This tutorial demonstrates how to implement common Matlab patterns in Rust using the ndarray ecosystem of linear algebra tools. It begins with simple matrix creation and progresses through basic statistics and parallel matrix operations.
Clone this repository with:
git clone https://github.com/benkay86/matlab-ndarray-tutorial.git
Compile and run the build_test example to verify that everything works.
cd matlab-ndarray-tutorial
cargo run --bin build_test
... cargo/rustc output ...
Running `target/debug/build_test`
Testing BLAS backend... OK.
Testing LAPACK backend... OK.
You will need to install Rust, C, and fortran compilers (typically gcc and gfortran). You will also need a blas/lapack backend such as openblas. These dependencies are provided on most modern operating systems. If you have difficulty getting openblas to work, open Cargo.toml and edit the last two lines to change system to static (this will significantly increase compilation time).
This tutorial is divided into workspaces by language. We strongly recommend beginning with the rust workspace.
- rust/
- The best place to get started.
- Review pros and cons of ndarray vs Matlab.
- Examples of ndarray using pure Rust.
- Work through basic examples in order.
- Refer to advanced examples as needed.
- python/
- Pass data seamlessly between Python's numpy and ndarray.
- Write numerical Python extensions in Rust.
- matlab/
- Need examples of how to write mex extensions in Rust.
- Your help writing this tutorial is welcome!
If you discover an error in this tutorial, have a question about an example, or would like to suggest an improvement, please open an issue on Github or e-mail the primary author benjamin@benkay.net.
If you have recently pulled this tutorial from ndarray version 0.14 to 0.15, you may get get bizarre trait errors like this one:
the trait FromPyObject<'_> is not implemented for PyReadonlyArray<'_, f64, Dim<IxDynImpl>>
If some of your code uses traits from ndarray 0.15 and other parts use the same trait from ndarray 0.14, Rust will see these as different, incompatible traits. To fix this, update your Cargo.lock file to make sure all packages depend on the same version of ndarray. Simply go to the root directory of this respository and run:
cargo update