitions:

    export interface PersonInfo {
        name: string
        id: PersonId
    }

    interface PersonId {}

What tool do you use to emit the typescript interfaces? I looked around for one of these a few months ago but didn't find anything at the time.

It's custom made. We are extracting it from our current project, anyway, so I'll ask the dev responsible if he would release it as open source project.

You can cheat by using Java reflection on the POJO, then it is just a "relatively" simple matter of printing strings.

" a remarkable improvement over plain JS" - Agreed. Optional types for the win. Types when you want them but just make something <any> when you want to get up to your old js tricks. Dart gives you a first class experience of optional typing with things like checked mode at runtime but I think typescript gets the 80/20 by enhancing js in a very clever way.

TypeScript has been great. With plain JavaScript you have to be outstandingly organized and diligent to maintain any sort of medium to large project. But this part of the process tends to get neglected, and eventually a point is reached where you can either a) break stuff at run-time in unpredictable ways, b) stop refactoring and deal with the consequences, or c) have a proper suite of unit tests. c) requires almost as much discipline as keeping the code clean in the first place, and a) or b) are not great options.

Luckily TypeScript helps out quite a lot and basically for free. Converting an existing JavaScript project to "naive" TypeScript takes hardly any effort, and as the project evolves it is easy to add type annotations and interfaces at any pace. Once a module is properly typed it (almost) stops breaking at run-time, yay! It's also very clear what will be executing in the browser. Using a cross-compiled language that follows the principle of least astonishment is a real treat.

Obviously it can't prevent compile-time errors, generate unit tests, or build a good application for you, but you do get quite a lot of help with the problems it can solve.

Some shameless promotion for a PR I worked on: If you're using the AMD features of TypeScript, support for named AMD modules also made it out in TypeScript 1.4. An example of the syntax can be found in the following file: https://github.com/Microsoft/TypeScript/blob/v1.4/tests/case...

I think the union type thing is cool. As someone who has never worked with something like that before I have to say that I'm not sure I like the conditional inside the function to check the type. I feel like it would be cool if there was some other construct to match on the argument type and then send the arguments to more specifically typed functions.

I am most excited about template strings, simply because they let you write multiline strings.

e.g. instead of

"hello\n" +

"world";

use

`hello

world`;

But who needs typescript for that?

http://tc39wiki.calculist.org/es6/template-strings/

That is precisely what TypeScript has implemented: ES6 string templates. Remember, TypeScript is JavaScript + optional types.

One nice thing about the template strings in TypeScript is you can use those and still compile to ES3/ES5 for today's browsers.

You can compile ES6 for today's browsers too. Check out 6to5.

> But who needs typescript for that?

Who needs javascript if you have dart? Who needs dart if you have go? Who needs go if you have erlang? Who needs erlang if you have c? Who needs c if you have assembly? Who needs assembly if you have binary code? Who needs binary code if you can just control electric signals by manually plugging and unplugging a cable to the wall?

Apparently all you need to build a good website is a power cord and some patience!

The JavaScript programmer in me welcomes that but the PHP programmer in me is a bit annoyed for having to remember another.. inconsistency.

Not surprising from MS - that was a feature in C# 1 and it immediately made me angry at every language that doesn't support it.

Typescript is really starting to look quite compelling. If we could have null-safety and ADTs I'd be using it in a heartbeat.

I'm hoping these features find their way in, too. I was just looking into their status yesterday, in fact, and came across these tracker threads:

* Non-nullable types: https://github.com/Microsoft/TypeScript/issues/185

* Sum types: https://github.com/Microsoft/TypeScript/issues/186

The short story is that there's some interest in both features, but there are challenges that someone needs to propose a solution to.

  TypeScript now supports using ‘let’ and ‘const’ in addition
  to ‘var’. These currently require the ES6 output mode, but
  we’re are investigating relaxing this restriction in future
  versions.

Would be quite nice to be able to use 'let', and just use variable renaming to generate ES5 code with unique names in the case where scopes overlap. I guess it might be coming?

You could probably use TS's ES6 output and pipe it into Traceur or 6to5 or something.

Variable hygiene is not sufficient in loops - you have to make IIFE's for that. Then you have to account that the code inside that IIFE might be referring to the arguments object of the outer function, so you have to forward the outer arguments object to the IIFE as well. It's not impossible but the end result is somewhat gnarly.

Frankly, use of "typeof" to tell between explicitly named types baffled me a bit. I hope they'll introduce pattern matching on type some time later.

The proposed syntax is

  function f(val:any): val is SomeType {
    do checking and return boolean
  }

and will be most likely available in 1.5.

There's plenty of discussion on this very topic in GitHub issues and PRs. The current "typeof" is inherited from JS and TS compiler doesn't actually do much to change that code. It would be great to see a nice pattern matching syntax, but people will have to figure out what the JS that pattern matching code would look like.

The discussion on Github for user-defined type guards (spion's comment) is at https://github.com/Microsoft/TypeScript/issues/1007

Advanced pattern matching, monads, etc. might be out of scope for TS, since it doesn't want to have major syntactical additions from regular JS.

Anyone have any guidance on navigating es6 output mode vs where firefox and chromium are these days?

Ie whats in typescript es6 output that will have issues?

Assuming this source is up to date[1] (and I'm reading it right), then the intersection of chrome ES6 features and Typescript ones is an empty set, while the Typescript ones are a strict subset of Firefox's ES6 features.

[1] http://kangax.github.io/compat-table/es6/#typescript

What is the relationship between the upcoming TypeScript 2.0 and ES6? Are they trying to achieve parity or keep evolving as a superset of standard JS?

Both?

"In addition to the type system improvements, one of the main goals for the upcoming TypeScript 2.0 release is to fully support the ECMAScript 6 standard. With TypeScript 1.4, we take another step towards this goal. In this release, we’ve added a new ES6 output mode, support for let and const, and support for ES6 template strings."

There will never be "types" in javascript,at most something like "guards" or runtime assertion checking, but they will never implement typescript type system. It's too late for that.

Its not too late, gradual typing can still be added. Maybe in ES7 or ES8.

EcmaScript 4 had a type system. Although ES4 never quite made it to browsers, there certainly was enough interest to standardize a type system in ES4. Microsoft is a part of "they" at ECMA determining the ES standard and it is certainly within the realm of possibility that if TypeScript proves people want it, a TypeScript-like type system could indeed become standard in some future EcmaScript edition. In some ways, TypeScript is Microsoft's playground for things that they would like to see eventually become standard.

It's also worthwhile to remember that gradually typed languages is /still/ an active research area (MSR has several people working on it, including around ES/TS, and there were adverts for sponsored PhDs around gradual typing in ES/TS around a year ago). I think it's still a bit early to burn something into the standard (once something ships, it'll be near impossible to ever remove; this is probably partly a reaction to ES4 which built a lot on fairly recent and ongoing research, arguably leading to its downfall), but I also think it's a direction most people within the committee want the language to take.

Indeed we had structural types in ES4 toward the end, and almost bluffed MS into folding (I'm told by a reliable source). But ES4 was trying for too much too soon.

At this point we really need to see ES6/7 and TypeScript + Flow "gene-culture co-evolution". As I said at the last TC39 meeting on the topic, big de-facto standard wins trump rushed de-jure standards any day.

/be

I don't understand the Flow "gene culture co-evolution" part. What do yo mean by that?

Google works, first hit for "gene-culture co-evolution" is

http://en.wikipedia.org/wiki/Dual_inheritance_theory.

The analogy is

* core JS standard = genes, DNA in full (mitochondrial [asm.js? :-P] and nuclear [dynamically typed JS]);

* Flow, TS, TS* and other research = culture.

Obviously it's a loose analogy, not an identity. There's no "Red Queen" (sex, sexual selection), but we do definitely see selection over time via developer adoption, what wins on performance and ergonomics, what feeds back from the ecosystem into the core standard APIs.

Also even a few new special forms: generators in ES6 (prototyped in ES4) were inspired by many experiments; mainly they constitute introgression from Python 2.5+.

JS as an evolutionary kernel (http://www.cc.gatech.edu/~sakhshab/evoarch-extended.pdf) consists of almost totally conserved material (always extended, backward compatibility required for adoption by developers and competing browsers), only slowly extended, while diverse languages and frameworks evolve above and below -- Node.js 14 years after I put JS in Netscape 2, e.g.

Feedback goes both ways, especially in the post-ES3 era where Firefox brought the browser market back to life. See ECMA-262 later editions (the Harmony era) rolling up de-facto standards such as Array map/reduce/forEach/etc.

The same feedback could and should happen with a standardized type system of some sort. It won't happen via design by committee, or picking one winner too soon.

Type system implementors and the JS stewards must communicate well for this to win. It's looking good so far, on Ecma TC39: JQuery, Facebook, Netflix, PayPal (Ebay originally, and again), Twitter all represented along with Apple, Google, Microsoft, and Mozilla. Also academic researchers from various universities, all of whom love type systems and theory :-).

/be

Masterfully applied analogy. Thank you.

"Never" is a long time... there's rumblings in the ES Discuss notes about "Type Annotations" https://esdiscuss.org/notes/2014-09-25.

I'm not so sure about that. Microsoft, Google, and Facebook have all implemented pretty similar typed supersets of JavaScript (TypeScript, AtScript and Flow). At least in Google's case, they are intentionally keeping AtScript similar to TypeScript in the hopes of influencing ES7. I expect that Microsoft and Facebook have similar ambitions.

Most of the JS runtimes already do some kind of type inference and checking in the JIT pipeline so providing those hints ahead of time is technically feasible.

> Most of the JS runtimes already do some kind of type inference

Problem is they are not available to users. I dont care what the JIT does,I want my IDE to help me write correct code,that's the goal of Typescript.

How do you update to this version if you just use the node version?

Have you tried 'npm update typescript'?

I have indeed. And when I run tsc -v I get 1.0.3.0

Maybe that is the stable?

The npm page shows 1.4.1 as latest. Maybe your json file is forcing a specific version.

Reply on news.ycombinator.com