Color histogram with floats

[zoccoler]

This adds the possibility to assign an array with floats (like representing some feature) as an overlay to the histogram and to the scatter plot.

• update documentation
• check if more tests are needed
• fix log scale problem for histogram

[codecov]

Codecov Report

Attention: Patch coverage is 91.66667% with 37 lines in your changes missing coverage. Please review.

Project coverage is 90.18%. Comparing base (6290895) to head (979103b).
Report is 57 commits behind head on main.

Files with missing lines	Patch %	Lines
src/biaplotter/artists.py	92.70%	20 Missing ⚠️
src/biaplotter/selectors.py	75.00%	12 Missing ⚠️
src/biaplotter/colormap.py	78.94%	4 Missing ⚠️
src/biaplotter/plotter.py	94.11%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #18      +/-   ##
==========================================
+ Coverage   89.65%   90.18%   +0.52%     
==========================================
  Files           7        8       +1     
  Lines         667      978     +311     
==========================================
+ Hits          598      882     +284     
- Misses         69       96      +27     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[zoccoler]

Hi @jo-mueller ,

I am adding here the possibility to display the scatter colors or an histogram overlay based o another features column (not just a categorical array with integers, but rather any float array with a continuous colormap - I renamed categorical_colormap to overlay_colormap).

You can get a feeling of the changes by running the Artists notebook again, which I updated.

Of course a histogram overlay completely hides the original histogram, but it may be useful if the user hide and show it quickly or change its opacity for example.

When you get some time, could you please give it a try to check if we can merge this?

[jo-mueller]

Hi @zoccoler ,

sorry for the late reply. I am a bit undecided on this implementation: On the one hand I like the simplicity of this: Simply change the colormap to something continuous, pass a float to the color indices and have your plot colored. On the other hand, it loads the same property (color) with potentially multiple meanings.

In the context of the clusters plotter, it would be hard to use colored scatter points in conjunction with the clustering. As an alternative suggestion: Would you think that it could make sense to forward the coloring done with color_indeces to the marker edge color? The coloring of the points according to a given feature could then still be done with the facecolor and the two colorings wouldn't superseed each other.

Let me know what you think!

[jo-mueller]

This is problematic, I think. By applying a color normalization here, there is no longer a clear mapping between color_indices and the actually drawn color in the plot.

Usually, what should work is

color = active_artist.overlay_colormap(color_indices)

and the colors I get should be exactly the same as in the plot. However, the colors in the plot pass through an additional normalization step. To recover the colors from the plot (like needed for the clusters plotter), I need to do this:

        color_indices = active_artist.color_indices
        norm = active_artist._get_normalization(color_indices)
        colors = active_artist._get_rgba_colors(color_indices, norm)

Maybe there could be an additional convenience method to wrap the normalization and to_rgba step? Or the normalization step could go into the __call__ method of the new BiaColormap?

[jo-mueller]

Hi @zoccoler , besides the above points, another problem for me is that with the color_indeces now also carrying continuous values, I cannot distinguish anymore between these two cases:

• A cluster was drawn on the plotter canvas and the cluster indeces are emitted
• Some continuous values were passed to the plotter and the values are subsequently emitted

In the scope of the clusters-plotter, this distinction is important though. Consider this workflow:

• I plot some continuous feature in the clusters-plotter
• When I start drawing and finish, the clusters should reset so that the MANUAL_CLUSTER_ID is plotted again.
• In this case, the hue selector in the widget should jump to MANUAL_CLUSTER_ID as well.
• I have no way of knowing to which value to set the hue selector from the cluster_indeces.

[zoccoler]

Hi @zoccoler , besides the above points, another problem for me is that with the color_indeces now also carrying continuous values, I cannot distinguish anymore between these two cases:

• A cluster was drawn on the plotter canvas and the cluster indeces are emitted
• Some continuous values were passed to the plotter and the values are subsequently emitted

In the scope of the clusters-plotter, this distinction is important though. Consider this workflow:

• I plot some continuous feature in the clusters-plotter
• When I start drawing and finish, the clusters should reset so that the MANUAL_CLUSTER_ID is plotted again.
• In this case, the hue selector in the widget should jump to MANUAL_CLUSTER_ID as well.
• I have no way of knowing to which value to set the hue selector from the cluster_indeces.

That is why I had placed this 😅 Does it help to solve that problem?

[jo-mueller]

That should help :)

[jo-mueller]

@zoccoler , just tested it, the new signal works on my end for clusters-plotter purposes. Can you have a look at PR #42 so that this PR can go through?

[zoccoler]

Hi @jo-mueller ,

This is now merged. I will make the new release early next week because I have to prepare a fix release for napari-phasors due to the breaking changes introduced here.

Please refer to the main branch for now while working in the clusters-plotter.