Bug summary

Despite being documented [here] and (https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.annotate.html)
here, providing ShinkA and ShrinkB are ignored in annotations.

Code for reproduction

import matplotlib.pyplot as plt
import numpy as np

fig, ax = plt.subplots()
x = np.linspace(0, 10, 100)
y = np.sin(x)
ax.plot(x, y)

# Annotate a some points with varying shrinkA values
annotation_x_values    = [3,4,5,6,7]
shrinkA_values = [0, 0.2, 0.4, 0.6, 0.8]
for shrinkA, x_anno in zip(shrinkA_values, annotation_x_values):
    y_anno = np.sin(x_anno)
    ax.annotate(
        f'shrinkA={shrinkA}',
        xy=(x_anno, y_anno),
        xytext=(x_anno+1, y_anno+0.2),
        arrowprops=dict(arrowstyle="->", shrinkA=shrinkA),
        fontsize=10
    )


plt.tight_layout()
plt.show()

Actual outcome

All the arrows in the example above look the same; ShinkA/ShrinkB has no effect.

Expected outcome

Expected that the beginning / end of the arrows are moved toward or away from the source/target of the arrow.

By way of example, the simplearrow shrink= parameter (when arrowstyle is omitted) works as expected:

import matplotlib.pyplot as plt
import numpy as np

fig, ax = plt.subplots(figsize=(5, 3))
x = np.linspace(0, 10, 100)
y = np.sin(x)
ax.plot(x, y)

# Annotate a some points with varying shrinkA values
annotation_x_values    = [3,4,5,6,7]
shrink_values = [0, 0.2, 0.4, 0.6, 0.8]
for shrink, x_anno in zip(shrink_values, annotation_x_values):
    y_anno = np.sin(x_anno)
    ax.annotate(
        f'shrink={shrink}',
        xy=(x_anno, y_anno),
        xytext=(x_anno+1, y_anno+0.2),
        arrowprops=dict(shrink=shrink),
        fontsize=10
    )
plt.tight_layout()
plt.show()

However then I cant use the desired arrowstyle='-->' or control which end of the arrow is shrunk.

Additional information

I have not been able to confirm, but I belive this comment from the documentation for FancyArrowPatch is the likely cause (link to doc source).

Alternatively if *path* is provided, an arrow is drawn along this path and *patchA*, *patchB*, *shrinkA*, and *shrinkB* are ignored.

My guess was that in the implementation of Axes.annotate, the *path* is provided?
I cant find if that is happening though.

I think debugging near this use set_positions() might be a good starting point?

Operating system

windows

Matplotlib Version

3.8.3

Matplotlib Backend

module://matplotlib_inline.backend_inline

Python version

Python 3.11.5

Jupyter version

VS Code Notebook (ipython Version: 8.17.2)

Installation

pip