Rust ❤️ Emacs

A community-driven port of Emacs to Rust.

GPLv3 license.

Table of Contents

Rust ❤️ Emacs

Why Emacs?

Emacs will change how you think about programming.

Emacs is totally introspectable. You can always find out 'what code runs when I press this button?'.

Emacs is an incremental programming environment. There's no edit-compile-run cycle. There isn't even an edit-run cycle. You can execute snippets of code and gradually turn them into a finished project. There's no distinction between your editor and your interpreter.

Emacs is a mutable environment. You can set variables, tweak functions with advice, or redefine entire functions. Nothing is off-limits.

Emacs provides functionality without applications. Rather than separate applications, functionality is all integrated into your Emacs instance. Amazingly, this works. Ever wanted to use the same snippet tool for writing C++ classes as well as emails?

Emacs is full of incredible software concepts that haven't hit the mainstream yet. For example:

Many platforms have a single item clipboard. Emacs has an infinite clipboard.
If you undo a change, and then continue editing, you can't redo the original change. Emacs allows undoing to any historical state, even allowing tree-based exploration of history.
Emacs supports a reverse variable search: you can find variables with a given value.
You can perform structural editing of code, allowing you to make changes without breaking syntax. This works for lisps (paredit) and non-lisps (smartparens).
Many applications use a modal GUI: for example, you can't do other edits during a find-and-replace operation. Emacs provides recursive editing that allow you to suspend what you're currently doing, perform other edits, then continue the original task.

Emacs has a documentation culture. Emacs includes a usage manual, a lisp programming manual, pervasive docstrings and even an interactive tutorial.

Emacs has a broad ecosystem. If you want to edit code in a niche language, there's probably an Emacs package for it.

Emacs doesn't have a monopoly on good ideas, and there are other great tools out there. Nonetheless, we believe the Emacs learning curve pays off.

Why Rust?

Rust is a great alternative to C.

Rust has a fantastic learning curve. The documentation is superb, and the community is very helpful if you get stuck.

Rust has excellent tooling. The compiler makes great suggestions, the unit test framework is good, and rustfmt helps ensure formatting is beautiful and consistent.

The Rust packaging story is excellent. It's easy to reuse the great libraries available, and just as easy to factor out code for the benefit of others. We can replace entire C files in Emacs with well-maintained Rust libraries.

Code written in Rust easily interoperates with C. This means we can port to Rust incrementally, and having a working Emacs at each step of the process.

Rust provides many compile-time checks, making it much easier to write fast, correct code (even when using multithreading). This also makes it much easier for newcomers to contribute.

Give it a try. We think you'll like it.

Why A Fork?

Emacs is a widely used tool with a long history, broad platform support and strong backward compatibility requirements. The core team is understandably cautious in making far-reaching changes.

Forking is a longstanding tradition in the Emacs community for trying different approaches. Notable Emacs forks include XEmacs, Guile Emacs, and emacs-jit.

There have also been separate elisp implementations, such as Deuce, JEmacs and El Compilador.

By forking, we can explore new development approaches. We can use a pull request workflow with integrated CI.

We can drop legacy platforms and compilers. Remacs will never run on MS-DOS, and that's OK.

There's a difference between the idea of Emacs and the current implementation of Emacs. Forking allows us to explore being even more Emacs-y.

Getting Started

Requirements

You will need Rust installed. Remacs uses unstable Rust features, so you will need to use nightly. The file rust-toolchain indicates the correct version to install; rustup will use that version as long as you are in the Remacs repository
```
 cd /path/to/remacs
 rustup install `cat rust-toolchain`
```
You will need a C compiler and toolchain. On Linux, you can do something like apt-get install build-essential automake. On macOS, you'll need Xcode.
You will need some C libraries. On Linux, you can install everything you need with:
```
 apt-get install texinfo libjpeg-dev libtiff-dev \
   libgif-dev libxpm-dev libgtk-3-dev libgnutls-dev \
   libncurses5-dev libxml2-dev
```
On macOS, you'll need libxml2 (via xcode-select --install) and gnutls (via brew install gnutls).

On macOS, the default makeinfo command in outdated, you'll need to update it (via brew install texinfo). To use the installed version of makeinfo instead of the built in (/usr/bin/makeinfo) one, you'll need to make sure /usr/local/opt/texinfo/bin is before /usr/bin in PATH.

Dockerized development environment

If you don't want to bother with the above setup you can use the provided docker environment. Make sure you have docker 1.12+ and docker-compose 1.8+ available.

To spin up the environment run

docker-compose up -d

First time you run this command docker will build the image. After that any subsequent startups will happen in less than a second.

