jluna: A modern Julia Wrapper for C++ (v1.0.0)

Julia is a beautiful language, it is well-designed, and well-documented. Julia's C-API is also well-designed, less beautiful, and much less... documented.
jluna aims to fully wrap the official Julia C-API, replacing it in projects with C++ as the host language, by making accessing Julia's unique strengths through C++ safe, hassle-free, and just as beautiful.

Table of Contents

1. Introduction
2. Features
   
3. Showcase
   
4. Documentation
   
5. Dependencies
   4.1 Julia 1.7.0+
   4.2 Supported Compilers: g++, clang++, MSVC
   4.3 CMake 3.12+
   
6. Installation
   
7. License
8. Authors

Anouncements

Note: jluna is currently not available for Apple aarch64 architectures, such as those used by the M1 or M2 MacBooks. See here for more information. jluna should still work for Windows 10, 11, Linux, and FreeBSD.

Features

• expressive, generic syntax
• create / call / assign Julia-side variables from C++
• full exception forwarding, verbose error messages with complete stacktrace
• std types & usertypes can be moved freely between Julia and C++
• call arbitrary C++ functions from Julia
• multidimensional, iterable array interface
• provides a custom thread pool that, unlike the C-API, allows for concurrent interfacing with Julia
• provides < 5% overhead functions, viable in performance-critical environments
• complete manual, installation guide, benchmark analysis, inline documentation for IDEs - all written by a human
• and more!

Showcase

(If you are looking for examples showing best-practice basic usage, please instead consult the manual)

Executing Julia Code

```cpp // execute multi-line Julia code Main.safe_eval(R"( f(x) = x^x^x vec = Int64[1, 2, 3, 4] )");

// call Julia functions with C++ values auto f = Main["f"]; std::cout << (Int64) f(3) << std::endl;

// mutate Julia-side values Main["vec"][2] = 999; Main.safe_eval("println(vec)");

// assign std objects to Julia variables Main["vec"] = std::vector{117, 118, 119, 120}; Main.safe_eval("println(vec)"); 2030534587 [1, 2, 999, 4] ['u', 'v', 'w', 'x'] ```

Array Interface

```cpp // array interface Array matrix = Main.safe_eval("return reshape([i for i in 1:(4*4)], 4, 4)");

// supports multi-dimensional indexing (and array comprehension, not shown here) matrix.at(0, 2) = 999; Main"println";

// even has generator expressions! auto generatedvector = Vector("(Char(i) for i in 97:104)"gen); Main"println"; [1 5 9 13; 2 6 10 14; 999 7 11 15; 4 8 12 16] ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h'] ```

Calling C++ Functions from Julia

```cpp Main.safeeval("cppfunction = () -> ()"); // forward declaration

// assign C++ lambda to Julia Function Main["cppfunction"] = asjulia_function( -> void { std::cout << "cpp prints: " << in << std::endl; } );

// call lambda, entirely Julia-side Main.safeeval(R"( cppfunction("whatjuliahandsit") )"); cpp prints: whatjuliahandsit

```

Documentation

Documentation, including a step-by-step installation and troubleshooting guide, tutorial, and index of all functions and objects in jluna is available here.

Dependencies

jluna aims to be as modern as is practical. It uses C++20 features extensively and aims to support the newest Julia version, rather than focusing on backwards compatibility.

For jluna you'll need: + Julia 1.7.0 (or newer) + cmake 3.12 (or newer) + C++ Compiler, one of - g++10 (or newer)
- clang++-12 (or newer)
- MSVC 19.34 (or newer)

On Unix, g++ or clang (installed using your package manager) are recommended.
On Windows, either use g++ provided by MinGW or MSVC provided by the Visual Studio C++ build tools.

In either case, make sure the compilers' version is as stated above, as jluna uses modern C++20 features extensively.

Installation & Troubleshooting

A step-by-step guide is available here. It is recommended that you follow this guide, instead of the highly abridged version below.

For IDEs: In many cases, simply opening the cloned jluna project in an IDE (such as VisualStudio, Atom, or CLion) will allow it to automatically set everything up for you. After initialization, simply run "install" from your build menu.

Command Line

Execute, in your bash console, in any public directory:

bash git clone https://github.com/Clemapfel/jluna cd jluna mkdir build cd build cmake .. -DJULIA_BINDIR=$(julia -e "println(Sys.BINDIR)") -DCMAKE_CXX_COMPILER=<C++ Compiler> -DCMAKE_INSTALL_PREFIX=<install directory> Where + <C++ Compiler> is the C++ compiler executable, e.g. g++, clang++, cl.exe, etc. + <install directory> is the desired install directory, omit this option to use the systems default directory

Then: make install ctest --verbose

Which will deposit the library to the specified system folder and run tests to make sure everything works.

Example Usage

For example, installing on a linux machine using g++:

git clone https://github.com/Clemapfel/jluna cd jluna mkdir build cd build cmake .. -DJULIA_BINDIR=$(julia -e "println(Sys.BINDIR)") -DCMAKE_CXX_COMPILER=/usr/bin/g++ sudo make install ctest --verbose Where ommitting DCMAKE_INSTALL_PATH makes CMake choose the default system path. sudo was necessary to write to that path.

Afterward, you can make jluna available to your library using

```cmake

inside your own CMakeLists.txt

findlibrary(jluna REQUIRED NAMES jluna PATHS ) targetlinklibraries( PRIVATE "${jluna}" "${}") `` Where +is the directory specified via-DCMAKEINSTALLPREFIX +is the Julia shared library (usually available in"${JULIABINDIR}/../lib") +` is the name of your library or executable

If any step of this does not work for you, please follow the installation guide instead.

Credits

jluna was designed and written by Clem Cords.

March 2022:

• CMake improvements by friendlyanon

Donations

Jluna was created with no expectation of compensation and made available for free. Consider donating to reward past work and support the continued development of this library:

• GitHub Sponsors
• PayPal

License & Citation

The current and all prior releases of jluna are supplied under MIT license, available here.

If you would like to cite jluna in your academic publication, you can copy the entry in CITATION.bib to your BibTeX bibliography, then use the \cite{jluna} command anywhere in your LaTeX source code.

Thank you for your consideration, C.