10 ms·

LunarML – Standard ML compiler that produces Lua/JavaScript

96 points by injuly 2y ago

humzashahid98 2y ago
This is a very cool project. There is also MLKit's SMLtoJs which compiles to Javascript, but not to Lua. https://github.com/melsman/mlkit/blob/master/README_SMLTOJS.md https://github.com/melsman/mlkit/blob/master/README_SMLTOJS....
- dybber 2y ago
  Which in turn compiles itself into Javascript, such that you can run the SMLtoJS compiler in the browser, e.g. it should be possible to make it accept <script> tags with SML code.
programjames 2y ago
I got tricked up by the acronym. This compiles the programming language "Standard ML [Meta Language]" to the programming languages Lua or JavaScript. It's not a machine learning compiler.
- hayley-patton 2y ago
  I can no longer google "ML Kit" and find the ML compiler, instead I find Google's ML library. You win some, you lose some.
- zem 2y ago
  those of us interested in the ML language family have been disappointed for years every time a headline turns out to be yet another machine learning thing, so I'm glad to see this one was actually an ML (:
- fortyseven 2y ago
  That's okay, I thought it was machine language at first.
_delirium 2y ago
Blog post announcing it: https://minoki.github.io/posts/2023-12-17-lunarml-release.html https://minoki.github.io/posts/2023-12-17-lunarml-release.ht...
sgt_bilko 2y ago
On work laptop (Windows) at the moment. Will get to it in the weekend, but can some kind soul tell me: is the lua and js generated code readable? (Standard for readable for me is ReScript)
weatherlight 2y ago
I've always loved Standard ML and feel it's never gotten the love that it deserved. It was my introduction to typed FP concepts via this course (Part A) https://www.coursera.org/learn/programming-languages https://www.coursera.org/learn/programming-languages
- mattpallissard 2y ago
  Agreed! But OCaml is close enough that I don't despair. Chris okasaki's purely functional data structures is an excellent intro to SML if your familiar with FP already.
  - weatherlight 2y ago
    I've heard good things about this book, I didn't realize the examples were in SML. :)
- Hirrolot 2y ago
  Is there any reason to use SML instead of OCaml in 2024?
  - SassyBird 2y ago
    Multicore programming that isn’t covered in stickers about being bleeding edge and not fully integrated with the rest of stdlib(s?). Since the 90s. :) And a minor point: operator arguments (as well as curried arguments) are evaluated left-to-right, not right-to-left like in OCaml.
  - someplaceguy 2y ago
    > Is there any reason to use SML instead of OCaml in 2024? * Many compilers and transpilers available. Don't like LunarML? Maybe try Poly/ML. Or MLton. Or MLKit. Or SML/NJ. Etc. All of the above are pretty much complete and production quality, and there are quite a few more. * The language has a formal specification which means there's a document formally specifying what all the constructs in the language should do. In contrast, if you wanted to build your own OCaml compiler/transpiler/interpreter, you'd basically have to follow whatever the main OCaml implementation does, which changes from one version to the next, so you'd always be playing catch-up. * The formal specification hasn't changed since 1997. This means that once written, your SML code should work the same forever. You no longer have to worry about your programs (or their dependencies) breaking when you upgrade to a new version of OCaml. * Compared to OCaml, your single-threaded programs will usually run faster if you compile them with MLton, mostly due to whole-program optimization / monomorphization and unboxed native integers, reals and arrays. * Some SML implementations have interesting extensions, which you can use (at the cost of being tied to that implementation). For example, MLKit supports memory regions (i.e. more efficient memory management in some cases). SML# (which is not related to .NET) supports seamless integration with SQL and other interesting features. Some implementations support multi-threading natively and/or different forms of parallelism. Most of them also support some form of FFI to interface with C or other languages. * SML programs are easier to port to CakeML than any other language, since CakeML is mostly an SML subset. CakeML not only allows you to formally verify that your program is correct (or that it has certain desirable properties), but it also has a compiler that is formally proven to be correct, so your compiled CakeML programs are (mostly) guaranteed to be free of miscompilation bugs as well.
psd1 2y ago
I've had a great time discovering Fable, which compiles F# to JS or TS. The cost of moving from VSCode to neovim has always put me off, but if I could write the config in F# it would tempt me more. How far is standard ML from F#? Is there any existing compiler for F# => Lua?
- lysecret 2y ago
  Have you been using Fable? What has been your experience?
  - psd1 2y ago
    Not in production. As an F# newbie, I hit some incompatibility early on with my initial library choices. Not everything works in fable. When I found Safe Stack, that went away. Then it was great. Hot reload; generated js is passable; I'm doing front end in F# with zero JS and almost no html.
    - JaggerJo 2y ago
      I’ve used Fable in production. It just works most of the time, but it’s not seamless.
- humzashahid98 2y ago
  Standard ML is pretty simple and easy to pick up, but a weak point is the lack of { record with updatedValue } syntax to create a new record which is the same as an existing one but with some value changed. You have to specify every field in the record which can be a pain depending on how large your record is. Standard ML has some nice features though and I think what I mentioned is the only serious shortcoming (except for ecosystem if that's important to you). Standard ML is also structurally (but strongly) typed, unlike F# and the vast majority of nominally typed languages, which is rare but nice. type person = { age : int, name : string } is the same as the type type animal = { age : int, name = string } but this hasn't been an issue foe me, and structural typing means you can avoid defining the type at all which is convenient for small records. For example, a function to increment age has no need for you to define the type. fun grow { name, age } = { name = name, age = age + 1 } I think Standard ML helps you understand the origin of some common FP idioms. Like tuples are really just records and SML treats them as such. The empty record {} is the same as the empty tuple (): both are the unit type. And a record like { 1 : "a", 2 : "b" } is exactly the same as the tuple ("a", "b").
ihumanable 2y ago
If you haven't taken a plunge into Standard ML or OCaml and are interested, I can't suggest this playlist enough https://youtu.be/MUcka_SvhLw?si=rc4ele_0WcjwLFR3 https://youtu.be/MUcka_SvhLw?si=rc4ele_0WcjwLFR3 It's such a lovely introduction to Ocaml and programming in general and has a free online textbook. It's a lot of videos but each one is 5-10 minutes so it's very easy to hop in and out.

