write_image assumes particular axisorder

[k-dominik]

Heyo, I just tried to naively write an ndarray to a an ome.zarr file, but didn't manage. My array was something like numpy.zeros((100, 200, 1), dtype='uint8') where I considered axes to be yxc.

of course when calling write_image I specified the axes argument. The scaler, however, will assume a certain axisorder. This leads to weird errors that are not super obvious.

I'd say this should be documented, or circumvented (e.g. making Scaler aware of axes, using a temporary adding the appropriate axes, etc...)

[will-moore]

Hi @k-dominik thanks for the feedback.
write_image should be doing a better job of validating the axes since the NGFF spec at https://ngff.openmicroscopy.org/latest/#image-layout says
All image arrays must be up to 5-dimensional with dimension order (t, c, z, y, x).
The scaler currently only downsamples in y and x dimensions, so it will certainly give strange behaviour if either of those are 1.

[k-dominik]

Hi @will-moore,

thx for the quick reply. I'd argue that this should be part of the documentation of the function - users might not necessarily the spec, whereas it's natural to read a docstring. Also, I'd argue it could also be totally natural to expect write_image to do all necessary conversions. Or at least, a meaningful exception should be thrown, as you suggest.

Cheers
Dominik