fortran-lang/minpack

Provide R bindings

ivan-pi opened this issue · 2 comments

There is a MINPACK wrapper for R called minpack.lm.

It might be interesting to check how many times it has been downloaded and how many R packages depend on it.

Having a C interface also makes the task of writing an R wrapper easier. Perhaps we can get in touch with the authors of that package at some point and ask them to use the new maintained version instead.

Just noticed the thread from @Beliavsky at Discourse, https://fortran-lang.discourse.group/t/gsoc-improve-nonlinear-least-squares-in-r/1735, on the topic of a GSoC'21 project about "Improving Nonlinear Least Squares in R".

The homepage of the GSoC project is here: https://arkab-ds.github.io/GSOC21-improveNLS/

One of their perceived problems is maintainability of the C/Fortran code used to wrap NLS (https://www.netlib.org/port/). NLS is a least squares code developed around the same time as MINPACK. I just found this to be a compelling example of an R project trying to move away from Fortran for various reasons.

Maybe we could coordinate with the R community on a joint GSoC project to "renew" the R MINPACK interface.

@ivan-pi do you want to take a lead on this? I think it would be great if we can tap into the R community and collaborate.