Safer Decimal Number (In)Equality Testing Operators

The %d==%, %d!=% %d<%, %d>%, %d<=%, %d>=% (in)equality operators perform decimal (type "double") number truth testing.
They are virtually equivalent to the regular (in)equality operators,
==, !=, <, >, <=, >=,
except for 2 aspects:

1. The decimal number (in)equality operators assume that if the absolute difference between any 2 numbers x and y is smaller than the Machine tolerance, sqrt(.Machine$double.eps), then x and y should be consider to be equal.
   For example: (0.1 * 7) == 0.7 returns FALSE, even though they are equal, due to the way decimal numbers are stored in programming languages like 'R' and 'Python'.
   But (0.1 * 7) %d==% 0.7 returns TRUE.
   

2. Only numeric input is allowed, so characters are not coerced to numbers.
   I.e. 1 < "a" gives TRUE, whereas 1 %d<% "a" gives an error.
   For character equality testing, see %s==% from the 'stringi' package.
   

Thus these operators provide safer decimal number (in)equality tests.

There are also the x %d{}% bnd and x %d!{}% bnd operators, where bnd is a vector of length 2, or a 2-column matrix (nrow(bnd)==length(x) or nrow(bnd)==1).
The x %d{}% bnd operator checks if x is within the closed interval with bounds defined by bnd.
The x %d!{}% bnd operator checks if x is outside the closed interval with bounds defined by bnd.

Moreover, the function is_wholenumber() is added, to safely test for whole numbers.

Usage

x %d==% y

x %d!=% y

x %d<% y

x %d>% y

x %d<=% y

x %d>=% y

x %d{}% bnd

x %d!{}% bnd

is_wholenumber(x, tol = sqrt(.Machine$double.eps))

Arguments

x, y

numeric vectors, matrices, or arrays.

bnd

either a vector of length 2, or a matrix with 2 columns and 1 row, or else a matrix with 2 columns where nrow(bnd)==length(x) (or can be recycled to be nrow(bnd)==length(x)).
The first element/column of bnd gives the lower bound of the closed interval;
The second element/column of bnd gives the upper bound of the closed interval.

tol

a single, strictly positive number close to zero, giving the tolerance.

Value

For the %d...% operators:
A logical vector with the same dimensions as x, indicating the result of the element by element comparison.
NOTE: Inf by Inf and -Inf by -Inf comparisons with the %d...% operators return NA.

For is_wholenumber():
A logical vector with the same dimensions as x, indicating the result of the element by element comparison.
NOTE: Inf, -Inf, NaN and NA all return NA for is_wholenumber().

Details

The operators described in this page are defined in terms of the existing base Logic operators, and should therefore be compatible with (S3) classes that have method dispatches defined for relational operators.

See also

tinycodet_safer

Examples

x <- c(0.3, 0.6, 0.7)
y <- c(0.1 * 3, 0.1 * 6, 0.1 * 7)
print(x)
#> [1] 0.3 0.6 0.7
print(y)
#> [1] 0.3 0.6 0.7

x == y # gives FALSE, but should be TRUE
#> [1] FALSE FALSE FALSE
x != y # gives TRUE, should be FALSE
#> [1] TRUE TRUE TRUE
x > y # not wrong
#> [1] FALSE FALSE FALSE
x < y # gives TRUE, should be FALSE
#> [1] TRUE TRUE TRUE

# same as above, but here the results are correct:
x %d==% y # correct
#> [1] TRUE TRUE TRUE
x %d!=% y # correct
#> [1] FALSE FALSE FALSE
x %d<% y # correct
#> [1] FALSE FALSE FALSE
x %d>% y # correct
#> [1] FALSE FALSE FALSE
x %d<=% y # correct
#> [1] TRUE TRUE TRUE
x %d>=% y # correct
#> [1] TRUE TRUE TRUE

# check if numbers are in closed interval:
x <- c(0.3, 0.6, 0.7)
bnd <- cbind(x - 0.1, x + 0.1)
x %d{}% bnd
#> [1] TRUE TRUE TRUE
x %d!{}% bnd
#> [1] FALSE FALSE FALSE

# These operators work for integers also:
x <- 1L:5L
y <- 1L:5L
x %d==% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d!=% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d<% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d>% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d<=% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d>=% y
#> [1] TRUE TRUE TRUE TRUE TRUE

x <- 1L:5L
y <- x + 1L
x %d==% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d!=% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d<% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d>% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d<=% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d>=% y
#> [1] FALSE FALSE FALSE FALSE FALSE

x <- 1L:5L
y <- x - 1L
x %d==% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d!=% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d<% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d>% y
#> [1] TRUE TRUE TRUE TRUE TRUE
x %d<=% y
#> [1] FALSE FALSE FALSE FALSE FALSE
x %d>=% y
#> [1] TRUE TRUE TRUE TRUE TRUE

# is_wholenumber:
is_wholenumber(1:10 + c(0, 0.1))
#>  [1]  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE  TRUE FALSE