Intended behaviour "crop" with extend=TRUE?

[jvandoninck]

When using crop with a raster and an extend that goes beyond the raster, it returns the original raster when extend=TRUE.

library(terra)
r <- rast(nrow=100, ncol=100, extent=ext(-100, 100, -45, 45))
e <- ext(0,150, 0, 30)
crop(r,e)
crop(r,e,extend=TRUE)

I would have expected that it extends the cropped raster to have the extent of e, as with:

extend(crop(r,e), e)

Is this the intended behaviour or a bug?

[rhijmans]

Thank you, that is a bug. The function did not expand rasters with no cell values.
I now see:

library(terra)
#terra 1.7.24
r <- rast(nrow=10, ncol=10, extent=ext(0, 10, 0, 10))

e <- ext(-10, 15, -10, 15) 
ext(crop(r, e))
#SpatExtent : 0, 10, 0, 10 (xmin, xmax, ymin, ymax)
ext(crop(r, e, extend=TRUE))
#SpatExtent : -10, 15, -10, 15 (xmin, xmax, ymin, ymax)
 
e <- ext(-10, 5, -10, 5)
ext(crop(r, e))
#SpatExtent : 0, 5, 0, 5 (xmin, xmax, ymin, ymax)
ext(crop(r, e, extend=TRUE))
#SpatExtent : -10, 5, -10, 5 (xmin, xmax, ymin, ymax)
 
e <- ext(0, 5, 0, 5)
ext(crop(r, e))
#SpatExtent : 0, 5, 0, 5 (xmin, xmax, ymin, ymax)
ext(crop(r, e, extend=TRUE))
#SpatExtent : 0, 5, 0, 5 (xmin, xmax, ymin, ymax)

[jvandoninck]

Thank you.
Out of curiosity, has it been considered changing the error message when extents of x and y to not overlap, but extend=TRUE, into returning an empty raster with ext=y? Perhaps with a warning message.

library(terra)
#terra 1.7.24

r <- rast(nrow=10, ncol=10, extent=ext(0, 10, 0, 10))
e <- ext(-10, -5, -10, -5)

crop(r, e, extend=TRUE)
#Error: [crop] extents do not overlap

I can imagine a few cases where this behaviour would be useful, though it can obviously easily be circumvented using a prior check based on intersect(ext(r), e), or simply using crop(extend(r,e),e), though the latter is potentially very inefficient in cases with a large distance between ext(r) and e.