refactor

refactor is an R package which aims to provide better handling of factors. Though R does have a special data type for factors, it isn't always explicity catered for in commonly used R functions, which can lead to unexpected and undesirable outcomes (e.g. see this Win-Vector blog post). This observation formed the inspiration for 're'-factor: which is essentially to 're'-visit functions likely to be used with factor data and where possible to wrap, extend or override them in order to better cater for factor data, or at the very least to provide warnings when the integrity of the data could be compromised by an operation.

Installation

refactor is not available on CRAN but you can easily install the latest development version from github using devtools:

# install.packages("devtools")
devtools::install_github("jonmcalder/refactor")

Overview

Below are a few examples to illustrate some scenarios in which refactor can improve on base R's handling of factors.

Please see the vignette for more details: vignette("refactor", package = "refactor")

Better factor creation with cfactor()

• get warnings whenever provided factor levels don't fully match those present in the data

string <- c("a", "b", "c")
factor(string, levels = c("b", "c", "d"))
#> [1] <NA> b    c   
#> Levels: b c d
cfactor(string, levels = c("b", "c", "d"))
#> Warning: the following levels were empty: 
#>  d
#> Warning: the following levels were removed: 
#>  a
#> [1] <NA> b    c   
#> Levels: b c d

• numerical sequences in strings are detected and utilized to obtain correct orderings for factor levels

hard_to_detect <- c("EUR 21 - EUR 22", "EUR 100 - 101", "EUR 1 - EUR 10", "EUR 11 - EUR 20")
factor(hard_to_detect, ordered = TRUE)
#> [1] EUR 21 - EUR 22 EUR 100 - 101   EUR 1 - EUR 10  EUR 11 - EUR 20
#> 4 Levels: EUR 1 - EUR 10 < EUR 100 - 101 < ... < EUR 21 - EUR 22
cfactor(hard_to_detect, ordered = TRUE)
#> [1] EUR 21 - EUR 22 EUR 100 - 101   EUR 1 - EUR 10  EUR 11 - EUR 20
#> 4 Levels: EUR 1 - EUR 10 < EUR 11 - EUR 20 < ... < EUR 100 - 101

Extension of the generic cut method to handle (discrete) integer data

• generate 'natural' intervals for factors when using cut on integer vectors
  • e.g. [1,3], [4,6], [7,9] as opposed to (0.5, 3.5], (3.5, 6.5], (6.5, 9.5]
• label factor levels more intuitively
  • e.g. 1-3, 4-6, 7-9 as opposed to [1,3], [4,6], [7,9], or (0,3], (3,6], (6-9]

x_int <- 1:9
cut.default(x_int, breaks = 3)
#> [1] (0.992,3.67] (0.992,3.67] (0.992,3.67] (3.67,6.33]  (3.67,6.33] 
#> [6] (3.67,6.33]  (6.33,9.01]  (6.33,9.01]  (6.33,9.01] 
#> Levels: (0.992,3.67] (3.67,6.33] (6.33,9.01]
cut(x_int, breaks = 3)
#> [1] 1-3 1-3 1-3 4-6 4-6 4-6 7-9 7-9 7-9
#> Levels: 1-3 4-6 7-9

Extension of the generic cut method to handle (non-numeric) ordered data

• cut usually requires numeric data, but refactor extends cut to handle other ordered data

some_letters <- factor(c('d','e','f','a','b','c','g','h','i'), ordered = TRUE)
cut(some_letters, breaks = c('a','c','f','i'), include.lowest = TRUE, ordered_result = TRUE)
#> [1] d-f d-f d-f a-c a-c a-c g-i g-i g-i
#> Levels: a-c < d-f < g-i

Contributions

Suggestions and feedback are most welcome.

Feel free to open an issue if you want to request a feature or report a bug, and make a pull request if you can contribute.