mlrng: Machine Learning in R - Next Generation

#' @import stringi
#' @import checkmate
#' @import data.table
#' @importFrom BBmisc vlapply viapply vcapply vnapply seq_row seq_col isFALSE
#' @importFrom digest digest
#' @importFrom fastmatch fmatch %fin%
#' @importFrom glue glue double_quote
#' @importFrom parallelMap parallelMap parallelExport parallelLibrary
#' @importFrom prettyunits pretty_sec
#' @importFrom R6 R6Class
#' @importFrom stats setNames predict
#' @importFrom utils data
#' @import phng
#'
#' @description
#' For bug reports and feature requests please use the tracker:
#' <https://github.com/mlr-org/mlrng>.
#' Package options are convered in \link{mlrng-config}.
#'
"_PACKAGE"

mlrng = new.env(parent = emptyenv())
mlrng$supported.col.types = c(
  "logical", "integer", "numeric", "character", "factor", "ordered"
)
mlrng$supported.learner.props = c(
  sprintf("feat.%s", mlrng$supported.col.types),
  "missings", "weights", "twoclass", "multiclass", "prob", "featimp", "parallel",
  "formula", "oobpreds", "se"
)
mlrng$result.states = c(
  "Result", "TrainResult", "PredictResult", "PerformanceResult", "ResampleResult", "BenchmarkResult"
)
mlrng$default.opts = list(
  mlrng.verbose = TRUE,
  mlrng.debug = TRUE,
  mlrng.keep.train.output = FALSE,
  mlrng.continue.on.learner.error = FALSE,
  mlrng.train.encapsulation = "evaluate"
)

.onLoad = function(libname, pkgname) { #nocov start
  backports::import(pkgname)
  backports::import(pkgname, "hasName", force = TRUE)
  parallelMap::parallelRegisterLevels(package = "mlrng", levels = c("resample", "benchmark", "tune"))

  # set default options
  opts = options()
  unset = which(!(names(mlrng$default.opts) %in% names(opts)))
  if (length(unset))
    options(mlrng$default.opts[unset])

  # read config file
  options(readMlrngConfig())

} #nocov end

.onUnload = function(libpath) { #nocov start
  pm.opts = getOption("parallelMap.registered.levels")
  pm.opts$mlrng = NULL
  options(parallelMap.registered.levels = pm.opts)
} #nocov end