If a numeric variable has only integer values, then make it an integer.

Users often accidentally create floating point numeric variables when they really mean integers, such as c(1, 2, 3), when they should have done c(1L, 2L, 3L). Before running as.integer() to coerce the variable, we'd rather be polite and ask the variable "do you mind being treated as if you are an integer?" This function checks to see if the variable is "close enough" to being an integer, and then coerces as integer. Otherwise, it returns NULL. And issues a warning.

safeInteger(
  x,
  tol = .Machine$double.eps,
  digits = 7,
  vmax = .Machine$integer.max,
  verbose = FALSE
)

Arguments

x

a numeric variable

tol

Tolerance value. Defaults to Machine$double.eps. See details.

digits

Digits value passed to the zapsmall function. Defaults to 7.

vmax

Maximum value allowed for an integer. Defaults to Machine$integer.max.

verbose

Default FALSE: print warnings about x

Value

Either an integer vector or the original variable

Details

First, calculate absolute value of differences between x and as.integer(x). Second, find out if the sum of those differences is smaller than tol. If so, then x can reasonably be coerced to an integer.

Be careful with the return. The correct return value for variables that should not be coerced as integer is uncertain at this point. We've tested various strategies, sometimes returning FALSE, NULL, or just the original variable.

Author

Paul Johnson <pauljohn@ku.edu> and Ben Kite <bakite@ku.edu>

Examples

x1 <- c(1, 2, 3, 4, 5, 6)
is.integer(x1)
#> [1] FALSE
is.double(x1)
#> [1] TRUE
is.numeric(x1)
#> [1] TRUE
(x1int <- safeInteger(x1))
#> [1] 1 2 3 4 5 6
is.integer(x1int)
#> [1] TRUE
is.double(x1int)
#> [1] FALSE
is.numeric(x1int)
#> [1] TRUE
x2 <- rnorm(100)
x2int <- safeInteger(x2)
head(x2int)
#> NULL
x3 <- factor(x1, labels = c(LETTERS[1:6]))
x3int <- safeInteger(x3)