Creating valid raster from numpy array

[fbunt]

I am trying to create a valid in memory raster from a numpy array. I'm running into a problem where operations on my from-numpy raster involving other rasters read in using rioxarray.open_rasterio are returning empty DataArrays. As an example I have an identical raster created in ArcGIS that should compare as equal, but again I get an empty result. This seems to indicate that xarray thinks the two DataArrays aren't aligned. As far as I can tell, they are though. Writing my raster to disk with to_raster and then reading in again with open_rasterio causes everything to work like I expect. What am I missing in order to create a valid in-memory raster?

Example code:

import numpy as np
import xarray as xr
import rioxarray as rxr
from affine import Affine


cs = np.array(
    [
        [
            [1, 2, 2, 1, 3, 4],
            [4, 7, 5, -1, 2, 6],
            [1, 4, 5, -1, 5, 1],
            [5, 8, 7, 5, 6, 6],
            [7, 3, 2, 6, 4, 6],
            [1, 3, 4, 4, 3, 2],
        ]
    ]
)
coords = [list(range(d)) for d in cs.shape]
dims = ["band", "y", "x"]
xcs = (
    xr.DataArray(cs, coords=coords, dims=dims)
    .astype("int32")
    .rio.write_nodata(-1)
    .rio.write_transform(Affine(1.0, 0.0, -0.5, 0.0, 1.0, -0.5))
)
print(f"xcs:\n{xcs}\n")
print(f"xcs transform:\n{xcs.rio.transform()}")

# ESRI output raster that has identical data
ecs = rxr.open_rasterio("costs.tif")
print(f"ecs:\n{ecs}\n")
print(f"ecs transform:\n{ecs.rio.transform()}")

print(f"xcs == ecs:\n{xcs == ecs}:\n")

xcs.rio.to_raster("test.tif")
rxcs = rxr.open_rasterio("test.tif")
print(f"rxcs:\n{rxcs}\n")
print(f"rxcs == ecs:\n{rxcs == ecs}")

Output:

xcs:
<xarray.DataArray (band: 1, y: 6, x: 6)>
array([[[ 1,  2,  2,  1,  3,  4],
        [ 4,  7,  5, -1,  2,  6],
        [ 1,  4,  5, -1,  5,  1],
        [ 5,  8,  7,  5,  6,  6],
        [ 7,  3,  2,  6,  4,  6],
        [ 1,  3,  4,  4,  3,  2]]], dtype=int32)
Coordinates:
  * band         (band) int64 0
  * y            (y) int64 0 1 2 3 4 5
  * x            (x) int64 0 1 2 3 4 5
    spatial_ref  int64 0
Attributes:
    _FillValue:  -1

xcs transform:
| 1.00, 0.00,-0.50|
| 0.00, 1.00,-0.50|
| 0.00, 0.00, 1.00|

ecs:
<xarray.DataArray (band: 1, y: 6, x: 6)>
array([[[ 1.,  3.,  4.,  4.,  3.,  2.],
        [ 7.,  3.,  2.,  6.,  4.,  6.],
        [ 5.,  8.,  7.,  5.,  6.,  6.],
        [ 1.,  4.,  5., -1.,  5.,  1.],
        [ 4.,  7.,  5., -1.,  2.,  6.],
        [ 1.,  2.,  2.,  1.,  3.,  4.]]])
Coordinates:
  * band         (band) int64 1
  * x            (x) float64 0.0 1.0 2.0 3.0 4.0 5.0
  * y            (y) float64 5.0 4.0 3.0 2.0 1.0 0.0
    spatial_ref  int64 0
Attributes:
    _FillValue:    -1.0
    scale_factor:  1.0
    add_offset:    0.0

ecs transform:
| 1.00, 0.00,-0.50|
| 0.00,-1.00, 5.50|
| 0.00, 0.00, 1.00|

xcs == ecs:
<xarray.DataArray (band: 0, y: 6, x: 6)>
array([], shape=(0, 6, 6), dtype=bool)
Coordinates:
  * band         (band) int64
  * y            (y) float64 0.0 1.0 2.0 3.0 4.0 5.0
  * x            (x) int64 0 1 2 3 4 5
    spatial_ref  int64 0:

rxcs:
<xarray.DataArray (band: 1, y: 6, x: 6)>
array([[[ 1,  2,  2,  1,  3,  4],
        [ 4,  7,  5, -1,  2,  6],
        [ 1,  4,  5, -1,  5,  1],
        [ 5,  8,  7,  5,  6,  6],
        [ 7,  3,  2,  6,  4,  6],
        [ 1,  3,  4,  4,  3,  2]]], dtype=int32)
Coordinates:
  * band         (band) int64 1
  * x            (x) float64 0.0 1.0 2.0 3.0 4.0 5.0
  * y            (y) float64 0.0 1.0 2.0 3.0 4.0 5.0
    spatial_ref  int64 0
Attributes:
    _FillValue:    -1.0
    scale_factor:  1.0
    add_offset:    0.0

rxcs == ecs:
<xarray.DataArray (band: 1, y: 6, x: 6)>
array([[[ True,  True,  True,  True,  True,  True],
        [ True,  True,  True,  True,  True,  True],
        [ True,  True,  True,  True,  True,  True],
        [ True,  True,  True,  True,  True,  True],
        [ True,  True,  True,  True,  True,  True],
        [ True,  True,  True,  True,  True,  True]]])
Coordinates:
  * y            (y) float64 0.0 1.0 2.0 3.0 4.0 5.0
  * band         (band) int64 1
  * x            (x) float64 0.0 1.0 2.0 3.0 4.0 5.0
    spatial_ref  int64 0

costs.tif.gz

[snowman2]

The main difference I see is that the coordinate order is different in the y axis. I would recommend sorting it.
Also, this may help: https://corteva.github.io/rioxarray/stable/examples/reproject_match.html

Update: the band is different

[fbunt]

Setting a CRS and then using reproject_match has the same result :(

[snowman2]

What happens if you skip the write_transform step? rioxarray will determine that from the coords.

[snowman2]

Looks like the value for band is different.

[fbunt]

Still gives an empty result

[fbunt]

Oh wow. I can't believe I missed that. It was the band. Thank you! And sorry for wasting your time