Dynamically resizing a plot to fit text annotations within the plot area

I am adding text annotations to points of a scatter plot as follows:

I do not know the length of the text annotations in advance. The figure above shows that the text annotations extend beyond the plot area.

I want to find the length of each text annotation in data coordinates to dynamically resize the plot by setting ylim to fit all the text annotations within the plot area.

I tried three solutions based on the following stack overflow answers.

I tested using two setups:

• macOS Catalina - 10.15.7

• Python - 3.8.14

• matplotlib - 3.3.1

and

• Windows 11 - 22H2

• Python - 3.10.2

• matplotlib - 3.5.1

import matplotlib.pyplot as plt
from matplotlib.transforms import TransformedBbox, Bbox


def get_text_object_height_1(text_obj, _ax):
    # Based on
    # https://stackoverflow.com/questions/24581194/matplotlib-text-bounding-box-dimensions

    # to get the text bounding box we need to draw the plot
    _fig = _ax.get_figure()
    _fig.canvas.draw()

    # get bounding box of the text in the data coordinates
    bb = text_obj.get_window_extent(renderer=_fig.canvas.get_renderer())
    transform = _ax.transData.inverted()
    trans_box = TransformedBbox(bb, transform)

    return trans_box.height


def get_text_object_height_2(text_obj, _ax):
    # Based on
    # https://stackoverflow.com/a/35419796/2912349
    # https://stackoverflow.com/questions/58854335/how-to-label-y-ticklabels-as-group-category-in-seaborn-clustermap/58915100#58915100

    # Must move the plot to see the last update. When the plot is saved, the last update is included.

    # the figure needs to have been drawn once, otherwise there is no renderer
    plt.ion()
    plt.show()
    plt.pause(1)  # Can make the pause smaller. Kept it lager to see the action.

    # get bounding box of the text in the data coordinates
    bb = text_obj.get_window_extent(renderer=_ax.get_figure().canvas.get_renderer())
    transform = _ax.transData.inverted()
    trans_box = TransformedBbox(bb, transform)

    plt.ioff()

    return trans_box.height


def get_text_object_height_3(text_obj, _ax):
    # Based on
    # https://stackoverflow.com/questions/5320205/matplotlib-text-dimensions

    # to get the text bounding box we need to draw the plot
    _fig = _ax.get_figure()

    # get text bounding box in figure coordinates
    renderer = _fig.canvas.get_renderer()
    bbox_text = text_obj.get_window_extent(renderer=renderer)

    # transform bounding box to data coordinates
    trans_box = Bbox(_ax.transData.inverted().transform(bbox_text))

    return trans_box.height


x = [1, 2, 3]
y = [5, 8, 7]
labels = ['mid length', 'short', 'this is a long label']

fig, ax = plt.subplots(dpi=300, figsize=(5, 3))
ax.scatter(x=x, y=y)

y_lim_max = 0
gap = 0.3

for idx, label in enumerate(labels):
    label_y_start = y[idx] + gap
    txt = ax.text(x[idx], label_y_start, label, rotation='vertical', fontdict=dict(color='black', alpha=1, size=8),
                  transform=ax.transData)

    # Can change to _2, _3 to see how other two methods work
    label_length = get_text_object_height_1(txt, ax)

    label_y_end = label_y_start + label_length
    # Just to show the computed label length
    plt.plot([x[idx], x[idx]], [label_y_start, label_y_end])
    y_lim_max = label_y_end if y_lim_max < label_y_end else y_lim_max

print(f'\t{y_lim_max:8.2f}')
plt.ylim(0, y_lim_max)
plt.tight_layout()

plt.show()
# plt.savefig('scaled_plot.png')

However, the computed text annotation lengths are shorter than the text annotations. Please note that I am plotting a line beside each text annotation to demonstrate the calculated length of the respective text annotation.

Am I doing something incorrectly?

Is there a way to get the correct dimensions of the text annotations?

Is there a different approach to resizing the plot area to include all the text annotations?