A few of my readers have noticed my absence on this blog and the seemingly slow development of AequilibraE, so I decided to share what is going on and what the plans are for the next 6 months.
The current version of AequilibraE (0.3.5.x) is the most powerful and robust version of AequilibraE yet, with emphasis on robust. This is the first version that had consistent answers to bug reports and followed with a few minor releases correction each one of the reported bugs.
To develop new features, however, I realized that they would require more of basic resources. Namely, things such as multi-class assignment (which also applies to Delaunay Lines) requires a proper matrix format, which has never been available in AequilibraE (all computation relied on matrices being imported “on-the-fly”).
Developing a matrix format that is both efficient and robust was no easy task (there is a reason why traditional software keeps its matrix format/routines so close to their chests), but I think I reached a pretty good compromise using some rather obscure features on NumPY originally developed to support MatPlotLib features. I will write more on that in the future.
Developing the matrix format opened a huge can of worms, leading me to develop a data format for keeping the binary data used in all procedures (such as zone-based data, assignment outputs, etc.) and re-factoring all procedures on trip distribution. It has been a daunting task.
On the bright side, I have decided to do the re-factoring while writing unit tests for all the code I touch, which means that the software should be much more robust going forward. It is a lot more work than just doing the re-factoring, but I am really enjoying developing “proper software”.