Skip to content

What Woxi Can Do That Mathematica Can't

Woxi implements only a subset of the Wolfram Language (see missing features), but being open source, self-contained, and written in Rust lets it do several things that Mathematica, WolframScript, and the Wolfram Engine can't.

Licensing & cost

  • No license checks \ Woxi never activates, phones home, or validates a license. It works offline, in CI pipelines, in containers, and on air-gapped machines without any setup.

  • Free of charge \ Mathematica is a paid product. Woxi, Woxi Studio, the Playground, and JupyterLite are free, including for commercial use under the terms of the AGPL.

  • Unlimited parallel instances \ A Wolfram license caps the number of kernels that may run at once. Starting another wolframscript while the slots are taken fails with a misleading "not activated" error. You can start as many woxi processes as your machine can handle, e.g. to evaluate a batch of scripts in parallel:

    ls *.wls | xargs -P 8 -n 1 woxi run
    

  • Open source \ Every algorithm can be read, audited, debugged, and fixed. Results don't depend on a black box.

Embedding

  • Embed it in other software \ Calling the Wolfram Language from another program (via WSTP, LibraryLink, or the Wolfram Client Libraries) requires a separately installed and licensed Wolfram Engine running as its own process, and shipping it with your product requires a commercial license. Woxi runs in-process as a library and can be bundled with your application:

    • Rust: the woxi crate (woxi::interpret("Plus[1, 2]"))
    • Python: pip install woxi (see Python)
    • JavaScript / WebAssembly: the woxi-wasm npm package
  • Runs in the browser \ Compiled to WebAssembly, Woxi evaluates code entirely client-side in the Playground and in JupyterLite — no server, no account.

  • Shareable links \ The Playground encodes your code in the URL, so you can share a runnable snippet with a link. For example, this one evaluates Range[9]: https://woxi.ad-si.com/playground/?code=EoQwdg5gpg2gnAXSA

  • Custom builds \ Since the full source is available, you can compile your own version of Woxi containing only the functions you need, which results in a much smaller binary.

  • Instant startup \ woxi eval starts and finishes in a few milliseconds, while launching a Wolfram kernel takes most of a second or more. This makes Woxi practical for shell pipelines and for running many small scripts.

Language & file formats

  • Unicode characters in input \ For example to calculate the circumference of a circle with radius 4:

    woxi eval 'N[2π * 4]'
    

  • Runs .nb notebooks directly \ WolframScript can only execute .wls / .m script files, but Woxi can evaluate the Input/Code cells of a Mathematica .nb notebook right from the command line:

    woxi run notebook.nb
    

  • Imports CERN ROOT files \ The Wolfram Language has no importer for the standard container format of particle physics, but Woxi decodes its 1-D and 2-D histograms, strings, and nested directories into an Association, and reads TTree branch data — flat leaves, leaf-count arrays, std::vectors, and TLorentzVectors — through element paths (see Import):

    woxi eval 'Import["experiment.root"]'
    woxi eval 'Import["experiment.root", {"ROOT", "events", "energy"}]'