Create R wrapper/interface for GEMMA about gemma

[timflutre]

+1 that would be great! maybe the first step to do is to refactor the core of the code into a stand-alone library, which would then be easy to call from R via Rcpp

[pjotrp]

I don't think GEMMA should be tied to R. We are using it from Python. You can still create an R wrapper that passes data in memory, but don't do away with the standard CLI. I would create a separate R package for this that may share the source code.

[timflutre]

Yes, I agree with @pjotrp, that's why I advocated for a library. In this way, the same library can be called by a simple .cpp file with a main function to be used from the CLI, but it can also be easily used by an R package via Rcpp, or from Python, or Julia, etc.

[pcarbo]

I agree @timflutre & @pjotrp.

[fboehm]

@timflutre From what I understand about libraries, I like the idea of creating a library, but I am new to C++. Can you say more about what we would need to do to create a library? How would we begin to refactor the core of the GEMMA code? Thanks!

[pcarbo]

@fboehm @timflutre It would involve replacing GEMMA's command-line interface, in which the data is read from files, with a single function that takes as input (1) the options, and (2) the data in the form of matrices (or perhaps data frames). The input could be passed from R to C++ via the .Call interface in R. However, I should warn you that designing this single function for interfacing the C++ code to R will be a substantial amount of work.

[timflutre]

@fboehm before refactoring anything, the first step is to automatize tests, as suggested in issue #38

[pjotrp]

We have tests. Closing this issue for now due to inactivity. Reopen if anyone wants to work on it.

[pjotrp]

Gemma2/lib is providing Python abstractions. An R wrapper may be part of that project.