ome_model: channel fixes

[sbesson]

This PR inclues a set of improvements for the Image/Pixels/Channel created by the ome_model library driven by in-progress IDR work on HCS RGB companion files /cc @dominikl

Changes

• the Channel.Name and Channel.Color attributes are optional as per the OME schema. In the case of many RGB images typically these values are irrelevant. 7837240 changes the Channel constructor and Image.add_channel method to turn these metadata fields into optional key/value pairs rather than mandatory arguments and handling missing arguments/None

• 1ae8b58 refines the current Image.validate() logic to account for the optional nature of the Channel element and the SamplesPerPixel value by using the two following checks:

  1. the number of Channel elements is less or equal to SizeC
  2. the sum of the SamplesPerPixels for each Channel is less of equal to SizeC

Testing

Without this PR the following examples should fail during the channel validation step. With this PR, it should create a valid OME-XML:

from ome_model.experimental import Image, create_companion
image = Image("test", 512, 512, 1, 3, 1)
image.add_tiff("test.tiff")
create_companion(images=[image])

from ome_model.experimental import Image, create_companion
image = Image("test", 512, 512, 1, 3, 1)
image.add_channel(samplesPerPixel=3)
image.add_tiff("test.tiff")
create_companion(images=[image])

In addition to manual testing, commits f22547e and following add Python unit tests covering the constructors of the Image and Channel classes as well as a tox infrastructure used byTravis to run these Python tests in a separate job.

Target: 6.2.0

[dominikl]

Looks good, but with your second testing example I'll get:

AssertionError: {'Image': {'ID': 'Image:0', 'Name': 'test'}, 'Pixels': {'ID': 'Pixels:0:0', 'DimensionOrder': 'XYZTC', 'Type': 'uint16', 'SizeX': '512', 'SizeY': '512', 'SizeZ': '1', 'SizeT': '1', 'SizeC': '3'}, 'Channels': [<ome_model.experimental.Channel object at 0x7fe4463cd400>], 'TIFFs': [<ome_model.experimental.TiffData object at 0x7fe44559ab38>]}

And I can't find where/if the tests are running in the travis output, maybe I'm just missing what to look for?

[sbesson]

Edited the snippet in the PR description, the default dimension order is XYZTC and my example had an image with SizeC=1 and one channel with SamplesPerPixels=3. So the AssertionError was expected (although the error message could be improved).

The PR changes had a third job to the Travis build where the Python library is installed and tests are executed - see https://travis-ci.org/github/ome/ome-model/jobs/718428821. Locally, this can be tested by running tox -e py36 or tox -e py37 depending on your environment.

[dominikl]

I think it was correct, because Image("test", 512, 512, 1, 1, 3) now gets interpreted as t=3, doesn't it?

[sbesson]

Yes, tou are correct. The problem was that my channel samples comparison was too stringent (< while it should have been <=) and my test was not checking the case of 1 RGB channel only. I have adjusted the test accordingly in https://github.com/ome/ome-model/pull/120/commits/2905ea9a01fed4f3dd47207822b183e304c27fdb so that it failed as expected by your testing and fixed the validation in https://github.com/ome/ome-model/pull/120/commits/49d212302bd30541e55ef64409b14377b069cd3a.

[joshmoore]

Happy with a general code review. Also happy to open issue for "add ome_model version to output". I do wonder (similar to my recent ponderings around ome-zarr-py) if we don't want to hook this repository up to a MxN matrix of file producers/consumers, something like:

| V producer \ consumer >   | bio-formats                       | omero             | ome-zarr-py                       | omero-ms-zarr     |
|-------------------------  |---------------------------------  |----------------   |--------------------------------   |---------------    |
| ome-model-py (TIFF)       | Easy                              | Easy              | with bf2raw (needs ZarrWriter)    | N/A (yet)         |
| bio-formats (TIFF)        | Easy                              | Easy              | "                                 | "                 |
| omero (TIFF Export)       | Easy                              | Easy              |                                   | "                 |
| omero-cli-zarr (ZARR)     | with raw2ometiff and ZarrReader   | as bio-formats    | Easy                              | "                 |
| omero-ms-zarr (ZARR)      | "                                 | "                 | Easy                              | "                 |
|                           |                                   |                   |                                   |                   |

[sbesson]

Thanks. Merging and I'll open 2 follow-up PRs: one for plane metadata and one for the version output.

Regarding compatibility, the closest official representation is the paragraph about tooling in https://docs.openmicroscopy.org/ome-model/6.1.2/. I could certainly see this turning into a producer/consumer or maybe read/write support matrix that we would keep updating as the community tooling grows.

I would assume we would need to discuss representation as I could imagine one table per container format ie. TIFF/ZARR.