numericalalgorithmsgroup
Pandas dependency
Great work on this! This seems to work pretty well for solving mode matching problems in optics, which is a very much non-convex (hence the name?) problem.
I have one gripe though: as far as I can tell, pandas is only used for diagnostics. Could this maybe be removed as an official dependency and instead be made available either as a setup extra or have Py-BOBYQA throw an error if pandas is not found when the diagnostics are run? I'm considering using the package with my own library but having pandas as a dependency when it's hardly used seems like overkill.
Thanks - this is a good idea. Without pandas, the diagnostics could also be made available as a dictionary.
I've made this change. For my own reference, I still need to test and update formal dependencies and documentation.