Use Raw Vectors to Minimize Memory Consumption of Factors

Uses raw vectors to minimize memory consumption of categorical variables with fewer than 256 unique values. Useful for analysis of large datasets involving variables such as age, years, states, countries, or education levels.

factor256

The goal of factor256 is to minimize the memory footprint of data analysis that uses categorical variables with fewer than 256 unique values.

Installation

You can install the development version of factor256 from GitHub with:

# install.packages("devtools")
devtools::install_github("HughParsonage/factor256")

Example

This is a basic example which shows you how to solve a common problem:

library(factor256)
x <- factor256(LETTERS)
typeof(x)
#> [1] "raw"
identical(recompose256(x), LETTERS)
#> [1] TRUE

library(data.table)
DT <-
  CJ(Year = 2000:2020,
     State = rep_len(c("WA", "SA", "NSW", "NT", "TAS", "VIC", "QLD"), 1000),
     Age = rep_len(0:100, 10000))
# pryr::object_size(DT)
# 3.36GB
for (j in seq_along(DT)) {
  set(DT, j = j, value = factor256(.subset2(DT, j)))
}
# pryr::object_size(DT)
# 630 MB