[doawoo]

I want to come out and say that a long time ago at a startup we needed to generate a very particular type of analysis graph for a human operator to review in our SaaS.

and I just straight up installed GNU Octave on the server and called out to it from python, using the exact code the mathematician had devised.

[drnick1]

These days however with all the AI coding tools that are available, it probably makes more sense to just ask Claude to port the Matlab/Octave script to Python and directly integrate it into your program. Numpy/Scipy often provide drop-in replacements for Matlab functions, even the names are the same in some cases.

I have gone further and asked AI to port working but somewhat slow numerical scripts to C++ and it's completely effortless and very low risk when you have the original implementation as test.

[doawoo]

See the benefit of just using what the original mathematician wrote, is that if they had a problem with a way the graph was rendering, or they wanted to tweak it, they just had to edit the code, no translation layer needed. It shipped like any other component of the product at the time.

[hatmatrix]

But then you lose the readability that is the core defense of MATLAB/Octave. Ok so NumPy is readable, but less pleasantly so.

[finbarr1987]

Yeah, this is a pretty common pattern: use a domain-specific tool where it fits (Octave for the math), and a general language for the product glue (Python). Same idea as infra work — lots of teams would rather express intent in Terraform than build it in Rust, because a DSL can be a cleaner fit for the job.

[grandiego]

For my thesis I did something similar: bash scripts to extract raw data from a Subversion repository, to be preprocessed with PHP scripts (now I would prefer Python but had more experience with PHP) for text extraction and csv output, and finally Octave did the math magic, generating tables and saving graphics in png format, ready for import into my Lyx document.

[fithisux]

The most sensible thing I've heard this year.