In romainfrancois/hybrid: What the Package Does (One Line, Title Case)

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

hybrid

The goal of hybrid is to ...

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("romainfrancois/hybrid")

Motivation

eval_hybrid() (temporary name) evaluates an expression with eval_tidy() with a data mask made of hybrid leaves. An hybrid leaf is essentially the ptype of a column marked with the "hybrid" class.

Giving leaves the hybrid class makes it possible to dispatch based on that class, so mean.hybrid and +.hybrid (currently for this POC) when passed hybrid trees (leaves included) make hybrid trees.

library(hybrid)

eval_hybrid(iris, mean(Sepal.Length))

For functions that don't define hybrid methods, the computation is however as quick as possible yet still maintaining some type stability because the ptype is used in place of the real column.

eval_hybrid(iris, median(Sepal.Length))

The idea is that standard R evaluation and dispatch takes place and if we end up with an hybrid object, then perhaps the c++ code can handle it, otherwise fall back.

# these return an hybrid object because +() and mean() can handle hybrid
eval_hybrid(iris, mean(Sepal.Length))
eval_hybrid(iris, Sepal.Length + Sepal.Width)
eval_hybrid(iris, Sepal.Length + Sepal.Width + Petal.Length)
eval_hybrid(iris, mean(Sepal.Length + Sepal.Width))

# not sensible to the call ast, as is evaluation based
plus <- function(x, y) {
  x + y
}

eval_hybrid(iris, identity(mean(Sepal.Length)))
eval_hybrid(iris, plus(Sepal.Length, Sepal.Width))

# No max() hybrid function so we get something not hybrid
eval_hybrid(iris, mad(Sepal.Length))

# + does not handle the pattern `hybrid + not-hybrid`
eval_hybrid(iris, mean(Sepal.Length) + 1)