mlr3pipelines

Releases

32

Needs compilation

No

Needs R

>= 3.3.0

Config/roxygen2/version

8.1.0

Config/testthat/edition

3

Config/testthat/parallel

true

• mlr3verse, installed by the mlr3verse package

• Machine learning frameworks, core

tests, Error on 1 platform

Running 'testthat.R' [143s]
Running the tests in 'tests/testthat.R' failed.
Complete output:
  > if (requireNamespace("testthat", quietly = TRUE)) {
  +   library("checkmate")
  +   library("testthat")
  +   library("mlr3")
  +   library("paradox")
  +   library("mlr3pipelines")
  +   test_check("mlr3pipelines")
  + }
  Starting 2 test processes.
  > test_Graph.R: Training debug.multi with input list(input_1 = 1, input_2 = 1)
  > test_PipeOp.R: Training test_autotrain
  > test_PipeOp.R: Predicting test_autotrain
  > test_PipeOp.R: Training test_autotrain
  > test_PipeOp.R: Training test_autotrain
  > test_PipeOp.R: Predicting test_autotrain
  > test_PipeOp.R: Predicting test_autotrain
  > test_PipeOp.R: Training test_autotrain
  > test_PipeOp.R: Training test_autotrain
  > test_PipeOp.R: Training test_autotrain
  > test_PipeOp.R: Predicting test_autotrain
  > test_PipeOp.R: Predicting test_autotrain
  > test_PipeOp.R: Predicting test_autotrain
  > test_PipeOp.R: Training test_autotrain
...[truncated]...
073741819
  Backtrace:
       ▆
    1. └─testthat::test_check("mlr3pipelines")
    2.   └─testthat::test_dir(...)
    3.     └─testthat:::test_files(...)
    4.       └─testthat:::test_files_parallel(...)
    5.         ├─withr::with_dir(...)
    6.         │ └─base::force(code)
    7.         ├─testthat::with_reporter(...)
    8.         │ └─base::tryCatch(...)
    9.         │   └─base (local) tryCatchList(expr, classes, parentenv, handlers)
   10.         │     └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
   11.         │       └─base (local) doTryCatch(return(expr), name, parentenv, handler)
   12.         └─testthat:::parallel_event_loop_chunky(queue, reporters, ".")
   13.           └─queue$poll(Inf)
   14.             └─base::lapply(...)
   1
[cut here]

Since March 2026

• backports
• checkmate
• cli
• data.table
• digest
• lgr
• mlr3
• mlr3misc
• paradox
• R6

Installing it adds

24 packages, 16.1 MB of source

Of those, need compilation

11

• bbotk
• bestNormalize
• dimRed
• evaluate
• fastICA
• future
• GenSA
• ggplot2
• glmnet
• htmlwidgets
• igraph
• kernlab
• knitr
• lme4
• MASS
• methods
• mlbench
• mlr3filters
• mlr3learners
• mlr3measures
• nloptr
• NMF
• quanteda
• ranger
• RANN
• rmarkdown
• rpart
• RSpectra
• smotefamily
• splines
• stopwords
• testthat
• themis
• visNetwork
• vtreat

Lines by kind (60,520)

Lines by language

Languages are counted from file extensions.

Files

378

R functions

210

Object systems

S3: 86 methods, R6: 103 classes

Median

2

Mean

3.72

Highest

167

Median lines per function

7

R functions

210

Calls

122

Unconnected

102

Groups

10

Has tests

Yes

Test frameworks

testthat, tinytest

Test cases

572

Snapshot tests

6

Test lines per line of code

0.91

• Adding new PipeOps

README

Yes, 490 words

Release notes file

Yes

Help pages

132

Lines of documentation

20,652

Commits

4,384

Open pull requests

14

Tagged releases

12

Latest tagged release

v0.12.0, 5 Sep 2026

In development

Default branch at 0.12.0-9000, ahead of CRAN 0.12.0 (checked 2026-10-25)

Issues closed in

363 days (median)

Pull requests closed in

8 days (median)

Pull requests merged

89%

Created

10 Oct 2017

Licence file reads as

LGPL-3.0

Method, repository practices

Across mlr-org

• Fix: Removed an R-devel deprecation warning from PipeOpFilter and kept the preproc() examples compatible with the declared minimum R version.
• Fix: GraphLearner had the guards of its internal validation and internal tuning extractors swapped, so a Graph supporting only one of the two properties reported nothing for the property it did support.
• feat: GraphLearner and PipeOpLearner gained a $best_valid_scores field, and GraphLearner collects it from the wrapped PipeOps via $.extract_best_valid_scores(), so msr("best_valid_score") can be used with a GraphLearner.
• Added a temporary compatibility alias that registers the former pima task as diabetes for released mlr3 versions older than 1.8.0. Retrieving the alias warns users to update to mlr3 1.8.0 or newer. This alias will be removed in a later update of mlr3pipelines.
• Fix: Re-running registration (e.g. when mlr3 is reloaded) no longer removes PipeOp properties added to mlr_reflections$pipeops$properties by extension packages.
• Switched from using digest::digest() to using mlr3misc::calculate_hash() for calculating the hash and phash of PipeOps, Graphs, and GraphLearners.
• Fix: Corrected registration of FilterEnsemble in mlr_filters using .prototype_args.
• Fix: PipeOpTargetMutate and PipeOpTargetTrafoScaleRange now correctly transform internal validation tasks during training.
• feat: PipeOpSmote, PipeOpSmoteNC, PipeOpADAS, and PipeOpBLSmote can now handle columns with role "name" by assigning the name synthetic.<pipeop id> to generated rows.
• New PipeOpMaterialize that materializes the active Task view, reducing size and simplifying structure of the Task's DataBackend.
• Fix: mlr_pipeops$add() now saves the namespace environment from which it was called to allow delayed evaluation in as.data.table(mlr_pipeops).
• PipeOpNMF now correctly rejects features containing missing or infinite values with an informative error message. Features containing negative values are still dropped implictly, but now this produces a warning to use selector_non_negative() for explicit column selection. This will become an error in the future.
• New selectors selector_positive(), selector_negative(), selector_non_negative(), selector_non_positive(), selector_non_zero(), and selector_non_missing(). The numeric selectors support na_ignore to control whether missing values are ignored when checking if all values satisfy the selector's condition.
• mlr_graphs no longer overrides the inherited $add() method.
• Simplified error messages from internal function check_types().
• Removed deprecated greplicate() function. Use ppl("greplicate") instead.
• Fix: PipeOpIsomap now explicitly requires igraph, RSpectra, and RANN; skip checks in tests were also updated.
• Fix: as_learner.Graph now handles ... as an argument.
• feat: as_learner.Graph now suppots the discard_state argument, same as as_learner.Learner.
• feat: PipeOpTargetInvert now propagates the extra slot of input prediction objects to the output prediction object, if the inverter fun does not handle it already.
• feat: PipeOpDateFeatures gains new features is_month_start, is_month_end, is_quarter_start, is_quarter_end, is_year_start, is_year_end, and is_leap_year.
• Fix: PipeOpDateFeatures no longer creates all-NA cyclic time-of-day features for Date features.