96 points by injuly 2y ago

This is a very cool project. There is also MLKit's SMLtoJs which compiles to Javascript, but not to Lua. https://github.com/melsman/mlkit/blob/master/README_SMLTOJS.md https://github.com/melsman/mlkit/blob/master/README_SMLTOJS....

dybber 2y ago
Which in turn compiles itself into Javascript, such that you can run the SMLtoJS compiler in the browser, e.g. it should be possible to make it accept <script> tags with SML code.

programjames 2y ago

I got tricked up by the acronym. This compiles the programming language "Standard ML [Meta Language]" to the programming languages Lua or JavaScript. It's not a machine learning compiler.

hayley-patton 2y ago
I can no longer google "ML Kit" and find the ML compiler, instead I find Google's ML library. You win some, you lose some.
zem 2y ago
those of us interested in the ML language family have been disappointed for years every time a headline turns out to be yet another machine learning thing, so I'm glad to see this one was actually an ML (:
fortyseven 2y ago
That's okay, I thought it was machine language at first.

_delirium 2y ago

Blog post announcing it: https://minoki.github.io/posts/2023-12-17-lunarml-release.html https://minoki.github.io/posts/2023-12-17-lunarml-release.ht...

sgt_bilko 2y ago

On work laptop (Windows) at the moment. Will get to it in the weekend, but can some kind soul tell me: is the lua and js generated code readable? (Standard for readable for me is ReScript)

weatherlight 2y ago

I've always loved Standard ML and feel it's never gotten the love that it deserved. It was my introduction to typed FP concepts via this course (Part A) https://www.coursera.org/learn/programming-languages https://www.coursera.org/learn/programming-languages

