[BUG] Plot artifact in baseline test images

[matthewfeickert]

Describe the bug

This was found while making test images for PR #161

The baseline test image for pull plots that was introduced in PR #149 has an axis artifact that is viewable in the pull plot on the sides (it additionally suffers from axis problems from being plotted on subplots).

Steps to reproduce

$ pip list | grep hist
boost-histogram               1.0.0
hist                          2.2.1
histoprint                    2.0.0

The following example is trying to replicate most of

hist/tests/test_plot.py

Lines 604 to 625 in e12f748

	def test_image_plot_pull():
	"""
	Test plot_pull by comparing against a reference image generated via
	`pytest --mpl-generate-path=baseline`
	"""
	np.random.seed(42)
	h = Hist(
	axis.Regular(
	50, -4, 4, name="S", label="s [units]", underflow=False, overflow=False
	)
	).fill(np.random.normal(size=100))
	def pdf(x, a=1 / np.sqrt(2 * np.pi), x0=0, sigma=1, offset=0):
	return a * np.exp(-((x - x0) ** 2) / (2 * sigma ** 2)) + offset
	fig, ax = plt.subplots()
	assert h.plot_pull(pdf)
	return fig

# example.py
import matplotlib.pyplot as plt
import numpy as np

from hist import Hist, axis


def pdf(x, a=1 / np.sqrt(2 * np.pi), x0=0, sigma=1, offset=0):
    return a * np.exp(-((x - x0) ** 2) / (2 * sigma ** 2)) + offset


if __name__ == "__main__":
    np.random.seed(42)

    hist_1 = Hist(
        axis.Regular(
            50, -4, 4, name="S", label="s [units]", underflow=False, overflow=False
        )
    ).fill(np.random.normal(size=1000))
    hist_2 = Hist(
        axis.Regular(
            50, -5, 5, name="S", label="s [units]", underflow=False, overflow=False
        )
    ).fill(np.random.normal(size=10000))

    fig = plt.figure(figsize=(10, 8))
    main_ax, pull_ax = hist_1.plot_pull(pdf)
    fig.savefig("debug_hist_1.png")

    fig = plt.figure(figsize=(10, 8))
    main_ax, pull_ax = hist_2.plot_pull(pdf)
    fig.savefig("debug_hist_2.png")

debug_hist_1.png	debug_hist_2.png

It can be seen that debug_hist_2.png (range: [-5, 5]) doesn't exhibit the behavior though debug_hist_1.png (range: [-4, 4]) does, and if you run with a range of [-6, 6] it will exhibit the behavior as well.

From the reproduced bug below I think this is due to the axis from the main axis being drawn over but the edges of it sticking out. I'm not sure though, and would defer to someone more expert in matplotlib. If you look at the artifact closely it appears to be the right hand side of the range limits (in this case 4 — it is easier to see with a range of 6).

[matthewfeickert]

@andrzejnovak have you ever come across things like this before?

[andrzejnovak]

Heh that's new. I wonder if it's related to how the subplots are created

hist/src/hist/plot.py

Line 251 in e12f748

grid = fig.add_gridspec(2, 1, hspace=0, height_ratios=[3, 1])

I tend to use plt.subplots() so I am not sure if that could affect anything

[henryiii]

The add_gridspec is used because the figure might already exist. This line is missing:

plt.setp(ax2.get_xticklabels(), visible=False)

The histogram line style needs to be improved, too, while I'm at it, I think.

[andrzejnovak]

Ah right, in which case it should use axis locators. Like cbars in mplhep.

I see you've already made another fix, so I'm gonna leave that to adding pull to mplhep

[matthewfeickert]

I see you've already made another fix, so I'm gonna leave that to adding pull to mplhep

Yeah I wasn't trying to say that it was a mplhep problem or anything, just that I was curious if you had seen this behavior in the past. 👍