WARNING: jl4R is moving to Rulia repository to design refactoring. There would be no more development of jl4R since the next step is now Rulia project.

Julia for R

This is an attempt to embed the julia language in R. Actually, very basic julia types are converted to R objects. This package is also very experimental

Install

1. Install Julia (all Operating System)

Install Julia. For Windows users don’t forget to select PATH in the installer.

2. Windows user setup

• Install RTools and a terminal with bash (for instance, Git Bash)
• Add Rscript in the PATH environment variable (see for example this page)

3. Bash installation (all Operating Systems)

In a terminal (tested on macOS M1 with julia-1.9.2:) with julia and Rscript binaries supposed to be in the PATH environment variable,

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/rcqls/jl4R/HEAD/inst/install.sh)"

How it works

getting started

require(jl4R)

## Le chargement a nécessité le package : jl4R

jl(`1.0`)

## 1.0

jl(1)

## 1.0

v_jl <- jl(c(1,3,2))
v_jl

## 3-element Vector{Float64}:
##  1.0
##  3.0
##  2.0

R(v_jl)

## [1] 1 3 2

jl() function

jl() as evaluation of julia expressions given as an one-length R character vector

jl(`[1,3,2]`)

## 3-element Vector{Int64}:
##  1
##  3
##  2

jl(`[1.0,3.0,2.0]`)

## 3-element Vector{Float64}:
##  1.0
##  3.0
##  2.0

jl(`(a=1,b=[1,3])`)

## (a = 1, b = [1, 3])

All these results are jlvalue objects which are R external pointers wrapping jl_value_t* values.

jl() as julia converter of R vectors

Above, an example for conversion of a vector of double was given. Below one completes with vector of character, logical and integer.

require(jl4R)
jl(c("one","three","two"))

## 3-element Vector{String}:
##  "one"
##  "three"
##  "two"

jl(c(TRUE,FALSE,TRUE))

## 3-element Vector{Bool}:
##  1
##  0
##  1

jl(c(1L,3L,2L))

## 3-element Vector{Int64}:
##  1
##  3
##  2

Notice that vector of length 1 are converted to atomic julia values.

require(jl4R)
jl(TRUE)

## true

jl(1L)

## 1

jl(1)

## 1.0

jl("1")

## "1"

To get a vector of length 1 in julia one has

require(jl4R)
jl("one", vector=TRUE) # or simply jl("one", TRUE)

## 1-element Vector{String}:
##  "one"

jl(TRUE, vector=TRUE) # or simply jl(TRUE, TRUE)

## 1-element Vector{Bool}:
##  1

jl(1L, TRUE)

## 1-element Vector{Int64}:
##  1

jl(1, TRUE)

## 1-element Vector{Float64}:
##  1.0

Notice that there is no need to set vector=TRUE when dim is not NULL:

jl(matrix("one"))

## 1×1 Matrix{String}:
##  "one"

Goal: conversion of julia structures used in statitictic to R

• DataFrame

require(jl4R)
jlusing(DataFrames)
jl(`(a=1,b=DataFrame(a=1:3,b=2:4))`) -> nt_jl
nt_jl
``$

    ## (\text{a} = 1, \text{b} = 3 \times 2 \text{DataFrame}
    ##  \text{Row} │ \text{a}      \text{b}
    ##      │ \text{Int64}  \text{Int64}
    ## ─────┼──────────────
    ##    1 │     1      2
    ##    2 │     2      3
    ##    3 │     3      4)

$`` r
R(nt_jl) # or toR(nt_jl)

## $a
## [1] 1
## 
## $b
##   a b
## 1 1 2
## 2 2 3
## 3 3 4

list(jltypeof(nt_jl), typeof(nt_jl), class(nt_jl))

## [[1]]
## @NamedTuple{a::Int64, b::DataFrame}
## 
## [[2]]
## [1] "externalptr"
## 
## [[3]]
## [1] "NamedTuple" "Struct"     "jlvalue"

• CategoricalArray

require(jl4R)
jlusing(CategoricalArrays)
ca_jl <- jl(`categorical(["titi","toto","titi"])`)
ca_jl

## 3-element CategoricalArray{String,1,UInt32}:
##  "titi"
##  "toto"
##  "titi"

R(ca_jl)

## [1] titi toto titi
## Levels: titi toto

list(jltypeof(ca_jl), typeof(ca_jl), class(ca_jl))

## [[1]]
## CategoricalVector{String, UInt32, String, CategoricalValue{String, UInt32}, Union{}} (alias for CategoricalArray{String, 1, UInt32, String, CategoricalValue{String, UInt32}, Union{}})
## 
## [[2]]
## [1] "externalptr"
## 
## [[3]]
## [1] "CategoricalArray" "Struct"           "jlvalue"

R Finalizers

Following the documentation on embedding julia, a system of preserved references to julia values has been created. An R finalizer is assiocated to each jlvalue object (in fact, an R external pointer wrapping some jl_value_t* value). Whenever the jlvalue is gabarged collected, the reference on the associated julia value is also dereferenced which is then cleaned up by the julia garbage collector.