18499875864

Committed 14 Oct 2025 02:26PM UTC coverage: 92.799% (-0.7%) from 93.522%

Build # 18499875864

Build Type

push

github

Committed by

web-flow

Commit Message

Enhance type safety and fix bugs across the codebase (#430)

* First Pyright cleanup

* TypeChecked game

* fixed introduced bugs in game and interaction_values

* Pyright Save Sampling

* TypeSafe Approximator

* Typechecked Datasets

* Explainer folder typechecked

* GameTheory Typechecked

* Imputer Typechecked

* Plot Typechecked

* Added static typechecking to pre-commit

* Refactoring

* Add pyright change to CHANGELOG

* Activate code quality show diff

* changed uv sync in pre-commit hook

* made fixtures local import

* Introduced Generic TypeVar in Approximator, reducing ignores

* Introduced Generic Types for Explainer. Approximator, Imputer and ExactComputer can either exist or not, depending on dynamic Type

* Bug fix caused through refactoring

* updated overrides

* tightened CoalitionMatrix to accept only bool arrays

* Remove Python reinstallation step in CI workflow

Removed the step to reinstall Python on Windows due to issues with tkinter. The linked GitHub issue was solved. Doing this as a first try.

* Add Python reinstallation and Tkinter installation steps

Reinstall Python and install Tkinter for Windows tests. prior commit did not help

* Fix command for installing Tkinter in workflow

* Update Windows workflow to install Tkinter via Chocolatey

* Remove Tkinter installation step from Windows workflow and adjust matplotlib usage for headless environments

* adapted some pyright types

* removed generics from explainer again

* tightened index type check

* made n_players None at assignment again

* moved comments

---------

Co-authored-by: Maximilian <maximilian.muschalik@gmail.com>

Run Details

304 of 360 new or added lines in 51 files covered. (84.44%)

12 existing lines in 9 files now uncovered.

4987 of 5374 relevant lines covered (92.8%)

0.93 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

86.57

/src/shapiq/plot/upset.py

"""This module contains the upset plot."""

from __future__ import annotations

from typing import TYPE_CHECKING

import matplotlib.pyplot as plt

from ._config import BLUE, RED

if TYPE_CHECKING:
    from collections.abc import Sequence

    from matplotlib.figure import Figure

    from shapiq.interaction_values import InteractionValues


def upset_plot(
    interaction_values: InteractionValues,
    *,
    n_interactions: int = 20,
    feature_names: Sequence[str] | None = None,
    color_matrix: bool = False,
    all_features: bool = True,
    figsize: tuple[float, float] | None = None,
    show: bool = False,
) -> Figure | None:
    """Plots the upset plot.

    UpSet plots[1]_ can be used to visualize the interactions between features. The plot consists of
    two parts: the upper part shows the interaction values as bars, and the lower part shows the
    interactions as a matrix. Originally, the UpSet plot was introduced by Lex et al. (2014)[1]_.
    For a more detailed explanation about the plots, see the references or the original
    [documentation](https://upset.app/).

    An example of this plot is shown below.

    .. image:: /_static/images/upset_plot.png
        :width: 600
        :align: center

    Args:
        interaction_values: The interaction values as an ``InteractionValues`` object.
        feature_names: The names of the features. Defaults to ``None``. If ``None``, the features
            will be named with their index.
        n_interactions: The number of top interactions to plot. Defaults to ``20``. Note this number
            is completely arbitrary and can be adjusted to the user's needs.
        color_matrix: Whether to color the matrix (red for positive values, blue for negative) or
            not (black). Defaults to ``False``.
        all_features: Whether to plot all ``n_players`` features or only the features that are
            present in the top interactions. Defaults to ``True``.
        figsize: The size of the figure. Defaults to ``None``. If ``None``, the size will be set
            automatically depending on the number of features.
        show: Whether to show the plot. Defaults to ``False``.

