Closed
Description
Bug summary
there are two thread pools working together.
savefig is in the one of them.
With matplotlib 3.2.0 it costs 1s.
With matplotlib 3.4.3 it costs 30s.
Code for reproduction
import time
from concurrent.futures import ThreadPoolExecutor, as_completed
import matplotlib
from matplotlib.figure import Figure
matplotlib.use('agg')
import matplotlib.pyplot as plt
import matplotlib.patches as patches
POOL_1 = ThreadPoolExecutor(3)
POOL_2 = ThreadPoolExecutor(3)
def _pull_stream():
while True:
a = [100]
if a[0] < 200:
continue
def generate_empty():
fig = Figure(figsize=(9.6, 12.8))
fig.savefig("./test.jpg")
fig.clear()
def draw_process():
start_time = time.time()
img = plt.imread("./test.jpg")
fig = plt.figure(figsize=(9.6, 12.8))
ax_var = plt.Axes(fig, [0., 0., 1., 1.])
ax_var.set_axis_off()
ax_var.imshow(img, aspect="equal")
fig.add_axes(ax_var)
for i in range(3):
for j in range(10):
ax_var.add_patch(patches.Rectangle((1, 1), 2, 3))
ax_var.text(4, 0, 'Here you are', family='fantasy', fontsize=12, style='italic', color='mediumvioletred')
plt.axis('off')
plt.savefig("./test_res.jpg")
plt.close('all')
return "cost {}s".format(time.time() - start_time)
def run():
POOL_1.submit(_pull_stream)
all_task = [POOL_2.submit(draw_process) for i in range(10)]
for future in as_completed(all_task):
data = future.result()
print("in main: get page {}s success".format(data))
if __name__ == '__main__':
generate_empty()
run()Actual outcome
i just write a simple draw process, for my real process i use add_patch(Rectangle) and text to add some elements for my pic.
matplotlib 3.2.0
matplotlib3.4.3
Expected outcome
get the same speed as 3.2.0
Additional information
No response
Operating system
No response
Matplotlib Version
3.4.3
Matplotlib Backend
agg
Python version
3.8.8
Jupyter version
No response
Installation
pip