More Mapper tests, small change to `max_fraction` in FirstSimpleGap and FirstHistogramGap

[wreise]

Reference issues/PRs None

Types of changes

• [ ] Bug fix (non-breaking change which fixes an issue)
• [ ] New feature (non-breaking change which adds functionality)
• [x] Breaking change (fix or feature that would cause existing functionality to change)

Description

• Adds test for the clone_pipeline parameter in gtda.mapper.visualization.plot_interactive_mapper_graph. The test currently fails.
• Makes a small change to the meaning of the max_fraction parameter in FirstSimpleGap and FirstHistogramGap: instead of the maximum number of clusters being max_fraction * (n_samples - 1), it is now more simply the floor of max_fraction * n_samples. Furthermore, the default value is changed from None to 1..

Checklist

• [x] I have read the guidelines for contributing.
• [x] My code follows the code style of this project. I used flake8 to check my Python changes.
• [x] My change requires a change to the documentation.
• [x] I have updated the documentation accordingly.
• [x] I have added tests to cover my changes.
• [x] All new and existing tests passed. I used pytest to check this on Python tests.

[ulupo]

@wreise I was able to put together a version which "works". I have not got to the bottom of the issue but it seems that we have bad conflicts/side effects between tests. It's as if the fig or pipe objects were changed simultaneously by different tests, causing the failures (but again, not sure). Placing the tests for plot_interactive_mapper_graph in a new file seems to have "solved" this. (?!)

[ulupo]

@wreise @lewtun the mystery is solved! The point was that ipywidgets somehow links together all widgets created in the same session, so that even after the Python object fig has seemingly been overwritten, fig.widgets still contains data from previously created widgets. This caused a problem when using @wreise's _get_widget_from_trait because this function could fetch the widget belonging to a previously created figure, instead of the correct one. This meant that fig did not actually observe a change, despite us thinking we caused one!

My fix involves turning _get_widget_from_trait into _get_widgets_from_trait which now returns a list of all possible matches. In the test code, all possible matching widgets are changed. It's not elegant (because ideally we would only change what's relevant), but it works!

@wreise I am hoping I can leave this PR for you to finish off, if you think you would be able to. As is, it already boosts code coverage substantially, but there are a few simple things which can be added (e.g. use of pandas dataframes, and maybe passing color_by_columns_dropdown=True). The coverage report (pytest gtda/mapper --cov --cov-report xml) is a good guide.

[wreise]

Thank you for your thorough investigation, @ulupo ! I should be able to finish this.

[ulupo]

@wreise thanks! And thanks for opening this a long time ago, even if we didn't know how to do it then. It served as a good reminder.