Changelog¶

What’s new in version 0.9¶

Fixes a bug where show_percentages used the incorrect denominator if filtering (e.g. min_subset_size) was applied. This bug was a regression introduced in version 0.7. (#248)
Align ylabels of subplots added using add_catplot. (#266)
Add a style_categories method to customize category plot styles, including shading of rows in the intersection matrix, and bars in the totals plot. (#261 with thanks to Marcel Albus).
Ability to disable totals plot with totals_plot_elements=0. (#246)
Ability to set totals y axis label (#243)
Added max_subset_rank to get only n most populous subsets. (#253)
Added support for min_subset_size and max_subset_size specified as percentage. (#264)

Allowed show_percentages to be provided with a custom formatting string, for example to show more decimal places. (#194)
Added include_empty_subsets to UpSet and query to allow the display of all possible subsets. (#185)
sort_by and sort_categories_by now accept ‘-’ prefix to their values to sort in reverse. ‘input’ and ‘-input’ are also supported. (#180)
Added subsets attribute to QueryResult. (#198)
Fixed a bug where more than 64 categories could result in an error. (#193)

Patch release 0.8.2 handles deprecations in dependencies.

Added add_stacked_bars, similar to add_catplot but to add stacked bar charts to show discrete variable distributions within each subset. (#137)
Improved ability to control colors, and added a new example of same. Parameters other_dots_color and shading_color were added. facecolor will now default to white if matplotlib.rcParams['axes.facecolor'] is dark. (#138)
Added style_subsets to colour intersection size bars and matrix dots in the plot according to a specified query. (#152)
Added from_indicators to allow yet another data input format. This allows category membership to be easily derived from a DataFrame, such as when plotting missing values in the columns of a DataFrame. (#143)

Support using input intersection order with sort_by=None (#133 with thanks to Brandon B).
Add parameters for filtering by subset size (with thanks to Sichong Peng) and degree. (#134)
Fixed an issue where tick labels were not given enough space and overlapped category totals. (#132)
Fixed an issue where our implementation of sort_by='degree' apparently gave incorrect results for some inputs and versions of Pandas. (#134)

Fixed issue with the order of catplots being reversed for vertical plots (#122 with thanks to Enrique Fernandez-Blanco)
Fixed issue with the x limits of vertical plots (#121).

Fixed large x-axis plot margins with high number of unique intersections (#106 with thanks to Yidi Huang)

Added option to display both the absolute frequency and the percentage of the total for each intersection and category. (#89 with thanks to Carlos Melus and Aaron Rosenfeld)
Improved efficiency where there are many categories, but valid combinations are sparse, if sort_by='degree'. (#82)
Permit truthy (not necessarily bool) values in index. (#74 with thanks to @ZaxR)
intersection_plot_elements can now be set to 0 to hide the intersection size plot when add_catplot is used. (#80)

Added from_contents to provide an alternative, intuitive way of specifying category membership of elements.
To improve code legibility and intuitiveness, sum_over=False was deprecated and a subset_size parameter was added. It will have better default handling of DataFrames after a short deprecation period.
generate_data has been replaced with generate_counts and generate_samples.
Fixed the display of the “intersection size” label on plots, which had been missing.
Trying to improve nomenclature, upsetplot now avoids “set” to refer to the top-level sets, which are now to be known as “categories”. This matches the intuition that categories are named, logical groupings, as opposed to “subsets”. To this end:
- generate_counts (formerly generate_data) now names its categories “cat1”, “cat2” etc. rather than “set1”, “set2”, etc.
- the sort_sets_by parameter has been renamed to sort_categories_by and will be removed in version 0.4.

Return a Series (not a DataFrame) from from_memberships if data is 1-dimensional.

Added from_memberships to allow a more convenient data input format.
plot and UpSet now accept a pandas.DataFrame as input, if the sum_over parameter is also given.
Added an add_catplot method to UpSet which adds Seaborn plots of set intersection data to show more than just set size or total.
Shading of subset matrix is continued through to totals.
Added a show_counts option to show counts at the ends of bar plots. (#5)
Defined _repr_html_ so that an UpSet object will render in Jupyter notebooks. (#36)
Fix a bug where an error was raised if an input set was empty.