Wrong result from calc_SPIE when there are only singletons

[AlbanSagouis]

I'm using the latest version of mobr (c5fd569) and R 4.0.2.

Here are 2 communities with 30 species and only singletons. One community has 3 and the other has 5. calc_SPIE() gives NA for the fist community which is the expected behaviour but not for the second one because the PIE value is not accurately = 1.

library(mobr)
x <- matrix(0, nrow = 2, ncol = 30)
x[1, 1:3] <- 1
x[2, 1:5] <- 1

calc_PIE(x)
[1] 1 1
calc_SPIE(x)
[1]           NA 9.007199e+15

calc_SPIE(x) should have given two NAs.

If we look into calc_SPIE():

  PIE = calc_PIE(x, replace = FALSE)
  SPIE = 1/(1 - PIE)
  SPIE 
[1]          Inf 9.007199e+15

Not the wanted result because:

  PIE == 1
[1]  TRUE FALSE

Seems to be a rounding problem, confirmed here:

  PIE = round(calc_PIE(x, replace = FALSE), 0)
  SPIE = 1/(1 - PIE)
  SPIE
[1] Inf Inf

I have no fix to commit to solve the issue.
Hope it helps,

[dmcglinn]

Thank you @AlbanSagouis. It seems that the simplest solution here is just to round PIE to the 10th digit or something. I propose we add PIE = round(PIE, 10) to the function calc_PIE. @T-Engel any concerns?

[T-Engel]

Thanks for spotting this, Alban. Yes, this is because of issues with the floating-point representation of numbers in R.
We used the "==" operator to check if PIE was 1 and that is not always safe as explained here.

I changed the code to use the "all.equal" notation instead. See here: fa996a8
I hope this has fixed it.