jlrs is a crate that provides access to most of the Julia C API, it can be used to embed Julia in Rust applications and to use functionality from the Julia C API when writing ccallable functions in Rust. Currently this crate is only tested on Linux in combination with Julia 1.6 and is not compatible with earlier versions of Julia.
An incomplete list of features that are currently supported by jlrs:
- Access arbitrary Julia modules and their contents.
- Call Julia functions, including functions that take keyword arguments.
- Exceptions can be handled or converted to their error message, optionally with color.
- Include and call your own Julia code.
- Use a custom system image.
- Create values that Julia can use, and convert them back to Rust, from Rust.
- Access the type information and fields of values. The contents of inline and bits-union fields can be accessed directly.
- Create and use n-dimensional arrays. The
jlrs-ndarrayfeature can be enabled for integration with ndarray. - Support for mapping Julia structs to Rust structs that can be generated by JlrsReflect.jl.
- Structs that can be mapped to Rust include those with type parameters and bits unions.
- An async runtime is available when the
asyncfeature is enabled, which can be used from multiple threads and supports scheduling JuliaTasks andawaiting the result without blocking the runtime.
Add this to your Cargo.toml:
[dependencies]
jlrs = "0.12"This crate depends on jl-sys which contains the raw bindings to the Julia C API, by default pregenerated bindings are used. If you want to generate the bindings at compile time, the use-bindgen feature must be enabled. In this case the bindings are generated by bindgen. You can find the requirements for using bindgen in their User Guide.
The recommended way to install Julia is to download the binaries from the official website, which is distributed in an archive containing a directory called julia-x.y.z. This directory contains several other directories, including a bin directory containing the julia executable.
In order to ensure the julia.h header file can be found, either /usr/include/julia/julia.h or /usr/local/include/julia/julia.h must exist, or you have to set the JULIA_DIR environment variable to /path/to/julia-x.y.z. This environment variable can be used to override the default. Similarly, in order to load libjulia.so you must add /path/to/julia-x.y.z/lib to the LD_LIBRARY_PATH environment variable. When the uv feature is enabled, /path/to/julia-x.y.z/lib/julia must also be added to LD_LIBRARY_PATH.
Support for Windows was dropped in jlrs 0.10 due to compilation and dependency issues. If you want to use jlrs on Windows you must use WSL. An installation guide to install WSL on Windows can be found on Microsoft's website. After installing a Linux distribution, follow the installation instructions for Linux.
The first thing you should do is use the prelude-module with an asterisk, this will bring all the structs and traits you're likely to need into scope. When embedding Julia, it must be initialized before it can be used. You can do this by calling Julia::init which returns an instance of Julia. Note that this method can only be called once while the application is running; if you drop it you won't be able to create a new instance but have to restart the application. If you want to use a custom system image, you must call Julia::init_with_image instead of Julia::init. If you're calling Rust from Julia everything has already been initialized, you can use CCall instead. If you want to use the async runtime, one of the initialization methods of AsyncJulia must be used.
With the Julia instance it's possible to include files containing Julia code by calling Julia::include. In order to actually create values that Julia can use and call Julia functions, Julia::scope must be used. This method takes a closure that takes two arguments, a Global and a mutable reference to a GcFrame. A Global is used to access Julia modules, the GcFrame is used to ensure Julia data isn't freed by the garbage collector while references to that data can be used from Rust.
Because you can use both a Global and a mutable reference to a GcFrame inside the closure, it's possible to access the contents of modules and create new values that can be used by Julia. The methods of Module let you access the contents of arbitrary modules, several methods are available to create new values.
The simplest is to call Value::eval_string, a method that takes two arguments. The first must implement the Scope trait, the second is a string which has to contain valid Julia code. The most important thing to know about the Scope trait for now is that it's used by functions that create new values to ensure the result is rooted. Mutable references to GcFrames implement Scope, in this case the Value that is returned is rooted in that frame, so the result is protected from garbage collection until the frame is dropped when that scope ends.
