Improve show method for Zattrs

[Artur-man]

Towards improving show method for Zattrs (#83). Lets discuss before moving forward, but no rush ...

Here are some examples:

> meta(image(x,1))
class: Zattrs
axes(3): c:channel y:space x:space
channels(3): 0 1 2
transformations(1): global
datasets(1): 0
> meta(image(x,2))
class: Zattrs
axes(3): c:channel y:space x:space
channels(3): 0 1 2
transformations(1): global
datasets(3): 0 1 2
> meta(label(x,1))
class: Zattrs
axes(2): y:space x:space
transformations(5): global scale translation affine sequence
datasets(1): 0
> meta(label(x,2))
class: Zattrs
axes(2): y:space x:space
transformations(1): global
datasets(3): 0 1 2
> meta(point(x))
class: Zattrs
axes(2): x y
transformations(1): global
> meta(shape(x))
class: Zattrs
axes(2): x y
transformations(1): global

[Artur-man]

Perhaps a slightly different show method for datasets under zattrs:

class: Zattrs
axes(3): c:channel y:space x:space
transformations(1): global
datasets(3):
- path(3): 0 1 2
- type(3): scale scale scale
- cTrans(3): (1,1,1) (1,2,2) (1,4,4)
channels(3): 0 1 2

I am a little confused about datasets$coordinatetransformations ... since there is another transformations list in the attributes. Will ask @LucaMarconato ... should we expect an arbitrary list of transformations for each dataset ?

[Artur-man]

Following here: https://ngff.openmicroscopy.org/latest/index.html
I think it is better to just list datasets below and summarize transformation, simply put each dataset has either a scale or translation transformation or both:

> x@images$blobs_multiscale_image@meta
class: Zattrs
axes(3): 
- name(3): c y x
- type(3): channel space space
transformations(1): global
datasets(3): 0 1 2
- 0: (scale:1,1,1) 
- 1: (scale:1,2,2) 
- 2: (scale:1,4,4) 
channels(3): 0 1 2

[LucaMarconato]

@Artur-man there are two sets of transformations because one set contains one transformation for each scale of a multiscale object. These transformations go between the intrinsic coordinate system of the scale to a different coordinate system.

Then there are transformations between different coordinate systems.

Here is a recent repository with many examples from the latest incoming specification (NGFF 0.6-dev1). The examples rendered on the NGFF website are too limited. Over the next weeks you can expect activities both on the repo above and in this repo: https://github.com/ome-zarr-models/ome-zarr-models-py with regard to coordinate transformations.

[Artur-man]

Here is the latest version based on what we discussed with @LucaMarconato. Let me know what you think.

> x@labels$blobs_labels@meta
class: Zattrs
axes(2): 
- name: y x
- type: space space
coordTrans(5):
- global: (identity)
- scale: (scale:[3,2])
- translation: (translation:[-50,10])
- affine: (affine:[[20,10,30],[50,40,60]])
- sequence: (scale:[3,2]), (translation:[-50,10])
datasets(1): 0
- 0: (scale:[1,1]) 

> x@images$blobs_multiscale_image@meta
class: Zattrs
axes(3): 
- name: c y x
- type: channel space space
coordTrans(1):
- global: (identity)
datasets(3): 0 1 2
- 0: (scale:[1,1,1]) 
- 1: (scale:[1,2,2]) 
- 2: (scale:[1,4,4]) 
channels(3): 0 1 2

[LucaMarconato]

Thanks @Artur-man for the update. Looks good to me, but I would add one piece of information: for the transformations not in datasets, I would add the input and output coordinate system names.