Inconsistent behavior of markevery between semilogx and semilogy plots

Bug report

The markevery option for plot, semilogx, semilogy, and loglog plotting methods has a weird difference of behavior between plot / semilogy and semilogx / loglog, when using markevery = (float, float).

Code for reproduction

import numpy as np
import matplotlib.pyplot as plt

# Generate some data
X = np.linspace(-10, 10, 1000)
Y = np.sin(X)**2

# Specify markers
markers = ['o', 'v', '^', '<', '>']

def markevery(i):
    # I tried a *lot* of values here, none were working for the 4 plots
    return (i / 50., 0.1)

plt.figure()

# First plot in normal plot
plt.subplot(221)
for i in range(10):
    plt.plot(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
plt.legend()
plt.title("plt.plot(): markers are here!")

# Then plot in semilogx
plt.subplot(222)
for i in range(10):
    plt.semilogx(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
plt.legend()
plt.title("plt.semilogx(): markers are NOT here!")

# And then semilogy
plt.subplot(223)
for i in range(10):
    plt.semilogy(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
plt.legend()
plt.title("plt.semilogy(): markers are here!")

# Finally plot in loglog
plt.subplot(224)
for i in range(10):
    plt.loglog(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
plt.legend()
plt.title("plt.loglog(): markers are NOT here!")

plt.suptitle("Weird behavior of markevery between plt.plot, plt.semilogx, plt.semilogy, plt.loglog")
plt.show()

Actual outcome

Expected outcome

One would expect the markers to be located in a similar fashion for each of the four plots method, plot, semilogx, semilogy, and loglog. The two plots on the left have markers correctly located (well spacen, with a different starting point for the different curves), and the two plots on the right do not show markers.
Note that I tried a lot of different values for the markevery parameters (here markevery = (i / 50., 0.1) for the various curves, with changing i to change the starting point of the markers), and always encounter this inconsistency between left and right.

Matplotlib version

Matplotlib version 2.0.0 (2.0.0+3337.gc15694b), with Python 3.5, on Ubuntu 16.10
Python installed with apt, latest version of Matplotlib installed with pip --upgrade.

Maybe I don't understand the semantic behind using markevery = (float, float), but as far as I can tell, this behavior is weird.
Thanks in advance!

By the way, the reason I am asking is because I use a similar trick of markevery = (i / 50., 0.1) in some research code showing half a dozen of plots in the same figure, and I am puzzled that it exhibits a different behavior between plot and semilogy: markers are perfectly displayed but not for semilogy plots...

Can't figure this bug, see matplotlib/matplotlib#8356

Thanks for the detailed report. What happens if you use plot and set the x/yscales afterwards?

Thanks for your quicky reply.
Well, it fixes the issue.

The corrected script can be:

import numpy as np
import matplotlib.pyplot as plt

# Generate some data
X = np.linspace(-10, 10, 1000)
Y = np.sin(X)**2

# Specify markers
markers = ['o', 'v', '^', '<', '>']

def markevery(i):
    return (i / 50., 0.1)

fig = plt.figure()

# First plot in normal plot
ax = fig.add_subplot(2, 2, 1)
for i in range(10):
    plt.plot(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
plt.legend()
plt.title("plt.plot(): markers are here!")

# Then plot in semilogx
ax = fig.add_subplot(2, 2, 2)
for i in range(10):
    plt.plot(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
    ax.set_xscale('log')  # Instead of using plt.semilogx
plt.legend()
plt.title("plt.plot() with ax.set_xscale('log'): markers are NOW here! Fixed!")

# And then semilogy
ax = fig.add_subplot(2, 2, 3)
for i in range(10):
    plt.plot(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
    ax.set_yscale('log')  # Instead of using plt.semilogy
plt.legend()
plt.title("plt.plot() with ax.set_yscale('log'): markers are here!")

