`lineplot` not working when `hue` changes along `units`

[RobinSobczyk]

Hi, I'd like to highlight that seaborn does not handle hue grouping when the hue variable is not constant inside units. Here is a MWE :

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns

df_data = []
for i in range(5):
    for j in range(5):
        df_data.append(
            {
                "Traj_id": i,
                "Step_id": j,
                "X": np.random.random(),
                "Y": np.random.random(),
            }
        )
df = pd.DataFrame(df_data)
sns.lineplot(
    data=df,
    x="X",
    y="Y",
    hue="Step_id",
    markers=True,
    sort=False,
    units="Traj_id",
    estimator=None,  # type: ignore
)
plt.show()

What is expected is color changing along the trajectories, but all I get is a blank figure (with a legend though) without getting any warning or error.
Could be interesting to have this feature to plot trajectory tendencies depending on step conditions.

[RobinSobczyk]

Don't know if it helps, but here's my code to use the function the way I wanted :

import matplotlib.pyplot as plt
import numpy as np
from matplotlib.collections import LineCollection
from matplotlib.lines import Line2D

def traj_plot(df, x, y, hue, units, order, hue_order = None):
    # sort df
    df = df.sort_values(order)

    fig, ax = plt.subplots(figsize=(8, 6))

    color_ids = df[hue].unique()
    if hue_order is not None:
        hue_order_dict = {hue_val : id for id, hue_val in enumerate(hue_order)}
        color_ids = sorted(color_ids, key = hue_order_dict.get) 
    cmap = plt.get_cmap("magma", len(color_ids) + 1)
    color_map = {hue_val: cmap(i) for i, hue_val in enumerate(color_ids)}

    # iterate trajs
    for _, traj_df in df.groupby(units):
        # get x, y, hue
        x_vals = traj_df[x].values
        y_vals = traj_df[y].values
        hue_vals = traj_df[hue].values

        # make (0,0) the starting point of the lines, this is just for my use case
        x_vals = np.insert(x_vals, 0, 0.0)
        y_vals = np.insert(y_vals, 0, 0.0)
        # the colors of each step, idk how to adapt when there is no starting point
        step_colors = [color_map[hue_val] for hue_val in hue_vals]

        # put the coords of the steps in an array
        points = np.array([x_vals, y_vals]).T.reshape(-1, 1, 2)
        segments = np.concatenate([points[:-1], points[1:]], axis=1)

        # put steps in line collection
        lc = LineCollection(segments, colors=step_colors, linewidths=2, alpha=0.2)
        ax.add_collection(lc)

    # categorical legend, you probably have your code for continuous values as well
    legend_elements = [
        Line2D([0], [0], color=color_map[val], lw=3, label=val) for val in color_ids
    ]
    ax.legend(handles=legend_elements, title=hue, loc="best")

    ax.set_xlabel(x)
    ax.set_ylabel(y)
    ax.axhline(0, color="gray", lw=0.5, linestyle="--")
    ax.axvline(0, color="gray", lw=0.5, linestyle="--")
    ax.autoscale()
    plt.tight_layout()
    plt.show()

For seaborn particularly, I'd add a line in the dataframe with (0,0) for the starting point, and probably "Step_id" = pd.NA, I don't know how to integrate that in the colors though as pd.NA should then be tolerated only once : for the first point of each trajectory.