Update Bokeh

[leehart]

Re: issue #845

[leehart]

About 100 linting errors for Bokeh to fix, over 9 files: https://github.com/malariagen/malariagen-data-python/actions/runs/21069645421/job/60595592158?pr=846

[leehart]

Down to 57 linting errors to fix for Bokeh, over 8 files: https://github.com/malariagen/malariagen-data-python/actions/runs/21073854626/job/60610492236?pr=846

[leehart]

Down to 38 errors in 8 files, mostly around a few obsolete keyword args in Bokeh's figure, namely: active_scroll, active_drag, active_inspect.

[leehart]

Since Bokeh docs & source code for v3.8.2 appear to still support the keyword args (active_scroll, active_drag, active_inspect, x_axis_type) for Bokeh's figure, and I didn't find an alternative, it seems safe to explicitly ignore these type checks with a note in the code.

• https://docs.bokeh.org/en/3.8.2/docs/reference/plotting/figure.html

However, this might need further checking and revisiting later.

There is still one linting error to resolve, namely Unexpected keyword argument "color_mapper" for "image" of "GlyphAPI" in anoph/cnv_data.py. There also appear to be some bokeh-related errors in the Jupyter Notebooks, which will require investigation.

[leehart]

Linting tests now pass: https://github.com/malariagen/malariagen-data-python/actions/runs/21182236081/job/60927279425?pr=846

Now working on the failing Notebook tests, which appear to be caused by the linting amends for the Bokeh upgrade, so some linting amends might need tweaking: https://github.com/malariagen/malariagen-data-python/actions/runs/21182236110/job/60927279285?pr=846

Unfortunately, other tests are still failing due to the sporadic segmentation fault, issue #835: https://github.com/malariagen/malariagen-data-python/actions/runs/21182236064/job/60927279399?pr=846

[leehart]

The Notebook tests now pass: https://github.com/malariagen/malariagen-data-python/actions/runs/21243730389/job/61127633039?pr=846

As do the linting and the integration tests:

• https://github.com/malariagen/malariagen-data-python/actions/runs/21243730439/job/61127633052?pr=846
• https://github.com/malariagen/malariagen-data-python/actions/runs/21243730457/job/61127633073?pr=846

However, the unit and coverage tests appear to fail (36 passed, 865 errors), when not halted by the segfault issue, even though tests all pass locally (873 passed, 310 warnings):

• https://github.com/malariagen/malariagen-data-python/actions/runs/21243730382/job/61127633080?pr=846
• https://github.com/malariagen/malariagen-data-python/actions/runs/21243730416/job/61127633041?pr=846

I will try to investigate those test failures, but we're obviously still going to be hampered by the sporadic segfault, so I'll return to trying to help investigate that (#835) soon. I was rather hoping that rejigging the package dependencies might magically resolve it, since we're currently out of ideas.

[leehart]

The tests/integration tests still pass locally (901 passed, 328 warnings) using Python 3.12.5.

Noting here that the change from tests.yml and coverage.yml running the tests under tests/anoph to testing everything under tests except tests/integration appears to date back to PR #686 from 06 Dec 2024, which relates to the open issue #682. That was when legacy_tests.yml was renamed integration_tests.yml, which exclusively runs tests under tests/integration.

• https://github.com/malariagen/malariagen-data-python/pull/686/files