    Returns:
        If ``show`` is ``True``, the function returns ``None``. Otherwise, it returns a tuple with
        the figure and the axis of the plot.

    References:
        .. [1] Alexander Lex, Nils Gehlenborg, Hendrik Strobelt, Romain Vuillemot, Hanspeter Pfister. UpSet: Visualization of Intersecting Sets IEEE Transactions on Visualization and Computer Graphics (InfoVis), 20(12): 1983--1992, doi:10.1109/TVCG.2014.2346248, 2014.

    """
    # prepare data ---------------------------------------------------------------------------------
    values = interaction_values.values
    values_ids: dict[int, tuple[int, ...]] = {
        v: k for k, v in interaction_values.interaction_lookup.items()
    }
    values_abs = abs(values)
    idx = values_abs.argsort()[::-1]
    idx = idx[:n_interactions] if n_interactions > 0 else idx
    values = values[idx]
    interactions: list[tuple[int, ...]] = [values_ids[i] for i in idx]

    # prepare feature names ------------------------------------------------------------------------
    if all_features:
        features = set(range(interaction_values.n_players))
    else:
        features = {feature for interaction in interactions for feature in interaction}
    n_features = len(features)
    feature_pos = {feature: n_features - 1 - i for i, feature in enumerate(features)}
    if feature_names is None:
        feature_names = [f"Feature {feature}" for feature in features]
    else:
        feature_names = [feature_names[feature] for feature in features]

    # create figure --------------------------------------------------------------------------------
    height_upper, height_lower = 5, n_features * 0.75
    height = height_upper + height_lower
    ratio = [height_upper, height_lower]
    if figsize is None:
        figsize = (10, height)
    else:
        if figsize[1] is None:
            figsize = (figsize[0], height)
        if figsize[0] is None:
            figsize = (10, figsize[1])

    fig, ax = plt.subplots(2, 1, figsize=figsize, gridspec_kw={"height_ratios": ratio}, sharex=True)

    # plot lower part of the upset plot
    for x_pos, interaction in enumerate(interactions):
        color = RED.hex if values[x_pos] >= 0 else BLUE.hex

        # plot upper part
        bar = ax[0].bar(x_pos, values[x_pos], color=color)
        label = [f"{values[x_pos]:.2f}"]
        ax[0].bar_label(bar, label, label_type="edge", color="black", fontsize=12, padding=3)

        # plot lower part
        # plot the matrix in the background
        ax[1].plot(
            [x_pos for _ in range(n_features)],
            list(range(n_features)),
            color="lightgray",
            marker="o",
            markersize=15,
            linewidth=0,
        )
        # add the interaction to the matrix
        y_pos = [feature_pos[feature] for feature in interaction]
        ax[1].plot(
            [x_pos for _ in range(len(interaction))],
            y_pos,
            color="black" if not color_matrix else color,
            marker="o",
            markersize=15,
            linewidth=1.5,
        )

    # beautify upper plot --------------------------------------------------------------------------
    min_max = (min(values), max(values))
    delta = (min_max[1] - min_max[0]) * 0.1
    ax[0].set_ylim(min_max[0] - delta, min_max[1] + delta)
    ax[0].set_ylabel("Interaction Value")
    ax[0].spines["top"].set_visible(False)
    ax[0].spines["right"].set_visible(False)
    ax[0].spines["bottom"].set_visible(False)
    ax[0].axhline(0, color="black", linewidth=0.5)  # add line at 0

    # beautify lower plot --------------------------------------------------------------------------
    ax[1].set_ylim(-1, n_features)
    ax[1].yaxis.set_ticks(range(n_features))
    ax[1].set_yticklabels(reversed(feature_names))
    ax[1].tick_params(axis="y", length=0)  # remove y-ticks
    ax[1].set_xticks([])  # remove x-axis
    ax[1].spines["top"].set_visible(False)
    ax[1].spines["right"].set_visible(False)
    ax[1].spines["bottom"].set_visible(False)
    ax[1].spines["left"].set_visible(False)
    # background shading
    for i in range(n_features):
        if i % 2 == 0:
            ax[1].axhspan(i - 0.5, i + 0.5, color="lightgray", alpha=0.25, zorder=0, lw=0)

