Aligning axvline and axhline with increasing linewidth

[djrobust]

I would like to draw a vertical line at x=0.5, and I want it to be visible, so I use

ax.axvline(x=0.5, linewidth=2)

However, it seems that the line gets fatter asymmetrically, expanding only to the right. If I set linewidth=20 this becomes even more visible.

How can I align the line to remain centered around x=0.5, even with increasing width?

Is there an argument in axvline/axhline equivalent to ax.bar's align='center'?

[tacaswell]

What backend are you using? I can not reproduce this:

(I made the ticks huge via ax.tick_params(width=5, length=15, color='k'))

[djrobust]

Okay I found out what the issue was. Using an IPython Notebook I had a graph like this

#!/usr/bin/python3
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

np.random.seed(42)

plt.rcParams.update({'font.size': 23, 'font.family': 'serif'})
f, ax = plt.subplots(1, 1, figsize=(16,10), dpi=600)

data = np.random.rand(50)
hist, bins = np.histogram(data, bins=99, range=(0.0, 1.0))

midpoints, hist = [0.5, 0.25]
ax.bar(midpoints, hist, align='center', width=0.9*(bins[1]-bins[0]),
           color='green')
ax.axis([0, 1, 0, 0.35])
ax.grid(True)
ax.axvline(x=0.5, linewidth=2, alpha=0.5)

The vertical line is not centered on my screen - because the bar and vertical line width are displayed using 7 and 2 pixels respectively. I then increased the scale of the graphic in my IPython Notebook using the bottom right edge, but the graphic did not rerender (why?) and hence the imbalance remained even after rescaling.

In that sense if we increase the linewidth from 1 to 2 pixels, we get an imbalance. I cannot reproduce the originally reported issue of 'expanding only to the right'.

In sum, this is probably not an issue with matplotlib, but just a property that comes from using pixels to display stuff.

[tacaswell]

There is also a snap property to artists, turning that off might help.

[djrobust]

Yes, snap=False fixed the issue. Thanks a lot!