python - Matplotlib: how to specify width of x-label bounding box

Question

I'm trying to create a stacked bar chart in MatPlotLib with two distinct x-labels on top and bottom. The upper one is supposed to have a bounding box with the same width as the bars themselves.

Plot that's not quite right

This is how I create the labels:

plt.tick_params(axis="both", left=False, bottom=False, labelleft=False)
plt.xticks(ind, diagram.keys())
ax.set_frame_on(False)

for label, x in zip([q[1] for q in diagram.values()], ind):
    ax.text(
        x, 1.05, '{:4.0%}'.format(label), 
        ha="center", va="center", 
        bbox={"facecolor": "blue", "pad": 3}
    )

diagram is a dictionary like {bottom-label: [[contents], top-label]}

So I guess my question boils down to: How do I manipulate the bounding boxs of the text objects?

Thanks a lot!

As per request, a runnable example:

import matplotlib.pyplot as plt
import numpy as np


def stacked_bar_chart(
        diagram, title="example question", img_name="test_image", width=0.7, clusters=None, show_axes=True,
        show_legend=True, show_score=True):
    """
    Builds one or multiple scaled stacked bar charts for grade
    distributions. saves image as png.
    :param show_score: whether the score should be shown on top
    :param show_legend: whether the legend should be shown
    :param show_axes: whether question name should be shown on bottom
    :param clusters: indices of clusters to be displayed.
    :param width: the width of the bars as fraction of available space
    :param title: diagram title
    :param img_name: output path
    :param diagram: dictionary: {x-label: [[grade distribution], score]}
    :return: nothing.
    """

    grades = {
        "sehr gut":     "#357100",
        "gut":          "#7fb96a",
        "befriedigend": "#fdd902",
        "ausreichend":  "#f18d04",
        "mangelhaft":   "#e3540e",
        "ungenügend":   "#882d23"
    }

    # select clusters
    if clusters is not None:
        diagram = {i: diagram[i] for i in clusters}

    # normalize score distribution => sum of votes = 1.0
    normalized = []
    for question in diagram.values():
        s = sum(question[0])
        normalized.append([x / s for x in question[0]])

    # transpose dict values (score distributions) to list of lists
    transformed = list(map(list, zip(*normalized)))

    # input values for diagram generation
    n = len(diagram)  # number of columns
    ind = np.arange(n)  # x values for bar center
    base = [0] * n  # lower bounds for individual color set
    bars = []
    fig, ax = plt.subplots()

    # loop over grades
    for name, grade in zip(grades.keys(), transformed):
        assert len(grade) == n, 
            "something went wrong in plotting grade stack " + img_name
        bar = plt.bar(ind, grade, width=width, color=grades[name], bottom=base)
        bars.append(bar)

        # loop over bars
        for i, (rect, score) in enumerate(zip(bar, grade)):
            # update lower bound for next bar section
            base[i] += grade[i]
            # label with percentage
            # TODO text color white
            ax.text(
                rect.get_x() + width / 2, rect.get_height() / 2 + rect.get_y(), "{0:.0f}%".format(score * 100),
                va="center", ha="center")

    # label diagram

    plt.suptitle(title)
    if show_axes:
        plt.tick_params(axis="both", left=False, bottom=False, labelleft=False)
        plt.xticks(ind, diagram.keys())
        ax.set_frame_on(False)

    else:
        plt.tick_params(axis="both", left=False, bottom=False, labelleft=False, labelbottom=False)
        plt.axis("off")

    # show score label above
    if show_score:
        for label, x in zip([q[1] for q in diagram.values()], ind):
            ax.text(
                x, 1.05, '{:4.0%}'.format(label),
                ha="center", va="center",
                bbox={"facecolor": "blue", "pad": 3}
            )

    # create legend
    if show_legend:
        plt.legend(
            reversed(bars), reversed([*grades]),
            bbox_to_anchor=(1, 1), borderaxespad=0)

    # save file
    plt.show()


diagram = {
    "q1": [[1, 2, 3, 4, 5, 6], 0.6],
    "q2": [[2, 3, 1, 2, 3, 1], 0.4]
}
stacked_bar_chart(diagram)

Answer 1

For arguments why setting the width of a text box to a defined width is hard see this question which is about setting the title text box width. In principle the answer over there could be used here as well - making this rather complicated.

A relatively easy solution would be to specify the x-position of the text in data coordinates and its y position in axes coordinates. This allows to create a rectangle as background for the text with the same coordinates such that it looks like a bounding box of the text.

import matplotlib.pyplot as plt
import numpy as np

ind = [1,2,4,5]
data = [4,5,6,4]
perc = np.array(data)/float(np.array(data).sum())
width=0.7 
pad = 3 # points


fig, ax = plt.subplots()
bar = ax.bar(ind, data, width=width)

fig.canvas.draw()
for label, x in zip(perc, ind):
    text = ax.text(
        x, 1.00, '{:4.0%}'.format(label),
        ha="center", va="center" , transform=ax.get_xaxis_transform(), zorder=4)
    bb= ax.get_window_extent()
    h = bb.height/fig.dpi
    height = ((text.get_size()+2*pad)/72.)/h
    rect = plt.Rectangle((x-width/2.,1.00-height/2.), width=width, height=height,
                         transform=ax.get_xaxis_transform(), zorder=3,
                         fill=True, facecolor="lightblue", clip_on=False)
    ax.add_patch(rect)


plt.show()