Hacker News new | past | comments | ask | show | jobs | submit login
The Earl Grey programming language (breuleux.github.io)
150 points by breuleux on May 7, 2015 | hide | past | favorite | 126 comments

Await, pattern matching and macros are great language features Node and CoffeeScript both lack. Nice work!

IcedCoffeeScript had await/defer three of four years ago: https://github.com/maxtaco/coffee-script/

But I agree, excellent set of features.

Async/await is coming in ECMAScript 7, so Node will have it soon.

Define "soon" haha. With the speed of adopting ECMAScript 6 this "soon" could become really long.

    fact(match) =  
       0 or 1 -> 1  
       n -> n * fact(n - 1)  
why not simply

    fact(0 or 1) -> 1
    fact(n) -> n * fact(n - 1)

I think the first one is a single function that has pattern matching for the parameter while the second one is defining the function twice and relying on the language to call the right one based on the parameter.

I think in Erlang the function identification is the name and the arity and this is as close I know of what you would like to have.

Erlang has both methods, pattern matching in function definition and case-clause pattern matching.

(some random link http://stackoverflow.com/questions/1050913/erlang-style-case...)

In haskell you can write exactly the following:

    fact 0 = 0
    fact 1 = 1
    fact n = n * fact (n-1)
It works as you would expect and really they aren't that dissimilar.

Small correction, fact 0 = 1

Erlang does it very much like GP's suggestion with the one downside of being 'noisy' with its punctuation (some might say 'clearer'):

  factorial(0) -> 1;
  factorial(N) -> N * factorial(N-1).

  factorial(Foo, 0) -> .. ;
  factorial(Foo, Bar) -> .. .
Semicolons separate partial definitions for a single function of a single arity, and periods end those definitions. Removing the punctuation might make parsing a bit more difficult (will there be more partial definitions?), so it's understandable to not offer it.

Off-topic: I'm kinda new here, what does GP stand for ?

The grandparent comment, as opposed to the parent comment or the original post.

I'd imagine you could compile them both down to the same thing, as there has to be a conditional at some point to determine what code to run (unless there was some kind of dynamic dispatch), I think it's just a matter of preference

Hyphens in variable names. It's a bold strategy for a scripting language, Cotton.

I think a lot of lisps use hyphens in variable names

There's plenty of precedent. It's just rare for a reason. I'm not sure allowing slightly easier access to CSS syntax justifies it.

I guess it depends a lot on your coding style. When I started allowing hyphens in variable names, I already had about 10,000 LOC of Earl Grey, including the whole compiler for it (seriously, it was a pretty late addition). In all that code, I only had to change one or two lines if I allowed hyphens, so I decided to just go ahead and do it.

And not even because of CSS, just because I think "do-thing" looks better than both "doThing" and "do_thing". Of course, I would think differently if I tended to write "a-b" instead of "a - b", but at the moment I quite enjoy this feature.

I'd love to hear why people prefer underscores over hyphens. I think I grew out of syntax fanboy-ism, but C and python __private__, _special_variable are everything but readable to me. They break the visual line too much. Even historically it was a weird symbol, originally a line break that made it into non-space separator in PL/1, then almost everywhere. Before that it was pure formatting, a typewriter glyph to be overtyped/composed on words to underline them.

> I'd love to hear why people prefer underscores over hyphens.

Because I like subtraction.

It's funny in the end. Parts of languages syntax reinforce other parts. In lisps you cannot write a-b as an arithmetic expression. Sequences of characters are symbols, and that's it. But also, lispers don't really care about infix notation[1] they have abstract polyadic operations: - + / * all take a list of zero or more arguments. (- x y z ...)

[1] And IMO people shouldn't, basic arithmetic isn't that frequent or complex enough to need such a dedicated treatment. But that's another lisp thing, you care more about having DSL opportunities for anything rather than a fixed set.

Just wondering though, how often do people write subtraction as "stuff-thing" instead of "stuff - thing"? I find myself typing the latter almost systemically, so hyphens never get in my way, but of course it's all too easy to be blind to the habits of others.

