unexpected maximum possible values for terra::writeRaster according to datatype

user_4003696 0
Viewed 171

When using terra::writeRaster, the maximum possible values allowed for writing depend on the datatype (INT1U, INT2S, INT2U...).

The documentation tells that "When writing integer values the lowest available value (given the datatype) is used [to store NA, I suppose] for signed types, and the highest value is used for unsigned values.". This should give the following range for unsigned types:

INT1U : 0-254 (2^8-1, minus one for NA storage)

INT2U : 0-65,534 (2^16-1, minus one for NA storage)

INT4U : 0-4,294,967,294 (2^32-1, minus one for NA storage)

However, for unsigned datatypes INT2U and INT4U, the maxima I observed on my machine do not fit these expectations:

INT2U : 65,532 INT4U : 4,294,967,292

Why this unexpected maximum values? I ask the question because it is not insignificant, for safe code writing, to exactly know these maximum values before writing files.

I am working under Windows 10. Here is a couple of code lines that I used to check:

library(terra)
terra version 1.3.4
Warning message:
package ‘terra’ was built under R version 4.0.5 

r <- rast(ncols=1, nrows=2)
values(r) <- c(65532,65533)

writeRaster(r,"test.tif",wopt=list(datatype="INT2U"))

t <- rast("test.tif")
values(t)
     lyr.1
[1,] 65532
[2,]   NaN
terra raster
1 Answers

With the development version I now get the expected result

library(terra)
r <- rast(ncols=1, nrows=4)
values(r) <- 65533:65536

2-byte unsigned integer

x <- writeRaster(r,"test.tif", datatype="INT2U", overwrite=TRUE)
values(x)
#     lyr.1
#[1,] 65533
#[2,] 65534
#[3,]   NaN
#[4,]   NaN

x <- writeRaster(r,"test.tif", datatype="INT2U", NAflag=0, overwrite=TRUE)
values(x)
#[1,] 65533
#[2,] 65534
#[3,] 65535
#[4,]   NaN

4-byte unsigned integer

values(r) <- 4294967293:4294967296
x <- writeRaster(r,"test.tif", datatype="INT4U", overwrite=TRUE)
values(x)
#          lyr.1
#[1,] 4294967293
#[2,] 4294967294
#[3,]        NaN
#[4,]        NaN


x <- writeRaster(r,"test.tif", datatype="INT4U", NAflag=0, overwrite=TRUE,)
values(x)
#          lyr.1
#[1,] 4294967293
#[2,] 4294967294
#[3,] 4294967295
#[4,]        NaN
user_6352450
Related