Deprecated

• mlr3learners.drat is now deprecated. The majority of learners in this organisation now live in mlr3extralearners. # mlr3learners.drat
• mlr3learners.drat is no longer being maintained and will stop functioning on 1st January 2021.
• Only mlr3learners.lightgbm is still being maintained in this organisation, for stable releases of all other learners please see mlr3extralearners.

mlr3learners.drat

drat package repository for the mlr3learners.drat project.
Install mlr3 learners from this repository using

install.packages(..., repos = "https://mlr3learners.github.io/mlr3learners.drat")

To omit having to pass the repos argument every single time, add the following to your .Rprofile:

options(repos = structure(c(getOption("repos"),
  mlr3learners = "https://mlr3learners.github.io/mlr3learners.drat")))

This will append this drat repo to your list of available repositories which install.packages() will check when trying to install a package. If you want a c/p command for this: The following will append the line above to your ~/.Rprofile while keeping your currently set CRAN mirror:

cran = getOption('repos')[['CRAN']] 
write(sprintf("options(repos = structure(c(CRAN = '%s',
  mlr3learners = 'https://mlr3learners.github.io/mlr3learners.drat')))",
   cran),
  file = "~/.Rprofile", append = TRUE)

Note: Restart your R session to make use of the new option.

With this, you are able to install any learner from the mlr3learners organization as simple as

install.packages("mlr3learners.kernlab")

Available packages

List of all available learners in this organization

##  [1] "mlr3learners.c50"             "mlr3learners.catboost"       
##  [3] "mlr3learners.coxboost"        "mlr3learners.extratrees"     
##  [5] "mlr3learners.flexsurv"        "mlr3learners.fnn"            
##  [7] "mlr3learners.gbm"             "mlr3learners.gss"            
##  [9] "mlr3learners.kerdiest"        "mlr3learners.kernlab"        
## [11] "mlr3learners.ks"              "mlr3learners.liblinear"      
## [13] "mlr3learners.locfit"          "mlr3learners.logspline"      
## [15] "mlr3learners.mboost"          "mlr3learners.nnet"           
## [17] "mlr3learners.np"              "mlr3learners.obliquersf"     
## [19] "mlr3learners.partykit"        "mlr3learners.penalized"      
## [21] "mlr3learners.pendensity"      "mlr3learners.plugdensity"    
## [23] "mlr3learners.proba"           "mlr3learners.pycox"          
## [25] "mlr3learners.randomforest"    "mlr3learners.randomforestsrc"
## [27] "mlr3learners.rweka"           "mlr3learners.sm"             
## [29] "mlr3learners.survival"        "mlr3learners.survivalsvm"

How it works

The package sources and binaries of the listed packages are build by the respective learner packages with the following logic:

• During a CI run when the version does not contain a devel indicator (i.e. .9000)
• On the first of each month

Binaries are build for the current R release and the previous release. Depending on the release date of a new major or minor R version it may take a while until the new binaries are available. The binaries are (re-)build on the first of each month.

The work behind the scenes is highly simplified by packages {tic} and {drat}.

How to add a mlr3learner to {mlr3learners.drat}

1. Check if you belong to the mlr3learners.drat team of the mlr3learners organization. If not, ask @pat-s to add you - otherwise, you do not have permission to add deploy keys to this repository.

2. Make sure you have a GITHUB_PAT set for the GitHub API calls. If you do not know what this is, read ?usethis::browse_github_token().

3. Make sure your repo already uses GitHub Actions. If not, call tic::use_ghactions_yml(deploy = TRUE).

4. Source init_mlr3learner_drat_deploy() from this Gist and call the function with the name of the repo, e.g. init_mlr3learner_drat_deploy("mlr3learners.mboost"). This will add a private key as a “secret” to your repo and a public deploy key to mlr3learners/mlr3learners.drat, making it possible to deploy to the latter from your repo.

5. Now the macOS and Windows runners will push their binaries to mlr3learners.drat for every no-devel version (i.e. versions without .9000 at the end). The Linux runner will push the .tar.gz file to install the learner from source.

Note

It is very important to only init one CI run per release and then switch directly to a dev version. Otherwise, binaries will be pushed during every CI run, bloating up the repo size of mlr3learners/mlr3learners.drat](https://github.com/mlr3learners/mlr3learners.drat).