Skip to contents

RS - Faster classes for R

The name RS is a combination of:

  • The R and S languages (S is R’s predecessor).
  • .rs - the file extension used for Rust files, and the TLD of Rust websites.

Install

Any of the following should work:

pak::pak("avhz/RS")
## or
renv::install("avhz/RS")
## or
devtools::install_github("avhz/RS")

Usage 

Class(
    "Dog",
 
    # Fields
    name  = t_char,
    age   = t_int,
    breed = t_char,
 
    # Methods
    bark = function(.self) cat(.self@name, "goes woof!\n")
)
 
fluffy <- Dog(name = "Fluffy", age = 3L, breed = "Golden Retriever")

Note that the fields are special objects like t_int, t_char, etc. These objects allow for attribute type validation.

If you pass an incorrect type for one of the fields, you will get an error like:

fluffy <- Dog(name = 1, age = 3L, breed = "Golden Retriever")
"Invalid type <'double'> passed for field <'name'>."
> fluffy@name
[1] "Fluffy"
> fluffy@age
[1] 3
> fluffy@breed
[1] "Golden Retriever"
> fluffy@bark()
Fluffy goes woof!

I used the slot operator @ for accessing attributes in RS, because the common $ operator is so pervasive in R code that I wanted an operator that made it more obvious that the object is an RS class, rather than a list, dataframe, environment, R6 class, etc.

Note: Inheritance is not currently supported and most likely won’t be unless there is overwhelming demand for it. Instead, composition is supported and preferred.

Benchmarks

The following shows the iterations per second for class instances for a number of R OOP libraries.

Note #1: the other libraries may provide more features and are currently more mature/complete offerings, in particular R6.

Note #2: creating Python class instances via reticulate inside the R interpreter is significantly slower than doing it directly in Python. Python classes are still much, much faster than my RS package (~30-40x faster when I last checked).