Downsampling with terra to arbitrary grids, resample() and project() only subset

I want to downsample model output to a coarser, arbitrary new grid. aggregate() does not do the trick, as I want to arrive at a specific grid. For some reason, resample() and project() only return a "zoomed" subset and not the desired product: the aggregated/downsampled data on a new grid.

I feel that the solution must be very obvious, but i could not find anything on stackoverflow/rseek at all.

Here is some example code:

library(terra)
# dummy starting raster
raw <- rast(ext(c(-0.5625, 359.4375, -89.70215791342, 89.70215791342)),nrow=160,ncol=320)
# with some "structure" to see the changes
values(raw) <- (1:ncell(raw))/10000+c(rep(1,ncell(raw)/4),rep(2,ncell(raw)/4),rep(1,ncell(raw)/4),rep(2,ncell(raw)/4))
# correct specs:
raw
plot(raw)

#desired new projection(?)
newgrid <- rast(xmin=1,xmax=128,ymin=1,ymax=64,nrow=64,ncol=128)#,crs="wgs84"

#modify with resample or project
new1 <- project(raw,newgrid,method="bilinear")
new2 <- resample(raw,newgrid,method="bilinear")

#that created only a subset/"zoom" and not a new map with different grid
plot(new1)
plot(new2)

I tried aggregate, resample and project functions of terra package in different variants. I want to transform my starting rast to a new one with the desired extent, resolution, etc. Aggregation does not work, because I want to have a specific grid; resample and project return only a subset instead of the full set of values on the new grid.

sessioninfo:

As @Chris noted, the extent of newgrid is smaller than that of raw, so only the overlapping areas will be resampled(). Note that even though terra::rast() has assigned WGS84/EPSG:4326 to your SpatRasters by default, it is good practice to do it explicitly.

To illustrate the issue:

library(terra)
library(ggplot2)
library(tidyterra)

# dummy starting raster
raw <- rast(ext(c(-0.5625, 359.4375, -89.70215791342, 89.70215791342)), 
            nrow = 160,
            ncol = 320,
            crs = "EPSG:4326")

values(raw) <- (1:ncell(raw))/10000 + c(rep(1,ncell(raw)/4),rep(2,ncell(raw)/4),
                                        rep(1,ncell(raw)/4),rep(2,ncell(raw)/4))


newgrid <- rast(ext(1, 128, 1, 64),
                nrow = 64,
                ncol = 128,
                crs = crs(raw))

new <- resample(raw, newgrid, method = "bilinear")

# For illustrative purposes
v1 <- vect(ext(raw), crs = "EPSG:4326")
v2 <- vect(ext(newgrid), crs = "EPSG:4326")

ggplot() +
  geom_spatvector(data = v1,
                  aes(colour = "raw\nextent")) +
  geom_spatraster(data = new) +
  geom_spatvector(data = v2,
                  aes(colour = "newgrid\nextent"),
                  lwd = 1,
                  fill = NA) +
  scale_colour_manual(name = "", values = c("firebrick", "black")) +
  theme_void()

One way to ensure your resampled Spatraster retains its original extent:

# Create new Spatraster with lower resolution, same extent as raw
newgrid <- rast(ext(raw), nrow = 64, ncol = 128, crs = "EPSG:4326")

# Resample raw using newgrid
new <- resample(raw, newgrid, method = "bilinear")

raw
# class       : SpatRaster 
# dimensions  : 160, 320, 1  (nrow, ncol, nlyr)
# resolution  : 1.125, 1.121277  (x, y)
# extent      : -0.5625, 359.4375, -89.70216, 89.70216  (xmin, xmax, ymin, ymax)
# coord. ref. : lon/lat WGS 84 (EPSG:4326) 
# source(s)   : memory
# name        :  lyr.1 
# min value   : 1.0001 
# max value   : 7.1200 

new
# class       : SpatRaster 
# dimensions  : 64, 128, 1  (nrow, ncol, nlyr)
# resolution  : 2.8125, 2.803192  (x, y)
# extent      : -0.5625, 359.4375, -89.70216, 89.70216  (xmin, xmax, ymin, ymax)
# coord. ref. : lon/lat WGS 84 (EPSG:4326) 
# source(s)   : memory
# name        :    lyr.1 
# min value   : 1.032195 
# max value   : 7.087905