Consider Rfast!

[JonasMoss]

Rfast is a package with many univariate densities implemented. Most if not all of the implementations have a higher speed than the implementations in this package and the overlap is large.

Do one of these:

1. Import Rfast and make use of their implementations;
2. Make univariateML GPL and adopt the code from Rfast.

[tripartio]

I like the idea of using Rfast. However, I recommend the first option: import Rfast and use its implementations rather than adopting its code directly. Two problems I see with the second option:

• The GPL is more restrictive than the MIT license, so that would be a step backwards in flexibility. (I am actually in the process of moving my own package ale from GPL to MIT, which involves reimplementing code that I had copied from a GPL project. I think its worthwhile moving in that direction. Of course, I'm biased because my package imports univariateML.)
• Rfast is 71% compiled in C++ whereas univariateML is 100% R. Moving to a compiled package has significant implications (users might have to compile from source sometimes), a step that I think would be better to avoid unless absolutely necessary.

[JonasMoss]

In the working branch https://github.com/JonasMoss/univariateML/tree/attributes-decorator I've used Rfast to construct tables when estimating the binomial (their function is much faster than base R). Now we need to find the Rfast candidates for ML and check them; it's important to test them, not only for speed, but for correctness.