In practice, Value::eval_string is relatively limited. It can be used to evaluate simple function calls like sqrt(2.0), but can't take any arguments. Its most important use-case is importing installed packages by evaluating an import or using statement. A more interesting method, Value::new, can be used with data of any type that implements IntoJulia. This trait is implemented by primitive types like i8 and char. Any type that implements IntoJulia also implements Unbox which is used to extract the contents of a Julia value.
In addition to evaluating raw commands with Value::eval_string, it's possible to call anything that implements Call as a Julia function. Value implements this trait because any Julia value is potentially callable as a function. Functions can be called with any number of positional arguments and be provided with keyword arguments. Both Value::eval_string and the trait methods of Call are unsafe because it's trivial to write a function like boom() = unsafe_load(Ptr{Float64}(C_NULL)), which causes a segfault when it's called, and call it with these methods.
As a simple example, let's convert two numbers to Julia values and add them:
use jlrs::prelude::*;
fn main() {
// Initializing Julia is unsafe because it can race with another crate that does
// the same.
let mut julia = unsafe { Julia::init().unwrap() };
let res = julia.scope(|global, frame| {
// Create the two arguments. Note that the first argument, something that
// implements Scope, is taken by value and mutable references don't implement
// Copy, so it's necessary to mutably reborrow the frame.
let i = Value::new(&mut *frame, 2u64)?;
let j = Value::new(&mut *frame, 1u32)?;
// The `+` function can be found in the base module.
let func = Module::base(global).function(&mut *frame, "+")?;
// Call the function and unbox the result as a `u64`. The result of the function
// call is a nested `Result`; the outer error doesn't contain to any Julia
// data, while the inner error contains the exception if one is thrown. Here the
// exception is converted to the outer error type by calling `into_jlrs_result`, this new
// error contains the error message Julia would have shown. Colors can be enabled by
// calling `Julia::error_color`.
unsafe {
func.call2(&mut *frame, i, j)?
.into_jlrs_result()?
.unbox::<u64>()
}
}).unwrap();
assert_eq!(res, 3);
}Many more features are available, including creating and accessing n-dimensional Julia arrays and nesting scopes. To learn how to use them, please see the documentation for the memory and wrappers modules.
Julia's ccall interface can be used to call extern "C" functions defined in Rust, for most use cases you shouldn't need jlrs. There are two major ways to use ccall, with a pointer to the function or a (:function, "library") pair.
A function can be cast to a void pointer and converted to a Value:
use jlrs::prelude::*;
// This function will be provided to Julia as a pointer, so its name can be mangled.
unsafe extern "C" fn call_me(arg: bool) -> isize {
if arg {
1
} else {
-1
}
}
fn main() {
let mut julia = unsafe { Julia::init().unwrap() };
julia.scope(|global, frame| {
// Cast the function to a void pointer
let call_me_val = Value::new(&mut *frame, call_me as *mut std::ffi::c_void)?;
// Value::eval_string can be used to create new functions.
let func = Value::eval_string(
&mut *frame,
"myfunc(callme::Ptr{Cvoid})::Int = ccall(callme, Int, (Bool,), true)"
)?.unwrap();
// Call the function and unbox the result.
let output = func.call1(&mut *frame, call_me_val)?
.into_jlrs_result()?
.cast::<isize>()?;
assert_eq!(output, 1);
Ok(())
}).unwrap();
}You can also use functions defined in dylib and cdylib libraries. In order to create such a library you need to add
[lib]
crate-type = ["dylib"]or
[lib]
crate-type = ["cdylib"]respectively to your crate's Cargo.toml. Use a dylib if you want to use the crate in other Rust crates, but if it's only intended to be called through ccall a cdylib is the better choice. On Linux, compiling such a crate will be compiled to lib<crate_name>.so.
The functions you want to use with ccall must be both extern "C" functions to ensure the C ABI is used, and annotated with #[no_mangle] to prevent name mangling. Julia can find libraries in directories that are either on the default library search path or included by setting the LD_LIBRARY_PATH environment variable on Linux. If the compiled library is not directly visible to Julia, you can open it with Libdl.dlopen and acquire function pointers with Libdl.dlsym. These pointers can be called the same way as the pointer in the previous example.
