16 releases

0.2.2 Aug 4, 2024
0.2.1 Dec 22, 2023
0.1.4 Sep 13, 2023
0.1.1 Oct 7, 2022
0.0.1 Mar 27, 2021

#16 in Science

Download history 11/week @ 2024-07-25 132/week @ 2024-08-01 10/week @ 2024-08-08 3/week @ 2024-08-15 16/week @ 2024-09-19 18/week @ 2024-09-26 1/week @ 2024-10-03

994 downloads per month

BSD-3-Clause

1MB
5K SLoC

Unofficial TEOS-10 GSW Oceanographic Toolbox in Rust

GSW for microcontrollers and full computers.

Note that we do follow TEOS-10 manual and references, but this library is not endorsed by TEOS-10 committee.

The initial motivation for developing GSW-rs was to support pure Rust firmware for microcontrollers used in autonomous oceanography platforms -- such as underwater gliders and Argo floats. Rust stands out as an ideal solution due to its robustness and reliability complemented by an integrated testing system and a robust package manager to handle dependencies. Our Rust implementation achieves speeds comparable with C while capitalizing on Rust's explicit and intuitive syntax which makes it an optimal choice for scientific applications.

We have already implemented approximately 53 of the public functions described on TEOS-10, and we do intend to eventually cover all of them. While we're working on that, we'll keep an inventory list.

Talks

We presented about goals and progress (as of Apr 2022) at the SEA Improving Scientific Software 2022, slides available here.

Minimum supported Rust version

Currently the minimum supported Rust version is 1.68.0

Installing

This crate was developed as a library, thus it is intended to be 'imported' by other libraries or used by applications. The recommended way to install Rust is using rustup. Check the book Rust Programming Language for more details. To include GSW in another crate:

cargo add gsw

To be sure it works properly in your platform, clone this repository and run: cargo test which will compile the library and run all the tests.

Features

From the Cargo Book: "Cargo 'features' provide a mechanism to express conditional compilation and optional dependencies.". The features defined in GSW-rs are:

  • capi: Include the C-API so that GSW-rs can be accessed as it was a C-library. For instance, the other GSW implementations based on GSW-C could be linked with GSW-rs instead by using this feature.
  • compat: Reproduces the GSW-Matlab implementation for compatibility.
  • invalidasnan: Returns NaN values on failure. The default behavior is to return an error.
  • nodgdz: Ignores vertical variations of gravity, i.e. no dependency on z. This might be useful on some numerical models.
  • std: Activate the Rust standard library. The default implementation does not rely on std so it can run in embedded systems.

For example, to compile it compatible with the official Matlab library:

cargo build --features compat

Repository structure

For anyone learning Rust, this repository might be overwhelming. We are doing many things here, hence more files and directories that would be strictly necessary.

  • The Cargo.toml and src are the fundamental Rust components. The first one contains some metadata describing the crate, while the second one groups the source code. This is the core of the crate.
    • Note that in src we group the modules following the TEOS-10 library card. We also isolate the constants, while avoiding repetition.
    • The src/lib.rs is the starting point of the library.
    • We split the tests so that those are close to the target functions. Usually all the unit tests would go together somewhere, but here, each function has so many tests that it would be an extra work to debug and maintain. Also, we add reference cases described in the scientific literature.
  • Validation tests are grouped outside the src, in the tests, which uses the reference data.
  • convert_refdata is an auxiliary crate. We don't intend to publish that one since it is only used to support GSW-rs. The GSW-Matlab provides a reference dataset, which we can use to validate our library. Since our goal is to also work with microcontrollers, we have to format that dataset in such a way that it can be used in microcontrollers, thus validated. There you'll find its own Cargo.toml and src. We use postcard to encode it.
  • We use FFI to expose our library, so it can be used outside as if it was a C-library. For instance, to test it, we link the official GSW-Python with our Rust library (GSW-rs) instead of the traditional GSW-C, and run the Python tests. Take a look on .github/workflows/gsw-python.yml to see how we link GSW-Python with GSW-rs. The directories utils, include, assets, and examples/usage-from-c are all related to that.
  • Integration tests are developed in tests, using the specially encoded verification values (using convert_refdata) in tests/data to validate our results. Note that there is one file per module, which are small enough to fit in the stack memory of resource-limited systems.

License

Licensed under the 3-Clause BSD License (LICENSE)

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the LICENSE, shall be licensed as above, without any additional terms or conditions.

Contributions should be done through GitHub, by forking the repository, creating a new branch, and pushing that new branch back as a Pull Request. Tests covering the new feature or bugfix must be included, and if relevant, the documentation updated. If not familiar with the procedure, we encourage to contact us and we will walk you through the process. Every contribution is valuable and will be recognized.

A note on tests. We don't follow the typical Rust pattern of grouping all the tests of the module together. As a scientific library, and a large one, we tend to keep tests right after the target tested, so it is easy to verify if a certain function covers the desired behavior. Whenever possible, we also add tests confirming specific values described in the literature.

A note on references: Please review and add the relevant literature for each function. It is particularly important to verify the coefficients and the valid range in the original literature.

Citation

If you use this library we kindly ask to cite all the three following references:

DOI

Castelao, G., & Irber, L. (2024). Gibbs Sea Water Oceanographic Toolbox of TEOS-10 implemented in Rust. Journal of Open Source Software, 9(93), 5988, https://doi.org/10.21105/joss.05988

@article{Castelao2024,
  doi = {10.21105/joss.05988},
  url = {https://doi.org/10.21105/joss.05988},
  year = {2024},
  publisher = {The Open Journal},
  volume = {9},
  number = {93},
  pages = {5988},
  author = {Guilherme P. Castelao and Luiz Irber},
  title = {Gibbs Sea Water Oceanographic Toolbox of TEOS-10 implemented in Rust},
  journal = {Journal of Open Source Software}
  }

Why should you also cite the two following references? We do appreciate you recognizing our efforts developing this library citing the reference above, but we can't take any credit for the theory behind it. There is a long list of publications that resulted in all these relations explored in this library. Although we encourage you to study and cite the ones relevant to your work, you should at least cite the two following ones. The manual 56 that summarizes all those publications, and the 'Getting started' publication which presents how to use the Toolbox.

  • IOC, SCOR and IAPSO, 2010: The international thermodynamic equation of seawater - 2010: Calculation and use of thermodynamic properties. Intergovernmental - Oceanographic Commission, Manuals and Guides No. 56, UNESCO (English), 196 pp.

  • McDougall, T.J. and P.M. Barker, 2011: Getting started with TEOS-10 and the Gibbs Seawater (GSW) Oceanographic Toolbox, 28pp., SCOR/IAPSO WG127, ISBN 978-0-646-55621-5.

As we review and expand this library, we add the relevant specific references within each function. You can find those in our source code or the manual.

Dependencies

~0.6–1.6MB
~25K SLoC