Time per edge in plotting lines with Matplotlib.

line_time_per_edge.py

import time

from matplotlib import pyplot as plt
import numpy as np


ns = [10, 30, 100, 300, 1000, 3000, 10000]
all_times = []

for n in ns:
    print(f"{n} edges")
    fig, ax = plt.subplots()
    ax.set(xticks=[], yticks=[])
    for spine in ax.spines.values():
        plt.setp(spine, visible=False)
    data = np.random.RandomState(0).random_sample((2, n + 1))
    ax.plot(*data, antialiased=True, solid_joinstyle="miter")

    def profile(max_time=1, timer=time.perf_counter):
        times = []
        base_start = timer()
        while timer() - base_start < max_time:
            start = timer()
            fig.canvas.draw()
            elapsed = timer() - start
            times.append(elapsed)
        return times

    times = profile()
    all_times.append(times)
    plt.close("all")

fig, axs = plt.subplots(len(ns), sharex=True)
min_time = np.min(np.concatenate(all_times))
max_time = np.max(np.concatenate(all_times))
bins = np.geomspace(.99 * min_time,  1.01 * max_time)
for ax, n, times in zip(axs, ns, all_times):
    ax.hist(times, bins, normed=True)
    ax.set(xlabel="$t$ (s)", xscale="log")
    ax.text(.95, .95, f"{n} edges (N={len(times)})",
            ha="right", va="top", transform=ax.transAxes)
    ax.label_outer()

fig, ax = plt.subplots()
# Use the minimum time as aggregate.
ax.plot(ns, [np.min(times) / n for n, times in zip(ns, all_times)], "o")
ax.set(xlabel="number of edges", ylabel="time per edge (s)",
       xscale="log", yscale="log")

plt.show()