If you have `stuff - thing` it’s no big deal, but once you start having `(a0-b0)/(a1-b1) + (x0-y0)/(x1-y1)` or whatever, then being able to save all the spaces starts to be kind of nice, especially if your math expressions get to be 60 characters long. There are a few times where I’ve definitely ended up with more readable code by using the presence or absence of a space as a way to group expressions. Also, I generally prefer to write things like `array[n+1]` or `array[n-1]` without the extra spaces.

This code example doesn’t actually have any examples of unspaced subtraction in it, but there are a bunch of other binary operators with space removed. In my opinion adding spaces around all the operators in this file would make the code less readable, especially if trying to follow along from the formal published spec describing the algorithm: https://github.com/jrus/chromatist/blob/master/src/ciecam.co...

I think using spacing for this can be a little misleading since it can obfuscate priority. To give an extreme example, 1e100 * 1e300/1e300 is infinity, because the multiplication is done before the division, but the formatting suggests it's the other way around. It's not too bad in that case but if you were to accidentally group an addition instead the mistake would be harder to spot. I would rather use parentheses all the time to be safe.

Still, you make a good point with the space savings. I guess I just find hyphens nice enough that I don't mind the tradeoff :)

I gave these kind of semantic spaces meaning when I wrote another implementation of AsciiMath[1] because I found it so intuitive. So `1-2 / 3-4 = (1-2) / (3-4)` but `1 - 2/3 - 4 = 1 - (2/3) - 4` like you'd expect.

[1]: http://runarberg.github.io/ascii2mathml/

In long and complicated mathematical expressions I find myself selectively using whitespace to identify logically matching pieces of code, similar to LaTeX' \left( and \right). If I couldn't write (x-y) that wouldn't be possible in the same way.

It's more a question of readability than writeability

Something like:

    result = balance-sheet - ledger-total + discount-total
Is harder to read than

    result = balanceSheet - ledgerTotal + discountTotal

But see, being used to hyphens in variable names, I must tell you I can't figure out which one I find easier to read. On one hand, the presence of hyphens and subtraction is a tad confusing, but on the other, I find "balance-sheet" to be generally nicer to read than "balanceSheet", so it feels like a tie to me. Then, on the other hand, you have things like "xs.for-each(x -> x + 1)" vs "xs.forEach(x -> x + 1)", where I really like the hyphen.

Anyhow, I feel that this is the kind of feature you just stop noticing after a while. I think we often tend to assume that what we are used to is more readable than what we aren't used to, but the brain will adapt to nearly anything. Doesn't mean we ought to go crazy with changes, but hey, I like hyphens.

> I find "balance-sheet" to be generally nicer to read than "balanceSheet"

Me too, and I've wondered if it's because it's activating the same part of my brain that gets aggravated when people decide to use capitals on seemingly random words within the sentence.

To me it's almost cognitively dissonant to see a capitalized letter in the middle of a word. In written language grammars, we have only capitals at the beginnings of new sentences (new ideas), so my brain sees camelcase terms and wants to separate the words into separate terms.

Also, it feels morally wrong that the first word typically doesn't get the capitalization (and often languages ask/require programmers to follow some arbitrary first letter capitalization rules...). Yes, that's right, it's about ethics in programming language design.

in a complex expression, I tend to use white space, and absence of it, to make it clearer, as well as parentheses. So, sometimes x2-x1 happens.

A search on github etc might give some idea of how common it really is - though their search wasn't precise enough last time I tried. Google code was (regex) but I think they closed down? Or you could just clone a few of your favorite projects and grep it.

But you don't like spaces in statements?

Because fanboy-ism could be parsed as fanboy - ism = fanboy minus ism.

You mean like this :

let some-var = a-number-another-number

instead of

let some-var = a-number - another-number

On the other hand, pinky fingers everywhere are collectively breathing sighs of relief from not having to push down the shift key nearly as often.

Don't pinky fingers press the hyphen key?

My ring finger does. Not sure which is more common.

Well, in standard touch typing, the pinky handles the hyphen.

It does? I never learned it that way. Using the ring finger is much more ergonomic for me than using the pinky finger.

Out of curiosity, what's that reason?

Because a hyphen is also the subtraction operator. It's extremely weird if a+b is valid syntax but a-b is not.

This language may require whitespace around all operators, though. It's not clear to me.

Ah that makes total sense, thanks!