# Finally plot in loglog
ax = fig.add_subplot(2, 2, 4)
for i in range(10):
    plt.plot(X, Y + i / 10., marker=markers[i % len(markers)], markevery=markevery(i), label=str(i))
    ax.set_xscale('log')  # Instead of using plt.loglog
    ax.set_yscale('log')  # Instead of using plt.loglog
plt.legend()
plt.title("plt.plot() with ax.set_xscale('log') and ax.set_yscale('log'): markers are NOW here! Fixed!")

plt.suptitle("Weird behavior of markevery between plt.plot, plt.semilogx, plt.semilogy, plt.loglog\nhttps://github.com/matplotlib/matplotlib/issues/8356")
plt.show()

And it gives what the first one should have given:

OK. Thanks, that'll help us track it down.

Is there a way to use set_xscale or set_yscale without having to store and use the figure and then the axis object?

Yes, but I always recommend that you explicitly create and store your Figure and Axes object.

You can do e.g., plt.xscale('log'), but then you have keep track in your head which Axes you're operating on. Kind of a mental burden, IMHO.

For instance, I would change your code to be

fig, axes = pyplot.subplots(nrows=2, ncols=2)
axes[0, 0].plot(...)
axes[0, 0].set_xscale('log')
ax.title(...)
ax.legend(...)
# yada yada

alternatively, you could do:

fig, ((ax1, ax2), (ax3, ax4))= pyplot.subplots(nrows=2, ncols=2)
ax1.plot(...)
ax1.set_xscale('log')
# yada yada

OK, I can confirm that setting the xscale/yscale manually instead of using plt.semilogx/plt.semilogy works also for my research code.

If you feel like you need help, I will try to fix the issue myself on matplotlib's code, if someone points me to the exact and correct function/part that needs to be modified. Don't hesitate to ask me, I will try my best!

Start digging at e.g., https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/axes/_axes.py#L1551

Seems to me that, for safety, the self.set_xscale('log', **d) could also be done after the l = self.plot(*args, **kwargs) ?
Doing this (hacky) change, for semilogx,semilogy & loglog, will fix the issues.

In fact, I am not sure that the proper way to use set_x/yscale is before the call to plot.
Could someone confirm?

If the semantic is not properly defined, the safer way to fix bug (but not the cleanest) is to call set_xscale before and after the call to plot(...)...

OK. Clearly, I don't enough about internals of matplotlib to really understand what you explained, but I'm glad to see you seem to know what's going on.
Thanks, and good luck!

Will this be included for the 2.1 release?

@Naereen

@Naereen we've milestoned it to be fixed by 2.1, but that's not a 100% guarantee.

Hi,
Any update? Any help needed?

Any help working on this would be appreciated :)

Does setting markevery on a line object correctly recompute the locations with the current scale?

Yes

This issue has been marked "inactive" because it has been 365 days since the last comment. If this issue is still present in recent Matplotlib releases, or the feature request is still wanted, please leave a comment and this label will be removed. If there are no updates in another 30 days, this issue will be automatically closed, but you are free to re-open or create a new issue if needed. We value issue reports, and this procedure is meant to help us resurface and prioritize issues that have not been addressed yet, not make them disappear. Thanks for your help!

This seems to have been fixed somewhere along the way, not too sure where:

This was fixed by #9477 (because the markevery calculation produces all nan and thus plots nowhere.)

Navigation Menu

Repository navigation

Inconsistent behavior of markevery between semilogx and semilogy plots #8356

Bug report

6 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Inconsistent behavior of markevery between semilogx and semilogy plots #8356

Description

Bug report

Activity

Naereen commented on Mar 21, 2017

phobson commented on Mar 21, 2017

Naereen commented on Mar 21, 2017

phobson commented on Mar 21, 2017

Naereen commented on Mar 21, 2017

phobson commented on Mar 21, 2017

phobson commented on Mar 21, 2017

Naereen commented on Mar 21, 2017

Naereen commented on Mar 22, 2017

phobson commented on Mar 22, 2017

Naereen commented on Mar 22, 2017

Naereen commented on Mar 22, 2017

6 remaining items

Naereen commented on Mar 23, 2017

Naereen commented on May 11, 2017

phobson commented on May 11, 2017

Naereen commented on Jun 19, 2017

tacaswell commented on Jun 19, 2017

Naereen commented on Jun 22, 2017

github-actions commented on Apr 10, 2023

ksunden commented on Apr 10, 2023

QuLogic commented on Apr 11, 2023

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions