This is a Lisp interpreter, written in Rust, intended to be embeddable as a library in a larger application for scripting purposes. Goals:
- Small footprint (both code size and memory usage)
- No runtime dependencies [1]
- Easy, ergonomic interop with native Rust functions
- Small but practical set of Lisp functionality
[1] cfg-if
is build-time, num-traits
add (I believe) no runtime presence,
and num-bigint
is entirely opt-in (at build time)
[dependencies]
rust_lisp = "0.18.0"
use std::{cell::RefCell, rc::Rc};
use rust_lisp::default_env;
use rust_lisp::parser::parse;
use rust_lisp::interpreter::eval;
fn main() {
// create a base environment
let env = Rc::new(RefCell::new(default_env()));
// parse into an iterator of syntax trees (one for each root)
let mut ast_iter = parse("(+ \"Hello \" \"world!\")");
let first_expression = ast_iter.next().unwrap().unwrap();
// evaluate
let evaluation_result = eval(env.clone(), &first_expression).unwrap();
// use result
println!("{}", &evaluation_result);
}
As you can see, the base environment is managed by the user of the library, as
is the parsing stage. This is to give the user maximum control, including
error-handling by way of Result
s.
The heart of the model is Value
, an enum encompassing every type of valid Lisp
value. Most of these are trivial, but Value::List
is not. It holds a recursive
List
data structure which functions internally like a linked-list.
into_iter()
and from_iter()
have been implemented for List
, and there is
also a lisp!
macro (see below) which makes working with Lists, in particular,
much more convenient.
Value
does not implement Copy
because of cases like Value::List
, so if you
read the source you'll see lots of value.clone()
. This almost always amounts
to copying a primitive, except in the Value::List
case where it means cloning
an internal Rc
pointer. In all cases, it's considered cheap enough to do
liberally.
The base environment is managed by the user of the library mainly so that it can
be customized. default_env()
prepopulates the environment with a number of
common functions, but these can be omitted (or pared down) if you wish. Adding
an entry to the environment is also how you would expose your Rust functions to
your scripts, which can take the form of either regular functions or closures:
fn my_func(env: Rc<RefCell<Env>>, args: &Vec<Value>) -> Result<Value, RuntimeError> {
println!("Hello world!");
return Ok(Value::NIL);
}
...
env.borrow_mut().define(
Symbol::from("sayhello"),
Value::NativeFunc(my_func)
);
env.borrow_mut().define(
Symbol::from("sayhello"),
Value::NativeFunc(
|env, args| {
println!("Hello world!");
return Ok(Value::NIL);
})
);
In either case, a native function must have the following function signature:
type NativeFunc = fn(env: Rc<RefCell<Env>>, args: &Vec<Value>) -> Result<Value, RuntimeError>;
The first argument is the environment at the time and place of calling (closures
are implemented as environment extensions). The second argument is the Vec of
evaluated argument values. For convenience, utility functions (require_arg()
,
require_int_parameter()
, etc) have been provided for doing basic argument
retrieval with error messaging. See default_environment.rs
for examples.
A Rust macro, named lisp!
, is provided which allows the user to embed
sanitized Lisp syntax inside their Rust code, which will be converted to an AST
at compile-time:
fn parse_basic_expression() {
let ast = parse("
(list
(* 1 2) ;; a comment
(/ 6 3 \"foo\"))").next().unwrap().unwrap();
assert_eq!(ast, lisp! {
(list
(* 1 2)
(/ 6 3 "foo"))
});
}
Note that this just gives you a syntax tree (in the form of a Value
). If you
want to actually evaluate the expression, you would need to then pass it to
eval()
.
The macro also allows Rust expressions (of type Value
) to be embedded within
the lisp code using { }
:
fn parse_basic_expression() {
let ast = parse("
(+ 3 1)").next().unwrap().unwrap();
let n = 2;
assert_eq!(ast, lisp! {
(+ { Value::Int(n + 1) } 1)
});
}
NOTE: When parsing lisp code from a string, dashes (-
) are allowed to be used
in identifiers. However, due to the limitations of declarative Rust macros,
these cannot be handled correctly by lisp! {}
. So it's recommended that you
use underscores in your identifiers instead, which the macro will be able to
handle correctly. The built-in functions follow this convention.
NOTE 2: The macro cannot handle the syntax for negative numbers! To get around this you can insert negative numbers as Rust expressions using the escape syntax, or you can parse your code as a string.
Sometimes if you're wanting to script an existing system, you don't want to convert your data to and from lisp-compatible values. This can be tedious, and inefficient.
If you have some type - say a struct - that you want to be able to work with
directly from your lisp code, you can place it in a Value::Foreign()
which
allows lisp code to pass it around and (native) lisp functions to operate on it:
struct Foo {
some_prop: f32,
}
let v: Value = Value::Foreign(Rc::new(Foo { some_prop: 1.0 }));
Special forms: define
, set
, defun
, defmacro
, lambda
, quote
, let
,
begin
, cond
, if
, and
, or
Functions (in default_env()
): print
, is_null
, is_number
, is_symbol
,
is_boolean
, is_procedure
, is_pair
, car
, cdr
, cons
, list
, nth
,
sort
, reverse
, map
, filter
, length
, range
, hash
, hash_get
,
hash_set
, +
, -
, *
, /
, truncate
, not
, ==
, !=
, <
, <=
, >
,
>=
, apply
, eval
Other features:
- Quoting with comma-escapes
- Lisp macros
- Tail-call optimization