Yes, because Lisp uses prefix notation and hence there is no ambiguity: a-b cannot be interpreted as subtraction in lisp but in EG ...?

Lisps do not have operators. This language does.

Imagine being able to type style.border-width in JS instead of style.borderWidth


inner = border - width

I like it. It goes further than coffeescript and livescript can. Especially with macros and better react support

One one hand, this is yet another syntactic reskin of ideas from modern imperative-land. On the other hand, it looks like a good one, by a thoughtful author who also cares about integration with existing platforms, and that’s not nothing.

Actually, the language looks and reads great. Keep up!

High-level comparison with Haxe:

EG and Haxe share: - compiles to Javascript - Macros - Pattern Matching - async/await (Haxe requires a 3rd party macro lib, but it integrates well)

EG has: - cleaner syntax - document-building DSL - "One of EG's primary goals is to be as compatible as possible with existing JavaScript libraries and frameworks and the node/iojs ecosystem." (Haxe makes compromises here to support other platforms)

Haxe has: - static type-checking - dead-code elimination - support for more platform targets (PHP, Python, C#, Java and more) - years of experience

I really like how the tooling to create a full application is already there, it isn't like "here's my new language, but there are no libraries, good luck!".

> Global variables need to be declared to be accessible: > globals: > document, google, React

If I'm getting this correctly, what a brilliant idea! Contain the shittiness by having one single location where all globals are declared. So very helpful.

Glad to see this addressed CoffeeScript's silly issue of variable scope by just requiring a `var` statement like regular JS.

Also, love the name :)

The "fact(match)" is very strange. In Haskell you have "lambda-case" and in Ocaml you have "fun" as syntaxes to define lambdas that pattern-match on the argument.

That said, I really like the inclusion of async and a DSL for documents. These two are things that benefit a lot from having language support.

"fact(match)" is just making use of a generic feature: the "match" keyword in a pattern dictates that the body defines a sequence of sub-patterns to match at that position. An argument uses pattern syntax, so it works there, but it works in other situations, for example this contrived example:

    f(x, y) =
       match x:
          {m, n} ->
             match m:
                <= 0 -> n + y
                else -> m + n + y
          n -> n + y
Can be rewritten:

    f(match x, y) =
       {match m, n} ->
          <= 0 -> n + y
          else -> m + n + y
       n -> n + y
So you can match hierarchically like that (sorry if the example is strange, it's not supposed to mean anything).

Is there any other language that does this? For me the second example doesn't look better than the first one.

I haven't seen this specifically, but I have seen other languages move common top level behaviors within functions into parameter lists.

Ruby, dart, coffescript allow setting of instance variables from the parameter list:

    class Example
      constructor: (@name) ->
        # do stuff

    class Example
      constructor: (name) ->
        @name = name
        # do stuff
Jonathan Blow's language allows the `using` keyword in parameter list, which desugars in much the same way.

I'm personally not a big fan of using this technique for pattern matching, but its better than requiring explicit match blocks everywhere (like Scala and rust).

I haven't seen the feature elsewhere. The second example looks better to me (syntax highlighting helps) and reduces redundancy and indent, but I guess views may vary on this. Good to know!

If your goal is to reduce redundancy than I think a "lambda-case" syntax might be a better fit. As a bonus it also lets you use this feature on anonymous functions!

https://downloads.haskell.org/~ghc/latest/docs/html/users_gu... http://stackoverflow.com/questions/1604270/what-is-the-diffe...

There already is a "lambda-case" syntax. "fact(match) = ..." is sugar for "fact = match -> ...". If one works, then so does the other.

Personally I much prefer the way I do it, because it is more powerful, works in more places, and requires no new syntactic forms.

Does 'each' serve as both forEach and map? It appears that way from the fib example.

Yes. The compiler detects whether the return value is used or not to determine whether to accumulate the results in an array or not. Also, unlike forEach and map you can use continue or break.

Looks interesting, but odd that the very first code snippet is an infinite loop.

God damn it how did I miss that. The condition was wrong, too. Fixed, thanks :)

Why not have a "Show HN" if you're the one who made it (assuming from the URL)?

I... actually did, a while ago. I got no feedback at all, so I thought I'd do things differently this time around. Might just be a matter of luck.

This is showing HN.

