Classify integer columns

Posted on by

The integerCols function below detects which numeric columns in a data frame contain only whole numbers, and converts those columns to integer class, so that they take up less space. It uses the multConvert function, which must be loaded too. Both functions are included in the latest version of the fuzzySim package (Barbosa, 2015).

integerCols <- function(data) {
  is.wholenumber <- function(x, tol = .Machine$double.eps ^ 0.5) {
    abs(x - round(x)) < tol
  }  # from ?is.integer examples
  all.cols <- 1:ncol(data)
  integer.cols <- rep(0, ncol(data))
  for (i in all.cols) {
    x <- na.omit(data[ , i])
    if(!is.numeric(x)) next
    if(!all(is.finite(x))) next
    if(min(is.wholenumber(x) == 1)) integer.cols[i] <- 1
  }
  multConvert(data, conversion = as.integer, cols = all.cols[integer.cols == 1])
}

[presented with Pretty R]

Usage example:

dat <- data.frame(
  a = 1:10,
  b = as.numeric(1:10),
  c = seq(0.1, 1, 0.1),
  d = letters[1:10]
)
 
str(dat)  # b is classified as 'numeric' although it contains only whole numbers
 
dat2 <- integerCols(dat)
 
str(dat2)  # b now classified as 'integer'

References:

Barbosa A.M. (2015) fuzzySim: applying fuzzy logic to binary similarity indices in ecology. Methods in Ecology and Evolution, 6: 853-858

Comment