The working directory with remacs will be mounted under the same path in the container so editing the files on your host machine will automatically be reflected inside the container. To build remacs use the steps from Building Remacs prefixed with docker-compose exec remacs, this will ensure the commands are executed inside the container.

Building Remacs

$ ./autogen.sh
$ ./configure --enable-rust-debug
$ make

For a release build, don't pass --enable-rust-debug.

The Makefile obeys cargo's RUSTFLAGS variable and additional options can be passed to cargo with CARGO_FLAGS.

For example:

$ make CARGO_FLAGS="-vv" RUSTFLAGS="-Zunstable-options --pretty"

Running Remacs

You can now run your shiny new Remacs build!

# Using -q to ignore your .emacs.d, so Remacs starts up quickly.
# RUST_BACKTRACE is optional, but useful if your instance crashes.
$ RUST_BACKTRACE=1 src/remacs -q

Rustdoc builds

You can use rustdoc to generate API docs:

# http://stackoverflow.com/a/39374515/509706
$ cargo rustdoc -- \
    --no-defaults \
    --passes strip-hidden \
    --passes collapse-docs \
    --passes unindent-comments \
    --passes strip-priv-imports

You can then open these docs with:

$ cargo doc --open

Porting Elisp Primitive Functions: Walkthrough

Let's look at porting numberp to Rust.

First, make sure you have configured and built Remacs on your system. You'll probably want to generate TAGS too, so you can jump to definitions of C functions.

This is the definition of numberp:

DEFUN ("numberp", Fnumberp, Snumberp, 1, 1, 0,
       doc: /* Return t if OBJECT is a number (floating point or integer).  */
       attributes: const)
  (Lisp_Object object)
{
  if (NUMBERP (object))
    return Qt;
  else
    return Qnil;
}

The DEFUN macro, in addition to defining a function Fnumberp, also creates a static struct Snumberp that describes the function for Emacs' Lisp interpreter.

In Rust, we define a numberp function that does the actual work then use an attribute (implemented as a procedural macro) named lisp_fn that handles these definitions for us:

// This is the function that gets called when
// we call numberp in elisp.
//
// `lisp_fn` defines a wrapper function that calls numberp with
// LispObject values. It also declares a struct that we can pass to
// defsubr so the elisp interpreter knows about this function.

/// Return t if OBJECT is a number.
#[lisp_fn]
fn numberp(object: LispObject) -> LispObject {
    LispObject::from_bool(object.is_number())
}

Additionally, lisp_fn can automatically translate LispObjects passed in as arguments into native Rust types:

// This function takes a double, and can also take an integer.
#[lisp_fn(min = "1")]
pub fn sleep_for(seconds: EmacsDouble, milliseconds: Option<EmacsInt>) -> LispObject {
    let duration = seconds + (milliseconds.unwrap_or(0) as f64 / 1000.0);
    if duration > 0.0 {
        // … etc
    }
}

Similarly, lisp_fn can automatically translate the return type:

#[lisp_fn(min = "1")]
pub fn atan(y: EmacsDouble, x: Option<EmacsDouble>) -> EmacsDouble {
    match x {
        None => y.atan(),
        Some(x) => y.atan2(x)
    }
}

The automatic translation signals a Lisp argument-type error if it sees an argument of the wrong type. LispObjects are therefore still the correct choice for functions which can handle disparate argument types (such as one that takes either a buffer object or a string containing a buffer name), or doesn't want to signal an error. Similarly, LispObject is still the correct choice of return type for functions which may return different types in different calls.

The elisp name of the function is derived from the Rust name, with underscores replaced by hyphens. If that is not possible (like for the function +), you can give an elisp name as an argument to lisp_fn, like #[lisp_fn(name = "+")].

Optional arguments are also possible: to make the minimum number of arguments from elisp different from the number of Rust arguments, pass a min = "n" argument.

The docstring of the function should be the same as the docstring in the C code. (Don't wonder about it being a comment there, Emacs has some magic that extracts it into a separate file.)

Finally, delete the old C definition.

You're done! Compile Remacs, try your function with M-x ielm, and open a pull request. Fame and glory await!

Porting Widely Used C Functions

If your Rust function replaces a C function that is used elsewhere in the C codebase, it needs to be exported. If the function is not a Lisp function (i.e. doesn't use the #[lisp_fn] macro), you need to manually mark it as #[no_mangle] and extern "C" to be exported with the correct ABI.