What was the inspiration to create yet another compile to JS language?

I guess I wanted macros and pattern matching without Lisp syntax, and I wanted to retain and use the JS/node ecosystem.

There's also a few language features I couldn't find anywhere else that I wanted to try out (my % operator, ad hoc exception classes, some pattern matching features like coercion and "match" inside a pattern to define sub-patterns, the each operator, some features of the macro system that I have yet to document, etc.)

It's kind of fun, really.

What does the % operator formally do? All I could find in the documentation was :

> Earl Grey's % operator can be used to easily build HTML, DOM, virtual DOM, and other things:

There are examples below that statement for common usage, but if you want a deeper understanding:

    tag.xyz %
       property = value
Will produce the data structure:

    {tags = {"tag", ".xyz"}
     props = {property = value}
     children = {child1, child2}}
as an instance of the ENode class. Then, transformer functions can process the data structure to generate HTML or other things, e.g.

    require: /html
    ==> <tag class="xyz" property="value">child1child2</tag>

Because folks feel there is a need for them, and it's a really good way to get one's feet wet in the compiler pound. Besides, a programming language (dialect) is to a programmer what different size wenches, pipes, or hoes are to a plumber. You need different tools for different jobs and a lot of it depends on your personal preference and what it will take you to be efficient. Besides, how we as engineers solve problems is just as much an expression of our personality as it is a statement about the correct operation of systems. But that's just my view, and I'm definitely biased on the subject because I'm working on a flavor of coffeescript that will compile to msft's typescript as we speak (well, in my spare time).

The ubiquity of the browser one would expect.

I love the name but it will get mixed up a lot on Google searches. Just like it was with the TV series "24" that matched every 24 that was to be found online.

I typed 'earl grey language' and the first three results refer to this, the HN article is in fifth position.

The same as typing '24 tv series' will mostly return relevant results.

As someone who currently uses LiveScript professionally, I look forward to trying out Earl Grey. At first glance it looks very promising.

I'm aware there are many people who don't see the value in creating "yet another compiles to JS language," or those who view them as simply "syntax sugar," but I for one very much appreciate the work done by yourself and others on similar projects.

Am I the only one childish enough to ask for the language to be called Earl Grey Hot in honor of Star Trek?

The JS transpiler should be named 'picard'

If we write a Webpack Hot-Module-Replacement plugin for it, the natural name would be earl-grey-hot.

This is also what I thought. I was disappointed to see no TNG references on the page.

We could always write a little utility in it called Engage to make things right.

I like Earl Grey tea, but I don't watch Star Trek, so it didn't even occur to me. Sorry about that :(

On the other hand I have to admit I like the idea of calling utilities Picard or Engage, they have a nice ring to them. Hmm.

If you do add those references the people will love you!

immediately went looking for a Tea.js.....

This language has it down to a tea.

Hope it fares better than Ceylon.

This is just extremely exciting. I love the language design and it looks usable as all hell. First compile-to-js language I've ever seen that got me excited, let alone the first one that i'd actually use outside of the browser. Congrats.


Gives me a good sip of the entirety: the syntax, macros, patmatching, integrating with React, and even how to gulp it and write tests with it. Yet succinct enough to scroll in one swipe.

Looks like another Coffeescript. Braces/parens are not bad, they structure the code and make it easier for the eyes. Lack of punctuation is as bad as too much punctuation.

What level of automated refactoring and IDE integration does it offer?

The maximum possible level: a self-hosting compiler.

Now add an absolutely vital library called Milk and watch tea lovers everywhere shudder.

Made correctly, earl grey (really any black tea) needs some milk fat to neutralize the tannic acid or it could turn your stomach.

I've thought milk in tea was so silly for years, and you've changed my mind in one sentence. Thank you, did not realize there was a practical reason for making it... taste worse.

...now it occurs that maybe the GP comment meant shudder in a good way. In which case, well played. Quite cricket, my good man.

I had actually noticed that I didn't feel great after drinking earl grey without milk, but never realized why. Thanks for mentioning this!

True story: One day I was out of milk, but figured it wouldn't be too bad to go without and have my tea anyway. Dropped the kids off at school, lost my stomach in the parking lot. 8:00am, lookin' classy at the private school.

I googled this because I find it interesting, but black tea does not contain tannic acid. It contains tannins, but of a completely different type (antioxydants which actually seem to be healthy). I couldn't find anything definitive on whether adding milk is a good idea or not. Maybe it helps for some people.

That's interesting. I wasn't 100% when I posted, it was from memory. On a little checking, it seems we're both right -- tannic acid is a subset of tannins. In any case, milk seems to help, whatever the mechanism.

earl-mocha should be bad enough.

There's nothing wrong with milk tea.

From the replies, it's clear the "one should never put tea in Earl Grey" meme isn't as widespread as I thought.

I drink milk in my tea a lot so I'd be fine with it haha

"Oh god, do we need another language that compiles to javascript?!"

—Everyone, probably

I think it's unfortunate that this is currently the top comment on this post. Someone puts hundreds of commits of work into a nifty language with a nice feature set (including compile-to-js) and awesome integrations, and this is what they get on HN...

I think this looks like a great project, we can never have too many programming languages to play around with.

>we can never have too many programming languages to play around with.

Wait until we start having legacy code in all those languages.

>we use 30 different compile-to-js languages used in our projects(that are no longer being developed) and you have to maintain it.

Shouldn't the tech leads in that case be picking one well maintained variant, then?

With what crystal ball? If you pick based on history, since you kinda have to, then you pick a more mature (older) language. If you're not careful with that you pick languages just before they go obsolete.

(Like a company I know which picked VB 6 and still hasn't fully migrated away from it. The same company picked Microsoft's AJAX demo as a basis for a JS framework and is still developing that even though MS long since abandoned it. Would it be better to use jQuery, Angular, Ember, React, Riot, etc? Well, those are too new and untested. Very conservative leadership.)

Anyway CoffeeScript and JavaScript work well together and you might want your code to be in CoffeeScript but you have to use a library, say Ember (also a real project at a different company) which means certain improvements, bug fixes, etc. to Ember are done in JS. There's two. I'm sure you could easily end up with multiple libraries written in multiple compile-to-JS languages.

Long-term code maintainability is a bit of a hard problem. You probably need to be constantly refactoring, rewriting, and re-inventing so you don't have too much old code in play anyway. Maybe. What do I know?

Why not? Maybe most of us are not going to use it, but the free market of growing programming languages will have an overall positive effect by inspiring other langs, such as many features of ES6 being inspired by CoffeeScript et cetera.

Competition is good, competition is creativity.

I like seeing the trends of what features show up frequently in new languages. These features manage to work their way into more popular languages, either as changes to existing ones like Java or C++11 or new languages like Swift and Go.

JS is the new JVM. Write once, run anywhere is still the dream, I'm glad new programming languages are taking advantage of JavaScript's versatility.

> Write once, run anywhere is still the dream

More like "write once, run in the browser". I don't think most compile-to-js languages care too much about "anywhere", they are just trying to get nicer alternatives in what is basically a platform (the browser) closed to anything except JS.

The examples in the article seem much more focused on NodeJS.

JavaScript will run anywhere

Javascript will run anywhere you have something to run it on, same as Java, or any other language.

It's better than the JVM. It provides an excellent, high-performance runtime for dynamic languages (JS), and an excellent, high-performance runtime for static, memory-unsafe languages (asm.js)!

Everyone says, “We don’t need any new programming languages cluttering up code bases,” but nobody volunteers to backport their own apps to COBOL.

There shall be no peace until every language compiles to Javascript.

Does a VM of a PC count as a JIT compiler? Can we compile VirtualBox to JS using one of those C->JS compilers? I'm sure we can come up with some stack of turtles here ...

That was my first reaction until I scrolled down the page a little. This language looks really nice.

I quite like this one, though,

is it just me, tiring quickly of this syntax?

I'm not a fan of using -> like CoffeeScript does. It doesn't seem to mean function-body / algorithm-to-do-the-thing to me. That operator makes sense for pointers or for function return value type expressions.

I think that's just because you are used to see it used that way. And it's not like CoffeeScript invented this.

Haskell does it:

    \x -> x + 1
OCaml does it:

    fun (x : int) -> x + 1
It is very common in functional languages to use -> both for function bodies and return types.

I suppose. Kinda like => for "maps to" in some math / theory classes.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact