Display an artist label instead of x y not working

[arnaudvre]

I'm using Seaborn, instead of the artist label I get _child0 for example (number is specific to a line). Interestingly on version 0.5.2-2 (apt default) the label shows intermittently, moving to 0.5.3 (pip latest/default) only x and y.

    df = pd.DataFrame(pandas_dict_for_dataframe)

    dfm = df.melt('Date', var_name='Item', value_name='Price')

    g = sns.catplot(x="Date", y="Price", hue='Item', data=dfm, kind='point', height=8, aspect=16 / 8., dodge=2)

    plt.xticks(rotation=45, ha='right')

    mplcursors.cursor(hover=True).connect("add", lambda sel: sel.annotation.set_text(sel.artist.get_label()))
    plt.show()

[anntzer]

Please provide a minimal reproducible example that I can directly copy-paste and run to test the issue.

[arnaudvre]

sure

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import mplcursors

data = {
  "target": [1, 2, 3, 4],
  "dave": [50, 44, 52, 41],
  "jane": [45, 41, 43, 39]
}

df = pd.DataFrame(data)

dfm = df.melt('target', var_name='participant', value_name='calories')

g = sns.lineplot(x="target", y="calories", hue='participant', data=dfm)

mplcursors.cursor(hover=True).connect("add", lambda sel: sel.annotation.set_text(sel.artist.get_label()))

plt.show()

[anntzer]

Thanks for the example. This appears to be because seaborn does not set the label properly on the relevant artist, but rather on a "fake" artist used only for the purposes of drawing the legend (see add_legend_data in seaborn's codebase). I guess they probably intentionally don't set the label at all on the real artist in order to prevent duplicate legend entries, but as a consequence, mplcursors cannot figure out the correct label.
As I don't use seaborn myself, I don't really have any suggestions for a fix.