rewrite tests/unit/image/test_image.py to use "truth" images rather than hardcoded values

[dmehring]

A more straightforward, lower maintenance way to write these tests is to store truth images and compare the "got" xds against the relevant truth image using a generic xds comparison alogorithm.

[dmehring]

Opened casangi/toolviper#40 for upload of truth images so they can be downloaded by tests.

[dmehring]

Hi @smcastro

I know you're waiting on this. I'm essentially done. I'm just waiting on the truth images to be placed on the data server so I can add the final download commands and test. But if you'd like, I think it's ok for you to start working using the current state of this branch as a starting point. You can just put the truth files in your directory so the tests can find them. You can find those on cv lustre in /lustre/cv/users/dmehring/zarr_for_tests. When the truth images are in place on the server, I'll make the small remaining modifications on this branch and you can merge that into yours if you create another. Or as far as I'm concerned you can just work on this branch and I'll pull any changes you make and add the small remaining changes once the truth images are in place. I'll leave it to you and @Jan-Willem to decide how to proceed with your work.

[smcastro]

Hi @dmehring ,

Thanks for the heads up. I had originally made a plan to refactor test_image.py to make it easier to maintain, to clearly map the functions from image.py to the test cases/classes and also to refactor the helper functions and fixtures and move them to the xradio.testing module, which you have already started to do in this branch. I have a question regarding the scope of the functions you moved to testing/assertions.py. Are those general assert functions that apply to any type of xarray dataset? Or are they specific for image types? I had originally planned to move all the compare and assert functions to under testing/image/assertions.py, but if the ones you moved are general they should indeed stay in the top testing directory as you have done. Below is part of the plan I had made for the refactoring, for your information.

It makes sense that I wait for this branch to merge and incorporate your changes into a revised refactoring plan. I am not in a hurry, and can wait for you to finish your work.

Refactoring Strategy

Phase 1: Create xradio.testing.image Module

Following the pattern in xradio.testing.measurement_set/, create:

src/xradio/testing/image/
├── __init__.py           # Export public functions
├── assertions.py         # Comparison/assertion helpers
└── fixtures.py           # Test data generation utilities

Helper Methods to Extract (from ImageBase and test classes)

These are comparison/assertion methods that will become standalone functions in assertions.py:

Old Method	Line	Source Class	New Function Name
dict_equality()	95	ImageBase	assert_dict_equality()
compare_sky_mask()	453	xds_from_image_test	assert_sky_mask_equal()
compare_time()	503	xds_from_image_test	assert_time_equal()
compare_polarization()	532	xds_from_image_test	assert_polarization_equal()
compare_frequency()	538	xds_from_image_test	assert_frequency_equal()
compare_vel_axis()	576	xds_from_image_test	assert_velocity_equal()
compare_l_m()	620	xds_from_image_test	assert_l_m_equal()
compare_ra_dec()	656	xds_from_image_test	assert_ra_dec_equal()
compare_beam_params_label()	686	xds_from_image_test	assert_beam_params_label_equal()
compare_sky_attrs()	693	xds_from_image_test	assert_sky_attrs_equal()
compare_xds_attrs()	719	xds_from_image_test	assert_xds_attrs_equal()
compare_image_block()	726	xds_from_image_test	assert_image_block_equal()
compare_uv()	872	xds_from_image_test	assert_uv_equal()
run_time_tests()	1679	make_empty_image_tests	assert_time_coord_valid()
run_polarization_tests()	1687	make_empty_image_tests	assert_polarization_coord_valid()
run_frequency_tests()	1693	make_empty_image_tests	assert_frequency_coord_valid()
run_velocity_tests()	1731	make_empty_image_tests	assert_velocity_coord_valid()
run_l_m_tests()	1739	make_empty_image_tests	assert_l_m_coords_valid()
run_right_ascension_tests()	1772	make_empty_image_tests	assert_right_ascension_coord_valid()
run_declination_tests()	1900	make_empty_image_tests	assert_declination_coord_valid()
run_u_v_tests()	2058	make_empty_image_tests	assert_u_v_coords_valid()
run_attrs_tests()	2082	make_empty_image_tests	assert_image_attrs_valid()

Helper Function Signatures

All assertion helpers should:

• Take xds (or similar) as first parameter
• Take expected values as second parameter
• Use descriptive error messages
• Handle edge cases (FITS vs CASA differences)

Naming Convention

• Assertion functions: assert_*_equal() pattern
• Fixture functions: descriptive names like create_test_image()
• Test classes: Test<FunctionName> pattern
• Nested test classes: Test<FunctionName><Format> for format-specific tests

[dmehring]

Hi @smcastro these functions are not image specific and can be used for comparing any two xr.Datasets. In particular, the public API method assert_xarray_datasets_equal will check all objects contained in the two datasets for equality (dims, coords, data vars, attrs, etc). eg it can be used for measurement set related xr.Datasets.