Source code style guide

In order to pass Travis checks on pull requests, the source has to be formatted according to the default style of rustfmt, as packaged with the Rust nightly in rust-toolchain. To do that, install rustfmt:

$ rustup component add rustfmt-preview

Make sure you uninstall the crate version of rustfmt first; The new component will install its own set of binaries.

$ cargo uninstall rustfmt
$ cargo uninstall rustfmt-nightly

Then you can run this in the checkout root to reformat all Rust code:

$ make rustfmt

Running tests

Run elisp and Rust tests in toplevel directory. If run in a subdirectory, only run the tests in that directory.

make check Run all tests as defined in the directory. Expensive tests are suppressed. The result of the tests for .el is stored in .log.
make check-maybe Like "make check", but run only the tests for files that have been modified since the last build.

Design Goals

Compatibility: Remacs should not break existing elisp code, and ideally provide the same FFI too.

Similar naming conventions: Code in Remacs should use the same naming conventions for elisp namespaces, to make translation straightforward.

This means that an elisp function do-stuff will have a corresponding Rust function Fdo_stuff, and a declaration struct Sdo_stuff. A lisp variable do-stuff will have a Rust variable Vdo_stuff and a symbol 'do-stuff will have a Rust variable Qdo_stuff.

Otherwise, we follow Rust naming conventions, with docstrings noting equivalent functions or macros in C. When incrementally porting, we may define Rust functions with the same name as their C predecessors.

Leverage Rust itself: Remacs should make best use of Rust to ensure code is robust and performant.

Leverage the Rust ecosystem: Remacs should use existing Rust crates wherever possible, and create new, separate crates where our code could benefit others.

Great docs: Emacs has excellent documentation, Remacs should be no different.

Non-Design Goals

etags: The universal ctags project supports a wider range of languages and we recommend it instead.

Contributing

Pull requests welcome, no copyright assignment required. This project is under the Rust code of conduct.

Help Needed

There's lots to do! We keep a list of low hanging fruit here so you can easily choose one. If you do, please open a new issue to keep track of the task and link to it.

Easy tasks:

Find a small function in lisp.h and write an equivalent in lisp.rs.
Add Rust unit tests. Currently we're relying on Emacs' own test suite.
Add docstrings to public functions in lisp.rs.
Tidy up messy Rust that's been translated directly from C. Run rustfmt, add or rename internal variables, run clippy, and so on.
Add Rust-level unit tests to elisp functions defined in lib.rs.

Medium tasks:

Choose an elisp function you like, and port it to rust. Look at rust-mod for an example.
Teach describe-function to find functions defined in Rust.
Expand our Travis configuration to run 'make check', so we know remacs passes Emacs' internal test suite.
Expand our Travis configuration to ensure that Rust code has been formatted with rustfmt
Set up bors/homu.
Set up a badge tracking pub struct/function coverage using cargo-doc-coverage.
Search the Rust source code for TODO comments and fix them.

Big tasks:

Find equivalent Rust libraries for parts of Emacs, and replace all the relevant C code. Rust has great libraries for regular expressions, GUI, terminal UI, managing processes, amongst others.
Change the elisp float representation to use nan boxing rather than allocating floats on the heap.

Rust Porting Tips

C Functions

When writing a Rust version of a C function, give it the same name and same arguments. If this isn't appropriate, docstrings should say the equivalent C function to help future porters.

For example, make_natnum mentions that it can be used in place of XSETFASTINT.

C Macros

For C macros, we try to define a fairly equivalent Rust function. The docstring should mention the original macro name.

Since the Rust function is not a drop-in replacement, we prefer Rust naming conventions for the new function.

For the checked arithmetic macros (INT_ADD_WRAPV, INT_MULTIPLY_WRAPV and so on), you can simply use .checked_add, .checked_mul from the Rust stdlib.

Assertions

eassert in Emacs C should be debug_assert! in Rust.

emacs_abort() in Emacs C should be panic!("reason for panicking") in Rust.

Safety

LispObject values may represent pointers, so the usual safety concerns of raw pointers apply.

If you can break memory safety by passing a valid value to a function, then it should be marked as unsafe. For example:

impl LispObject {
    // This function is unsafe because it's accessing a raw pointer
    // without doing any checking. We assume the current value is a
    // pointer to a string.
    unsafe fn as_string_unchecked(self) -> LispStringRef {
        LispStringRef::new(unsafe { mem::transmute(self.get_untaggedptr()) })
    }

    // This function is safe because it verifies that the pointer is
    // tagged as a string.
    fn as_string_or_error(self) -> LispStringRef {
        if self.is_string() {
            unsafe { self.as_string_unchecked() }
        } else {
            wrong_type!(Qstringp, self)
        }
    }
}

ljos/remacs