W.r.t. the test, the closest PR I can find didn't add a test, and looking through the existing tests, I'm not sure how we would add a test for this.

Maybe @martinRenou or @maartenbreddels can you point us to a PR that did something similar and had a test for it?

The process for producing the error is here, but it requires manual brushing.

Since recently we added galata tests in bqplot. Galata should have everything for "manually" testing the brush and checking that it renders properly. But that requires good knowledge of galata which I don't have...

Can this be bug be triggered programmatically from Python? That should help create a visual test. I can look into it.

Thanks @martinRenou, it looks like that should work. However, in investigating programmatically setting the brush selected_y values another concern popped up.

This PR originally introduces using the domain index for the values in the brush's selected traits which is different than the existing behavior which uses the domain values itself. It might be more consistent to deal with the indices in the domain, however to minimize changes to other users it looks like the:

  typedRange(values) {
    return new Int32Array(values.map(this.domain.indexOf, this.domain));
  }

could be replaced with:

  typedRange(values) {
    return values;
  }

There is also a new galata test notebook ordinal_scale.ipynb which tests two different figures. One figure uses the ordinal scale with strings while the other uses the ordinal scale with integers. The test notebook also programatically sets the brush's selected values and captures the output screenshots.

Since recently we added galata tests in bqplot. Galata should have everything for "manually" testing the brush and checking that it renders properly. But that requires good knowledge of galata which I don't have...

Can this be bug be triggered programmatically from Python? That should help create a visual test. I can look into it.

@martinRenou : @dfreeman06 added a test using Galata that we believe is working and testing the brushes programmatically, at least locally it is passing, would it be possible to run the workflow to see if everything is on the up and up?

Just a friendly ping to the BQPlot team to see if there is something else we should do for this PR. Also curious if @maartenbreddels has any issues with this as per his PR.

y'all let me know if there's anything i can do to push this forward...

Thanks for adding tests. Ideally, we would add a galata test that simulates a selection with the mouse and test that the selected_y attribute gets changed properly. But I don't know yet how to do that, so let's merge and see later!

Thanks!

Everybody's real busy, for sure, but Is there a timeline for a .30 release? Anything we can do to help?