How about hosting C/C++/Fortran/Rust code in Julia packages?

mkitti · August 21, 2023, 12:30am

I have a draft recipe here:

Migrating Fortran part of an R package (mrfDepth) into Julia

opened 06:11PM - 18 Aug 23 UTC

closed 06:27PM - 22 Aug 23 UTC

Hey guys, I hope this issue helps with the migration of some legacy code to J…ulia and also paves the way for the addition of similar packages in the future. The issue is about migrating the Fortran part of an R package, namely mrfDepth, which is hosted as an R package in [mrfDepth](https://github.com/cran/mrfDepth). The main interest is the `rdepthmedian` function which is used for estimating the Deepest Regression parameters, a well known robust regression estimator. It is simply used like ```R R> rdepthmedian(maxit = 10000, x = stackloss) $deepest intercept slope var. 1 slope var. 2 slope var. 3 -35.37610619 0.82522124 0.44247788 -0.07964602 ``` where the first parameter is intercept and the remaining part represents the partial slops. When I compile all of the fortran code in a single binary using ```shell $ gfortran -shared -fPIC *.f ``` I can bring all of the functionality into Julia using the FFI features. The estimated parameters can be re-calculated in Julia using ```Julia X = convert(Matrix, stackloss) n, p = size(X) n = Int32(n) p = Int32(p) A = zeros(Float64, p) maxit = Int32(10000) iter = Int32(1) MDEPAPPR = Int32(3) result = ccall((:sweepmedres_, "./a.out"), Cint, (Ref{Float64}, # X Ref{Int32}, # n Ref{Int32}, # np Ref{Float64}, # A Ref{Cint}, # maxit Ref{Cint}, # iter Ref{Cint} # MDEPAPPR ), X, n, p, A, maxit, iter, MDEPAPPR) println(A) ``` and the codes returns the same estimates as we obtained using R: ```julia [0.8252212389746946, 0.44247787604338223, -0.0796460177080886, -35.37610619462] ``` where the last term is intercept and first 3 estimates are partial slopes. The problem is, the package is an R package and I failed to make it done using BinaryBuilder.jl. - Can any talented friends in this area help migrating this functionality into Julia? - Can we standardize the toolchain that inputs an R package and prepare a jdll output in an easy way? Sorry if I am bothering you for a specific issue. Thank you for considering my message.

Topic		Replies	Views
Considering switch to Julia -- modelling examples linked to R / comparison to Stan / production environment New to Julia	1	538	June 5, 2021
[ANN] Embedding Rust Library in a Julia Package General Usage announcement , rust	3	3357	September 27, 2018
Fortran modules and Julia modules General Usage question , fortran	8	1626	August 18, 2020
Developing an R package using Julia for a statistical model Statistics	21	810	December 23, 2024
Julia version of the (fortran) IMSL library? New to Julia statistics	2	900	August 20, 2019

How about hosting C/C++/Fortran/Rust code in Julia packages?

Related topics