Skip to content

Latest commit

 

History

History
22 lines (20 loc) · 2.7 KB

new-in-this-release.md

File metadata and controls

22 lines (20 loc) · 2.7 KB
Raw

What's new in plyrmr 0.2.0

Features

  • redesigned the API to avoid duplication, split jobs between the all powerful transmute, the transform-like but humbly-named bind.cols and the specialized, dplyr-lifted select.
  • magic.wand now knows special tricks to deal with non-standard argument evaluation
  • .data variable in transmute, bind.cols and where expressions allows to refer to the whole chunk of data, not just one column, for instance to select odd columns: transmute(mtcars, odd = .data[, c(T,F)]).
  • new pipe operators %|% and %!%: select(group(select(input(data), select2.args), group.args), select1.args) can be written like data %|% input %|% select(select1.args) %|% group(group.args) %|% select(select2.args) in a unix shell like fashion. a %!% b is a shorthand for function(x) x %|% a %|% b. Works with complex expressions and piping into arguments other than the first using the special . variable, as in: LETTERS %|% grep("A", .) or LETTERS %|% paste0(.,.,.). Similar to features found in packages vadr, dplyr and magrittr.
  • optimizations for nested grouping, saves a job in certain circumstances.
  • where takes a single argument now. The multiple argument form just calculated a logical & of all the arguments, which is an arbitrary default and one more thing to remember. Therefore where(data, cond1, cond2, cond3) becomes where(data, cond1 & cond2 & cond3). More clear, same verbosity and symmetric to where(data, cond1 | cond2 | cond3), and we all love symmetries, don't we?
  • transmute allows fractional recycling transmute(mtcars, even = c(FALSE, TRUE), triad = 1:3, .cbind = TRUE).
  • .columns argument for transmute, allows to select multiple cols with a string character: transmute(mtcars, .columns=c("carb", "cyl", "mpg"))`.
  • removed pryr dependency with modest borrowing of code courtesy of the always supportive @hadley
  • prettier default names for new columns
  • better composition rules for mergeable and non-mergeable operations, information about mergeability is stored in the operation itself. Mergeability in this context is when an operation is associative and commutative, which allows an important optimization in Mapreduce.
  • operations also encapsulate information about being able to deal with multiple groups at once, but this is still work in progress.

Bugs

  • name changes in recursive group: mtcars %|% input %|% gather %|% select(sum(carb)) just works.
  • column duplication
  • more aggressive freezing of environments to make delayed evaluation and pipe optimization semantically neutral
  • incorrect writing to "native" format when custom expected
  • uncountable corner cases (things like 0-rows, 0-columns etc), but there may be more.