9 releases

✓ Uses Rust 2018 edition

0.3.4 Nov 20, 2019
0.3.3 Nov 7, 2019
0.3.1 Sep 16, 2019
0.2.2 Sep 1, 2019
0.1.1 May 23, 2019

#68 in Parser tooling

Download history 17/week @ 2019-08-05 2/week @ 2019-08-12 16/week @ 2019-08-26 25/week @ 2019-09-02 23/week @ 2019-09-09 115/week @ 2019-09-16 19/week @ 2019-09-23 18/week @ 2019-09-30 3/week @ 2019-10-07 19/week @ 2019-10-14 45/week @ 2019-10-21 10/week @ 2019-10-28 37/week @ 2019-11-04 12/week @ 2019-11-11

114 downloads per month

GPL-3.0-or-later

420KB
7.5K SLoC

WLambda - Embeddable Scripting Language for Rust

This crate provides you with a small and simple embeddable scripting language. It's syntax gravitates around functions and argument composition for functions. A core concept is, that everything is callable. It could be viewed as LISP without parenthesis. Or as a mixture of Perl, JavaScript and LISP/Scheme.

Here are some of it's properties:

  • Simple syntax. For a reference look at the WLambda Language Reference and the parser.
  • Easily embeddable into Rust programs due to a simple API.
  • It's about getting things done quickly, so performance is not a main priority. Current performance is roughly in the ball park of (C)Python. Which means, it's too slow if you need speed. But fast enough if you are not primarily concerned about speed.
  • No garbage collector. Garbage collection relies only on reference counting.
  • Main data structures are Lists and Maps.
  • No exceptions, except panics. Error handling is accomplished by a specialized data type. It can be thought of as dynamic counterpart of Rust's Result type.
  • Easy maintenance of the implementation.
  • Custom user data implementation using VValUserData.

The embedding API relies on a data structure made of VVal nodes.

Here you can find the WLambda Language Reference.

Example WLambda Code

Just a quick glance at the WLambda syntax and semantics.

More details for the syntax and the provided global functions can be found in the WLambda Language Reference.

# This is a comment

# Definition:
!a = 10;

# Assignment:
.a = 20;

# List variable definition:
!a_list = $[1, 2, 3, 4];

# Map assignment:
!a_map = ${a = 10, b = 20};

# Function definition/assignment:
!a_func = {
    _ + _1  # Arguments are not named, they are put into _, _1, _2
};

a_func[2, 3];   # Function call
a_func 2 3;     # Equivalent function call

# Shortened one statement function definition:
!do_something_to = \_ * 2;

# There is no `if` statement. Booleans can be called
# with two arguments. The first one is called when the boolean
# is true, the second one is called when the boolean is false.
(a == 10) {
    # called if a == 10
} {
    # called if a != 10
};

# Counting loop:
!sum = $&0; # Defining a reference that can be assignment
            # from inside a function.

# `range` calls the given function for each iteration
# and passes the counter as first argument in `_`
range 0 10 1 { # This is a regular function.
    .*sum = $*sum + _; # $* is a dereferencing operator
                       # and .* starts a reference assignment
};

# `range` loop with `break`
!break_value = range 0 10 1 {
    (_ == 5) { break 22 };
};

# Returning early from functions:
!some_fun = \:some_fun_lbl { # \:xxx defines a function label for returning
    !x = 10;
    .x = do_something_to x;
    (x > 20) {
        return :some_fun_lbl 20; # explicit argument for return returns from
                                 # the specified block.
    }
    .x = 20;
    x
};

# `return` implicitly jumps to the topmost $nul label
# you may specify a small unused label like `_` to jump out some unnamed func:
!some_fun = {
    !(x) = @;
    (x == 20) \:_{ return 30 } # returns from some_fun, not from the if-branch
};

# Error reporting:
    # There are special error values, that will make the program panic
    # if they are not handled correctly at statement block level:
    !some_erroring_func = {
        return $error "An error happened!"
    };
    !value = some_erroring_func[];
    # on_error calls the first argument if the second argument
    # is an error value.
    on_error {
        # handle error here, eg. report, or make a new error value
        !(err_value, line, col, file) = @;
        std:displayln err_value;
    } value;

    !handle_err = { std:displayln _ };

    # with the ~ operator, you can chain it nicely:
    on_error {|| handle_err[_] } ~ some_erroring_func[];
    # or without ~:
    on_error {|| handle_err[_] } (some_erroring_func[]);
    # or with |
    some_erroring_func[] | on_error {|| handle_err[_] };

    # _? transforms an error value, and returns it from the current
    #    function. optionally jumping outwards.

    std:assert_eq (str {
        _? ~ $e "ok"; # is with an error value the same as: `return $e "ok"`
    }[]) "$e[98,17:<wlambda::eval>(Err)] \"ok\"";

    _? 10; # passes the value through

!report_my_error = { std:displayln _ };

!some_erroring_func = {
    on_error {
        report_my_error _;
    } block :outer {
        # do something...
        (_ != 10) {
            return :outer $error "Something really failed"
            # same as, with the difference, that _? only returns
            # from :outer if it is an error value.
            _? :outer $error "Something really failed"
        }
        # do more ...
    }
    # cleanup ...
};

# Basic OOP:
# std:weaken to make any closure capture of some_obj a weak reference, so
# we don't get any cyclic references:
!some_obj = $&${};
some_obj.do_something = {
    # do something here with some_obj captured (weakly)
    # from the upper lexical scope.
};
some_obj.do_something[]; # Method call

Currently there are many more examples in the test cases in compiler.rs.

