Skip to contents

Changelog

Source: NEWS.md

distplyr 0.3.0

Every verb is now built on distionary’s support objects, so a transformed distribution says where its probability lives: support(), its atoms, and discrete or mixed moments all work on the result. Verbs no longer state a range or .vtype; both are read from the support. invert(), for one, derives the support of 1 / X for a distribution spanning zero by mapping each side of it separately. Requires distionary 0.2.0.

New verbs

  • trim_left() and trim_right() condition a distribution on one side of a point, with knot saying what becomes of the mass sitting exactly on it: "keep" it (the default, since a point is not to one side of itself), "discard" it with the trimmed side, or "split" it. See ?trim.

  • graft_left() and graft_right() replace one end of a body with another model of that end. Name the tail either as tail_excess, measured from the knot and moved there, or as tail_absolute, on the body’s own scale and conditioned beyond the knot. knot names the side that mass sitting exactly on the knot belongs to — the "body" (the default), the "tail", or "split" between them. See ?graft.

Other changes

  • Distributions built by a verb print as a summary rather than unfolding every distribution they were built from: each is named with pretty_name(), and a mixture’s components are listed with their weights.

Bug fixes

  • mix(), maximize() and minimize() no longer treat different components as the same one. Results may change for transformed distributions.

  • Verbs no longer discard a distribution that merely carries the name "Null". Null-ness is read with is.na().

distplyr 0.2.0

CRAN release: 2025-12-08

  • Initial CRAN release.

distplyr 0.1.5

  • Updates to package infrastructure in the process of migrating to a new GitHub Organization.

distplyr 0.1.4

  • Fix graft distributions so that they can evaluate on NA.

distplyr 0.1.3

  • Default evaluation methods, and base distributional forms, have been moved to a new package, distionary. distplyr focusses on manipulation verbs only.
  • Math method now applies to finite distributions.
  • Ops methods are now available for arithmetic operations (+, -, *, and /) on a single distribution, along with the verbs shift(), multiply(), invert(), and flip().
  • graft_left() and graft_right() are fully functional, and slice_left() and slice_right() are now also available.

distplyr 0.1.2

  • If you have the tibble package installed, distplyr will now output tibbles wherever data frames were previously output.

Breaking changes

  • The get_ prefix has been removed from distributional quantities. get_mean() is now mean(), etc.
    • For now, the get_ prefix still holds for distributional representations, like get_cdf().
  • Make your own distribution object with distribution() instead of dst(), and checked with is_distribution().

distplyr 0.1.1

This patch both fixes some problems in the previous release, as well as offering a step towards a bigger expansion.

  • Some change in the functional representations:
    • Changed random number generation from randfn, a functional representation, to the realise() and realize() functions.
    • Changed probfn representation to be more specific: pmf or density
  • Added the enframe suite of functions.
  • Implement the beginnings of being able to specify your own distribution, with the set_ suite of functions, after making an empty distribution with dst().

Additionally, there’s some internal rearrangement, where the get functions call the eval functions, not vice versa.

distplyr 0.1.0

The first version of distplyr is now available! Its functionality is rather limited at the moment, but is still useful, especially for its capability to handle a discrete component of a distribution. Here are the main features:

  • Base distributions include step distributions, Gaussian, Uniform, and generalized Pareto.
  • Operations include grafting (right) and mixing
  • Distribution properties included are moment-related quantities, and extreme value index.
  • Distribution representations are mostly comprehensive, perhaps only missing mean excess function and moment generating function.

Take a look at the “Vision” vignette to get a sense of where this package is headed.