`igneous image xfer` to change chunk size on 4-channel precomputed image creates empty data

[chourroutm]

Hi, I created a Neuroglancer precomputed dataset with 4 channels (RGBA) from a per-slice dataset, this is its info:

{
  "data_type": "uint8",
  "num_channels": 4,
  "scales": [
    {
      "chunk_sizes": [
        [
          1024,
          1024,
          1
        ]
      ],
      "encoding": "raw",
      "key": "16545.0_16545.0_16545.0",
      "resolution": [
        16545.0,
        16545.0,
        16545.0
      ],
      "size": [
        4655,
        5161,
        4991
      ],
      "voxel_offset": [
        1999,
        2769,
        7720
      ]
    }
  ],
  "type": "image"
}

I tried to change the chunk size to (256,256,256) with the following command:

igneous -p 50 image xfer --mip 0 --skip-downsample --chunk-size 256,256,256 --fill-missing /data/dec_brain_slices.precomputed/ /data/dec_brain.precomputed/

But the output is empty with (0,0,0,0) values everywhere in Neuroglancer. This is its info:

{
  "data_type": "uint8",
  "num_channels": 4,
  "scales": [
    {
      "chunk_sizes": [
        [
          256,
          256,
          256
        ]
      ],
      "encoding": "raw",
      "key": "16545.0_16545.0_16545.0",
      "resolution": [
        16545.0,
        16545.0,
        16545.0
      ],
      "size": [
        4655,
        5161,
        4991
      ],
      "voxel_offset": [
        1999,
        2769,
        7720
      ]
    }
  ],
  "type": "image"
}

Thus, is there an issue with my parameters in the igneous command?

[chourroutm]

Hi @william-silversmith, do you need more information about this issue?

[william-silversmith]

Hi sorry for the delay, I'll take a look at this next week! We don't use many 4-channel volumes in my lab, so I suppose that code path isn't well tested...