API Usage Examples

Basic API Usage

Here is how you can quickly evaluate a piece of WLambda code:

let s = "$[1,2,3]";
let r = wlambda::compiler::eval(&s).unwrap();
println!("Res: {}", r.s());

More Advanced API Usage

If you want to quickly add some of your own functions, you can use the GlobalEnv add_func method:

use wlambda::vval::{VVal, VValFun, Env};

let global_env = wlambda::GlobalEnv::new_default();
global_env.borrow_mut().add_func(
    "my_crazy_add",
    |env: &mut Env, _argc: usize| {
        Ok(VVal::Int(
              env.arg(0).i() * 11
            + env.arg(1).i() * 13
        ))
    }, Some(2), Some(2));

let mut ctx = wlambda::compiler::EvalContext::new(global_env);

// Please note, you can also add functions later on,
// but this time directly to the EvalContext:

ctx.set_global_var(
    "my_crazy_mul",
    &VValFun::new_fun(|env: &mut Env, _argc: usize| {
       Ok(VVal::Int(
          (env.arg(0).i() + 11)
        * (env.arg(1).i() + 13)))
    }, Some(2), Some(2)));


let res_add : VVal = ctx.eval("my_crazy_add 2 4").unwrap();
assert_eq!(res_add.i(), 74);

let res_mul : VVal = ctx.eval("my_crazy_mul 2 4").unwrap();
assert_eq!(res_mul.i(), 221);

Possible Roadmap

There are several things that can be added more or less easily to WLambda. But I am currently working on making the language more complete for real world use. So my current goals are:

  • Add namespacing and importing for managing the global environment.
  • Make namespaces for utility functions in the areas:
    • List handling
    • Map handling
    • Iteration
    • Basic I/O for testing purposes (WLambda is for embedding, there are currently no goals to provide a binary beyond basic needs.)
  • Improve and further document the VVal API for interacting with WLambda.
  • Make VVal::Sym hold an interned string instead of a String instance.

Future plans could be:

  • Prototyped inheritance, sketched out like this:

        !proto = ${ print = { std:displayln _ }, };
        !o = to_obj ${ _proto_ = proto };
        o.print(123);
    
        # MetaMap(Rc<RefCell<std::collections::HashMap<String, VVal>>>),
        # => invokes _proto_ lookup on field access (not write)
    
  • Augment functions with tagged values:

        !tag = 123;
        !v = tag 10 tag;
        !fun = { println("not tagged!") };
        .fun = add_tag fun tag { println("tagged with 123"); }
        fun[v]; # prints "tagged with 123"
        fun[10]; # prints "not tagged!"
    
        # TagFun(Rc<RefCell<std::collections::HashMap<String, Rc<VValFun>>>>),
    
  • There are currently no plans to change the internal evaluator from a closure tree to a VM and/or JIT speedup. However, help is appreachiated if someone is able to significantly speed up the evaluation without too much breakage.

License

This project is licensed under the GNU General Public License Version 3 or later.

Why GPL?

Picking a license for my code bothered me for a long time. I read many discussions about this topic. Read the license explanations. And discussed this matter with other developers.

First about why I write code for free at all:

  • It's my passion to write computer programs. In my free time I can write the code I want, when I want and the way I want. I can freely allocate my time and freely choose the projects I want to work on.
  • To help a friend or member of my family.
  • To solve a problem I have.

Those are the reasons why I write code for free. Now the reasons why I publish the code, when I could as well keep it to myself:

  • So that it may bring value to users and the free software community.
  • Show my work as an artist.
  • To get into contact with other developers.
  • And it's a nice change to put some more polish on my private projects.

Most of those reasons don't yet justify GPL. The main point of the GPL, as far as I understand: The GPL makes sure the software stays free software until eternity. That the user of the software always stays in control. That the users have at least the means to adapt the software to new platforms or use cases. Even if the original authors don't maintain the software anymore. It ultimately prevents "vendor lock in". I really dislike vendor lock in, especially as developer. Especially as developer I want and need to stay in control of the computers I use.

Another point is, that my work has a value. If I give away my work without any strings attached, I effectively work for free. Work for free for companies. I would compromise the price I can demand for my skill, workforce and time.

This makes two reasons for me to choose the GPL:

  1. I do not want to support vendor lock in scenarios. At least not for free. I want to prevent those when I have a choice. And before you ask, yes I work for a company that sells closed source software. I am not happy about the closed source fact. But it pays my bills and gives me the freedom to write free software in my free time.
  2. I don't want to low ball my own wage and prices by giving away free software with no strings attached (for companies).

If you need a permissive or private license (MIT)

Please contact me if you need a different license and really want to use my code. As long as I am the only author, I can change the license. We might find an agreement.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in WLambda by you, shall be licensed as GPLv3 or later, without any additional terms or conditions.

Authors

Dependencies

~4.5MB
~93K SLoC