Add a Julia Interface for Cantera (Cantera.jl)

[Fastaxx]

Changes proposed in this pull request

• Add a new Julia interface (Cantera.jl) under interfaces/julia, linking directly against the native libcantera through generated CLib bindings (no Python dependency).
• Extend sourcegen with CLib recipes covering kinetics derivatives, _ddX getters, ReactorNet state, standard-state species properties, and heat production/release rates, and use them to generate the low-level Julia bindings (src/generated/*.jl).
• Wrap the generated bindings in a public API (Solution, ReactorNet, Reactor, Domain/Sim1D, Multiphase, Func1, connectors, reaction-diagnostics, etc.), including thermodynamic, kinetic, transport, mixture, and 1-D flame functionality.
• Add a test suite (test/runtests.jl and per-module tests) with reference values, examples (examples/*.jl), and API documentation (docs/, README.md).
• Add a CI workflow that builds libcantera from source and runs the Julia test suite.

If applicable, fill in the issue number this pull request is fixing

Cantera/enhancements#81

If applicable, provide an example illustrating new features this pull request is introducing

using Cantera
gas = Solution("gri30.yaml")
set_TPX!(gas, 1000.0, one_atm, "H2:2, O2:1, N2:4")
net = ReactorNet(IdealGasReactor(gas))
advance!(net, 1e-3)
temperature(gas)          # ignited temperature [K]

AI Statement (required)

Limited use of generative AI.
Standard or boilerplate code snippets were generated with AI and manually reviewed;
all design, logic, and implementation decisions were made by the contributor.
Examples: IDE code-completions or brief LLM queries for common patterns.

For additional information on Cantera's AI policy, see
https://github.com/Cantera/cantera/blob/main/CONTRIBUTING.md#restrictions-on-generative-ai-usage

Checklist

• The pull request includes a clear description of this code change
• Commit messages have short titles and reference relevant issues
• Build passes (scons build & scons test) and unit tests address code coverage
• Style & formatting of contributed code follows contributing guidelines
• AI Statement is included
• The pull request is ready for review

[speth]

Thanks for contributing this, @Fastaxx. Overall, I think this looks really interesting, and I'm looking forward to seeing where this goes. A few general comments in addition to my in-line comments:

• I'm a little torn on the low-level binding generator -- rather than having generate_bindings work off the generated CLib headers, I feel like it would be a little cleaner to have at least this layer of the interface directly generated by sourcegen, with Julia as a new submodule there alongside clib and dotnet. On the other hand, this works, so we can go with it for now and if it breaks later, maybe that will be the impetus for a rewrite.

• While this is a pretty comprehensive start, it would be useful to know what future plans you have for this, if any. For example, what the path is to getting documentation that we can integrate with the other API docs, and the path toward packaging this without users having to build Cantera from source (how does Julia handle packages with dependencies on compiled libraries?) If you want to post about this in the Cantera/enhancements#81 thread, that would be very helpful.

• All the new source files (excluding the generated ones and the examples) should contain the license/copyright preamble:

# This file is part of Cantera. See License.txt in the top-level directory or
# at https://cantera.org/license.txt for license and copyright information.

• I think you may be understating the degree to which this was implemented by an LLM. There's no inherent problem with contributing AI-generated code, but it does help the review process to understand where this came from. Can you please update the "AI Statement" to provide a little more detail about how you created this?

• Please add yourself to AUTHORS.md

[speth]

I guess while this is very much a pre-release feature, we can leave this as an 0.x version, it should be updated to match the Cantera version once this matures.

[Fastaxx]

Hello @speth,
Thank you very much for your detailed and constructive feedback. I started using Julia three years ago for my PhD thesis. This Cantera connector allows me to access and integrate data from the Cantera project into my research.

I have incorporated a large number of changes following your comments.

However, I have some questions regarding best practice concerning the following points.

• bindings : I used the low-level approach with generate-bindings for my experiment. However, it seems to me that we could use Yggdrasil to generate and distribute the low-level bindings via a pre-compiled Cantera_jll package, as i try in another project.

• Documentation within the project. Is it better to use the Doxygen template from the repository, or to host the documentation via the Julia Documenter package with deployment to gh-pages?

• Regarding the AI statement, i would say a Medium Use of AI, particularly the generation of bindings, with which I am not yet familiar. Also the automatic generation of documentation. I did, however, want to structure the interface myself, as well as write the tests and carry out functional verification.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.15%. Comparing base (39cfc14) to head (af3cff4).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2149   +/-   ##
=======================================
  Coverage   78.15%   78.15%           
=======================================
  Files         453      453           
  Lines       55433    55433           
  Branches     9117     9117           
=======================================
  Hits        43324    43324           
  Misses       9060     9060           
  Partials     3049     3049           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[speth]

• Documentation within the project. Is it better to use the Doxygen template from the repository, or to host the documentation via the Julia Documenter package with deployment to gh-pages?

I think the docs should be integrated and hosted with the rest of the Cantera documentation. There are two approaches to doing this. One, which we use with the Matlab toolbox, is to have Sphinx be the documentation generator for the Julia code. That depends on there being a (good) Sphinx plugin for Julia, and I don't know if that exists or not. The other, which is what we do with C++, is to let the language-specific documentation generator (Doxygen for C++ or Documenter for Julia) generate its own pages, doing what we can to configure the output for a similar style, and then merge the resulting outputs. In the case of Doxygen, we get some interlinking capability due to the doxysphinx extension, but you can also just use relative HTML links if the paths are predictable. Again, I'm not sure what's available here.

Take a look at doc/SConscript to see how the doc build process is driven. If we have scons build the Julia docs along with the Sphinx and Doxygen ones as part of the docs workflow in main.yml, then updated ones will automatically be deployed and accessible at https://cantera.org/dev/reference after each PR is merged.

[Fastaxx]

Thank you for the suggestions. I’ve taken your comments on board. I’ve also added the documentation. Is that all right with you, overall?

[Fastaxx]

I pushed two new commits.

• sourcegen for the low-level bindings. Went with your first instinct : generate_bindings.jl is gone, replaced by a julia submodule under interfaces/sourcegen/src/sourcegen/ alongside clib and csharp, scaffolding from the same ct*.yaml specs. An interfaces/julia/SConscript runs it during the build, so scons build emits src/generated/ and using Cantera works with no generator step. One less step in CI too. The regex parser for the CLib headers is gone with it, and the output is unchanged: all 455 wrappers across the 14 headers are identical to what the script emitted, tests pass.

• About the handles.jl question. Yes, the views now hold a reference to the owning Solution. Added a test; each view is built from its own Solution, since views sharing an owner keep each other alive and would mask a regression in any
  single accessor.