zonal removing NA/NaN values from x when using custom fun

[mwip]

Problem

When using zonal on a raster x that contains NA or NaN values, these are omitted prior to passing x into in custom functions. This does not happen when using built-in functions like fun = "isNA".

Expected behavior

This, to me, is unexpected behavior, as I would expect the vector passed to the custom function to include NA and NaN.

Example

See this minimal reproducible example:

library(terra)

values_rast <- rast(matrix(c(1, 2, 3, NA), ncol = 2))
names(values_rast) <- "values"

zones_rast <- rast(matrix(c(1, 1, 2, 2), ncol = 2))
names(zones_rast) <- "zones"

plot(c(values_rast, zones_rast), colNA = "red")

zonal(x = values_rast, z = zones_rast, function(x) {
  print(x)
  return(sum(is.na(x)))
})

# [1] 1 2
# [1] 3     # <------ unexpected, should be c(3, NA)
#   zones values
# 1     1      0
# 2     2      0  # <------ unexpected, should be 1

zonal(x = values_rast, z = zones_rast, fun = "isNA")
#   zones values
# 1     1      0
# 2     2      1  # <---------- expected

System details

OS/Kernel: Linux 5.15, Ubuntu 20.04, x86_64,
Tested on terra version 1.7.29 (latest version on CRAN) and 1.7.30 (via devtools::install_github("rspatial/terra"))

The issue was reproducible on a Windows machine, too.

[rhijmans]

Thanks, I now get:

library(terra)
v <- rast(matrix(c(1, 2, 3, NA), ncol = 2))
names(v) <- "values"
z <- rast(matrix(c(1, 1, 2, 2), ncol = 2))
names(z) <- "zones"

zonal(v, z, \(x) sum(is.na(x)))
#  zones values
#1     1      0
#2     2      1

zonal(v, z, "isNA")
#  zones values
#1     1      0
#2     2      1


zonal(v, z, \(x) mean(x))
#  zones values
#1     1    1.5
#2     2     NA

zonal(v, z, "mean")
#  zones values
#1     1    1.5
#2     2     NA


zonal(v, z, \(x) mean(x, na.rm=T))
#  zones values
#1     1    1.5
#2     2    3.0

zonal(v, z, \(x, ...) mean(x, ...), na.rm=T)
#  zones values
#1     1    1.5
#2     2    3.0

zonal(v, z, "mean", na.rm=T)
#  zones values
#1     1    1.5
#2     2    3.0

[mwip]

I can confirm this fixed the issue. Thank you for taking the time and resolving this that quickly. I really appreciate your dedication!