If the library is visible to Julia you can access it with the library name. If call_me is defined in a crate called foo, the following should workif the function is annotated with #[no_mangle]:
ccall((:call_me, "libfoo"), Int, (Bool,), false)One important aspect of calling Rust from other languages in general is that panicking across an FFI boundary is undefined behaviour. If you're not sure your code will never panic, wrap it with std::panic::catch_unwind.
Most features provided by jlrs including accessing modules, calling functions, and borrowing array data require a Global or a frame. You can access these by creating a CCall first. Another method provided by CCall is CCall::uv_async_send, this method can be used in combination with Base.AsyncCondition. In particular, it lets you write a ccallable function that does its actual work on another thread, return early and wait on the async condition, which happens when CCall::uv_async_send is called when that work is finished. The advantage of this is that the long-running function will not block the Julia runtime, There's an example available on GitHub that shows how to do this.
The async runtime runs Julia in a separate thread and returns a handle that can be shared across threads. The handle can be used to send new tasks to the runtime, multiple tasks can run in parallel by scheduling a function call as a new Julia Task. While the Julia Task has not completed, the runtime can switch to another task. To use this feature you must enable the async feature flag:
[dependencies]
jlrs = { version = "0.12", features = ["async"] }The struct AsyncJulia is exported by the prelude and lets you initialize the runtime in two ways, either as a blocking task or as a thread. The first way should be used if you want to integrate the async runtime into a larger project that uses async_std.
The easiest way to interact with Julia when using the async runtime is by using AsyncJulia::blocking_task, which can be used to send a closure like the one in the first example and call it. While this closure has not completed the runtime is blocked, the methods that schedule a function call as a new Julia Task can't be used.
In order to write non-blocking tasks, you must implement either the AsyncTask or GeneratorTask trait. An AsyncTask can be called once, its async run method replaces the closure; this method takes a Global and a mutable reference AsyncGcFrame. The AsyncGcFrame provides mostly the same functionality as GcFrame, but can also be used to call the methods of the CallAsync trait. These methods schedule the function call on another thread and return a Future. While awaiting the result the runtime can handle another task.
A GeneratorTask can be called multiple times. In addition to run it also has an async init method. This method is called when the GeneratorTask is created and can be used to prepare the initial state of the task. The frame provided to init is not dropped after this method returns, which means this initial state can contain Julia data. Whenever a GeneratorTask is successfully created a GeneratorHandle is returned. This handle can be used to call the GeneratorTask which calls its run method once. A GeneratorHandle can be cloned and shared across threads.
You can find basic examples that show how to implement these traits in the examples directory of the GitHub repository.
The restriction that Julia can be initialized once must be taken into account when running tests that use jlrs. The recommended approach is to create a thread-local static RefCell:
use jlrs::prelude::*;
use std::cell::RefCell;
thread_local! {
pub static JULIA: RefCell<Julia> = {
let julia = RefCell::new(unsafe { Julia::init().unwrap() });
julia.borrow_mut().scope(|_global, _frame| {
/* include everything you need to use */
Ok(())
}).unwrap();
julia
};
}Tests that use this construct can only use one thread for testing, so you must use cargo test -- --test-threads=1, otherwise the code above will panic when a test tries to call Julia::init a second time from another thread.
If these tests also involve the async runtime, the JULIA_NUM_THREADS environment variable must be set to a value larger than 2.
If you want to run jlrs's tests, both these requirements must be taken into account: JULIA_NUM_THREADS=3 cargo test -- --test-threads=1
In order to map a struct in Rust to one in Julia you can derive or implement ValidLayout, Unbox, and Typecheck. If the struct in Julia has no type parameters and is a bits type you can also derive IntoJulia, which lets you use the type in combination with Value::new.
You should normally not need to implement these structs or traits manually. The JlrsReflect.jl package can generate the correct Rust struct and automatically derive the supported traits for types that have no tuple or union fields with type parameters. The reason for this restriction is that the layout of tuple and union fields can be very different depending on these parameters in a way that can't be expressed in Rust.
These custom types can also be used when you call Rust from Julia with ccall.