    # adjust whitespace
    plt.subplots_adjust(hspace=0.0)

    if not show:
        return fig
    plt.show()
    return None

1	"""This module contains the upset plot."""
2
3	from __future__ import annotations	1✔
4
5	from typing import TYPE_CHECKING	1✔
6
7	import matplotlib.pyplot as plt	1✔
8
9	from ._config import BLUE, RED	1✔
10
11	if TYPE_CHECKING:	1✔
12	from collections.abc import Sequence	×
13
NEW 14	from matplotlib.figure import Figure	×
15
UNCOV 16	from shapiq.interaction_values import InteractionValues	×
17
18
19	def upset_plot(	1✔
20	interaction_values: InteractionValues,
21	*,
22	n_interactions: int = 20,
23	feature_names: Sequence[str] \| None = None,
24	color_matrix: bool = False,
25	all_features: bool = True,
26	figsize: tuple[float, float] \| None = None,
27	show: bool = False,
28	) -> Figure \| None:
29	"""Plots the upset plot.
30
31	UpSet plots[1]_ can be used to visualize the interactions between features. The plot consists of
32	two parts: the upper part shows the interaction values as bars, and the lower part shows the
33	interactions as a matrix. Originally, the UpSet plot was introduced by Lex et al. (2014)[1]_.
34	For a more detailed explanation about the plots, see the references or the original
35	[documentation](https://upset.app/).
36
37	An example of this plot is shown below.
38
39	.. image:: /_static/images/upset_plot.png
40	:width: 600
41	:align: center
42
43	Args:
44	interaction_values: The interaction values as an ``InteractionValues`` object.
45	feature_names: The names of the features. Defaults to ``None``. If ``None``, the features
46	will be named with their index.
47	n_interactions: The number of top interactions to plot. Defaults to ``20``. Note this number
48	is completely arbitrary and can be adjusted to the user's needs.
49	color_matrix: Whether to color the matrix (red for positive values, blue for negative) or
50	not (black). Defaults to ``False``.
51	all_features: Whether to plot all ``n_players`` features or only the features that are
52	present in the top interactions. Defaults to ``True``.
53	figsize: The size of the figure. Defaults to ``None``. If ``None``, the size will be set
54	automatically depending on the number of features.
55	show: Whether to show the plot. Defaults to ``False``.
56
57	Returns:
58	If ``show`` is ``True``, the function returns ``None``. Otherwise, it returns a tuple with
59	the figure and the axis of the plot.
60
61	References:
62	.. [1] Alexander Lex, Nils Gehlenborg, Hendrik Strobelt, Romain Vuillemot, Hanspeter Pfister. UpSet: Visualization of Intersecting Sets IEEE Transactions on Visualization and Computer Graphics (InfoVis), 20(12): 1983--1992, doi:10.1109/TVCG.2014.2346248, 2014.
63
64	"""
65	# prepare data ---------------------------------------------------------------------------------
66	values = interaction_values.values	1✔
67	values_ids: dict[int, tuple[int, ...]] = {	1✔
68	v: k for k, v in interaction_values.interaction_lookup.items()
69	}
70	values_abs = abs(values)	1✔
71	idx = values_abs.argsort()[::-1]	1✔
72	idx = idx[:n_interactions] if n_interactions > 0 else idx	1✔
73	values = values[idx]	1✔
74	interactions: list[tuple[int, ...]] = [values_ids[i] for i in idx]	1✔
75
76	# prepare feature names ------------------------------------------------------------------------
77	if all_features:	1✔