mattpallissard 2y ago
Agreed! But OCaml is close enough that I don't despair. Chris okasaki's purely functional data structures is an excellent intro to SML if your familiar with FP already.
- weatherlight 2y ago
  I've heard good things about this book, I didn't realize the examples were in SML. :)
Hirrolot 2y ago
Is there any reason to use SML instead of OCaml in 2024?
- SassyBird 2y ago
  Multicore programming that isn’t covered in stickers about being bleeding edge and not fully integrated with the rest of stdlib(s?). Since the 90s. :) And a minor point: operator arguments (as well as curried arguments) are evaluated left-to-right, not right-to-left like in OCaml.
- someplaceguy 2y ago
  > Is there any reason to use SML instead of OCaml in 2024? * Many compilers and transpilers available. Don't like LunarML? Maybe try Poly/ML. Or MLton. Or MLKit. Or SML/NJ. Etc. All of the above are pretty much complete and production quality, and there are quite a few more. * The language has a formal specification which means there's a document formally specifying what all the constructs in the language should do. In contrast, if you wanted to build your own OCaml compiler/transpiler/interpreter, you'd basically have to follow whatever the main OCaml implementation does, which changes from one version to the next, so you'd always be playing catch-up. * The formal specification hasn't changed since 1997. This means that once written, your SML code should work the same forever. You no longer have to worry about your programs (or their dependencies) breaking when you upgrade to a new version of OCaml. * Compared to OCaml, your single-threaded programs will usually run faster if you compile them with MLton, mostly due to whole-program optimization / monomorphization and unboxed native integers, reals and arrays. * Some SML implementations have interesting extensions, which you can use (at the cost of being tied to that implementation). For example, MLKit supports memory regions (i.e. more efficient memory management in some cases). SML# (which is not related to .NET) supports seamless integration with SQL and other interesting features. Some implementations support multi-threading natively and/or different forms of parallelism. Most of them also support some form of FFI to interface with C or other languages. * SML programs are easier to port to CakeML than any other language, since CakeML is mostly an SML subset. CakeML not only allows you to formally verify that your program is correct (or that it has certain desirable properties), but it also has a compiler that is formally proven to be correct, so your compiled CakeML programs are (mostly) guaranteed to be free of miscompilation bugs as well.

psd1 2y ago

I've had a great time discovering Fable, which compiles F# to JS or TS. The cost of moving from VSCode to neovim has always put me off, but if I could write the config in F# it would tempt me more. How far is standard ML from F#? Is there any existing compiler for F# => Lua?

lysecret 2y ago
Have you been using Fable? What has been your experience?
- psd1 2y ago
  Not in production. As an F# newbie, I hit some incompatibility early on with my initial library choices. Not everything works in fable. When I found Safe Stack, that went away. Then it was great. Hot reload; generated js is passable; I'm doing front end in F# with zero JS and almost no html.
  - JaggerJo 2y ago
    I’ve used Fable in production. It just works most of the time, but it’s not seamless.
humzashahid98 2y ago
Standard ML is pretty simple and easy to pick up, but a weak point is the lack of { record with updatedValue } syntax to create a new record which is the same as an existing one but with some value changed. You have to specify every field in the record which can be a pain depending on how large your record is. Standard ML has some nice features though and I think what I mentioned is the only serious shortcoming (except for ecosystem if that's important to you). Standard ML is also structurally (but strongly) typed, unlike F# and the vast majority of nominally typed languages, which is rare but nice. type person = { age : int, name : string } is the same as the type type animal = { age : int, name = string } but this hasn't been an issue foe me, and structural typing means you can avoid defining the type at all which is convenient for small records. For example, a function to increment age has no need for you to define the type. fun grow { name, age } = { name = name, age = age + 1 } I think Standard ML helps you understand the origin of some common FP idioms. Like tuples are really just records and SML treats them as such. The empty record {} is the same as the empty tuple (): both are the unit type. And a record like { 1 : "a", 2 : "b" } is exactly the same as the tuple ("a", "b").

ihumanable 2y ago

If you haven't taken a plunge into Standard ML or OCaml and are interested, I can't suggest this playlist enough https://youtu.be/MUcka_SvhLw?si=rc4ele_0WcjwLFR3 https://youtu.be/MUcka_SvhLw?si=rc4ele_0WcjwLFR3 It's such a lovely introduction to Ocaml and programming in general and has a free online textbook. It's a lot of videos but each one is 5-10 minutes so it's very easy to hop in and out.