NASTRAN-95 VS squarify

Well, NASTRAN was one of the first Solvers developed and available. You can download predecessor if current NASTRAN for free from github: https://github.com/nasa/NASTRAN-95

I haven’t tried to use the public, open source version, but shouldn’t NASTRAN make this list?

A larger legacy FORTRAN project is NASTRAN-95, a finite element structural analysis code (again from NASA) that consists of almost 2000 individual source files as well as substantial example case input and output files.

Finally, consider how long it took to answer these questions for each file. The Castlequest source code is distributed among 16 files; how long would it take to assess the whole codebase by viewing bitmaps versus skimming source files in an IDE or text editor? By comparison, the NASTRAN-95 project consists of almost 1900 separate source files distributed among 10 directories - would the graphical or the textual approach be more effective on a project of that size? Would either approach be effective at that scale?

I think there's a common misconception that numerical codes should be accessible without documentation, and without a background in the subject.

Take NASTRAN-95 [1] or SPICE2 [2], for example. Both have good documentation in the form of manuals, books, and papers. If you have the mechanical or electrical background, you should be able to understand the docs explaining the implementation, and then you should be able to understand the code. It doesn't matter that there are no comments anywhere, or that GOTO is used. Aside from some cosmetic differences, it looks pretty much like what you'd write today with MATLAB or Python.

I picked these examples because they're publicly available, but I would guess that most Fortran still in use has no public visibility. In my experience with commercial numerical engineering software, Fortran lives on in under-the-hood components written by subject matter experts for performing specific tasks. It doesn't matter that programmers are unfamiliar with the language, because only subject matter experts are allowed to modify the code anyway.

To be clear, I'm not defending the many examples of unstructured, undocumented academic code that grows until it's essential to an organization despite being buggy and virtually unmaintainable. But those would have been terrible to read no matter which language was used.

I've never seen the language itself be a significant barrier to understanding in a business context. And it's not like we have a problem where there are all these active open-source projects that could be so much better if only they were written in a different language.

[1] https://github.com/nasa/NASTRAN-95

[2] https://ptolemy.berkeley.edu/projects/embedded/pubs/download...

Nastran bulk data: https://github.com/nasa/NASTRAN-95/blob/master/um/BULK.TXT Similar case, but modern nastran versions have somewhere in the 1000s (I think) card types.

NASTRAN-95/um at master · nasa/NASTRAN-95 (github.com)

The code to produce the treemap layout is mostly using Erik Cederstrand's squarify treemap layout algorithm, and can be seen at https://github.com/dbabbitt/StatsByCountry/blob/master/treemaps/Prison%20Population%20by%20US%20State%20Colored%20by%20White%20Percent.ipynb.

Conveniently, treemap generation is not terribly complex; the original papers describing the algorithms for squarifying and cushioning treemaps are readily available. Further, treemaps have been used to visualize a wide variety of hierarchical data beyond filesystem contents. There are libraries in a number of languages which simplify treemap creation, for example the Python library squarify generates treemap images using the Matplotlib plotting library.

Tools used: Python and Squarify.