78	features = set(range(interaction_values.n_players))	1✔
79	else:
80	features = {feature for interaction in interactions for feature in interaction}	1✔
81	n_features = len(features)	1✔
82	feature_pos = {feature: n_features - 1 - i for i, feature in enumerate(features)}	1✔
83	if feature_names is None:	1✔
84	feature_names = [f"Feature {feature}" for feature in features]	1✔
85	else:
86	feature_names = [feature_names[feature] for feature in features]	1✔
87
88	# create figure --------------------------------------------------------------------------------
89	height_upper, height_lower = 5, n_features * 0.75	1✔
90	height = height_upper + height_lower	1✔
91	ratio = [height_upper, height_lower]	1✔
92	if figsize is None:	1✔
93	figsize = (10, height)	1✔
94	else:
95	if figsize[1] is None:	×
96	figsize = (figsize[0], height)	×
97	if figsize[0] is None:	×
98	figsize = (10, figsize[1])	×
99
100	fig, ax = plt.subplots(2, 1, figsize=figsize, gridspec_kw={"height_ratios": ratio}, sharex=True)	1✔
101
102	# plot lower part of the upset plot
103	for x_pos, interaction in enumerate(interactions):	1✔
104	color = RED.hex if values[x_pos] >= 0 else BLUE.hex	1✔
105
106	# plot upper part
107	bar = ax[0].bar(x_pos, values[x_pos], color=color)	1✔
108	label = [f"{values[x_pos]:.2f}"]	1✔
109	ax[0].bar_label(bar, label, label_type="edge", color="black", fontsize=12, padding=3)	1✔
110
111	# plot lower part
112	# plot the matrix in the background
113	ax[1].plot(	1✔
114	[x_pos for _ in range(n_features)],
115	list(range(n_features)),
116	color="lightgray",
117	marker="o",
118	markersize=15,
119	linewidth=0,
120	)
121	# add the interaction to the matrix
122	y_pos = [feature_pos[feature] for feature in interaction]	1✔
123	ax[1].plot(	1✔
124	[x_pos for _ in range(len(interaction))],
125	y_pos,
126	color="black" if not color_matrix else color,
127	marker="o",
128	markersize=15,
129	linewidth=1.5,
130	)
131
132	# beautify upper plot --------------------------------------------------------------------------
133	min_max = (min(values), max(values))	1✔
134	delta = (min_max[1] - min_max[0]) * 0.1	1✔
135	ax[0].set_ylim(min_max[0] - delta, min_max[1] + delta)	1✔
136	ax[0].set_ylabel("Interaction Value")	1✔
137	ax[0].spines["top"].set_visible(False)	1✔
138	ax[0].spines["right"].set_visible(False)	1✔
139	ax[0].spines["bottom"].set_visible(False)	1✔
140	ax[0].axhline(0, color="black", linewidth=0.5) # add line at 0	1✔
141
142	# beautify lower plot --------------------------------------------------------------------------
143	ax[1].set_ylim(-1, n_features)	1✔
144	ax[1].yaxis.set_ticks(range(n_features))	1✔
145	ax[1].set_yticklabels(reversed(feature_names))	1✔
146	ax[1].tick_params(axis="y", length=0) # remove y-ticks	1✔
147	ax[1].set_xticks([]) # remove x-axis	1✔
148	ax[1].spines["top"].set_visible(False)	1✔
149	ax[1].spines["right"].set_visible(False)	1✔
150	ax[1].spines["bottom"].set_visible(False)	1✔
151	ax[1].spines["left"].set_visible(False)	1✔
152	# background shading
153	for i in range(n_features):	1✔
154	if i % 2 == 0:	1✔
155	ax[1].axhspan(i - 0.5, i + 0.5, color="lightgray", alpha=0.25, zorder=0, lw=0)	1✔
156
157	# adjust whitespace
158	plt.subplots_adjust(hspace=0.0)	1✔
159
160	if not show:	1✔
161	return fig	1✔
162	plt.show()	×
163	return None	×

mmschlk / shapiq / 18499875864

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous