New Swig/Boost.Python alternative?

I just saw this on the Python C+±SIG email list. Haven’t had a chance to look at the code, but the description sounds very promising:


1 Like

Should we plan on setting up a small-scale prototype to see if pybind11 is a nicer glue than SWIG?

cc/ @timj in the spirit of your lsstx experiment.

I skimmed some documentation, and agree it sounds promising. I think @natelust was planning to try it out in one of his “20% time” projects, so he might have some comments on how well it works in practice in a few weeks.

An additional concern would be about the long term maintenance and stability of the project: so far, it seems to be a one person effort developed over the course of the summer. In other words, the bus factor is pretty low. Before investing too much time on it, we’d need to consider if whether we could count on it in the long term (e.g. by allocating our own resources to keep it alive if the original author abandons it).