Repository navigation
Errorbar semilogy switch to log y removed error bars when crosses into negative #9149
Description
Bug report
Bug summary
When plotting error bars on data points, where the data is positive but the bar with the data-error is entering negative values, switching to log Y axes using semilogy() means that some data points are plotted without error bars at all (both y-err bars disappear).
Code for reproduction
create a file example.dat:
x dx y dy
1.046850 0.0029168 0.042551 0.02420
1.506679 0.0030152 0.027573 0.02252
3.028597 0.1959554 0.021218 0.00974
3.656348 0.1034119 0.002788 0.01059
4.949535 0.4693247 0.006276 0.00969
from astropy.io import ascii
from pylab import figure, errorbar,semilogy
t = ascii.read('example.dat')
figure()
errorbar(t['x'],t['y'],xerr=t['dx'],yerr=t['dy'],fmt='s')
This shows a linear plot with errorbars fine. The third data point error bar extends into negative territory.
semilogy()
Now the plot has a log y axis, but the error bar of the third point is gone.
# Paste your code here
#
#from astropy.io import ascii
from pylab import figure, errorbar,semilogy
t = ascii.read('example.dat')
figure()
errorbar(t['x'],t['y'],xerr=t['dx'],yerr=t['dy'],fmt='s')
#This shows a linear plot with errorbars fine. The third data point error bar extends into negative territory.
semilogy()
#Now the plot has a log y axis, but the error bar of the third point is gone.
Actual outcome
# If applicable, paste the console output here # #
Expected outcome
An errorbar in the log plot, surely the positive bar, and the negative going a few (3 or 5) decades down (perhaps with an arrow ending).
Matplotlib version
- Operating System: OSX
- Matplotlib Version: 2.0.2
- Python Version:2.7
- Jupyter Version (if applicable):
- Other Libraries: astropy
anaconda astroconda
Activity
dstansby commented on Sep 1, 2017
Yeah, we change the behaviour of this recently to fix some other issues (#8836). There is a note in the docstring of errorbar that what you're describing is expected behaviour, but I do agree that we should find some way to put the errorbars back in and have them go right down.
tacaswell commented on Sep 1, 2017
I think the issue here is that exactly what should be done with the negative is not clear, I think there are at least 4 options:
See https://matplotlib.org/examples/statistics/errorbar_limits.html for all the ways to have finer grain control over how limits are drawn (if you want to use log then negative values are probably non-physical and you really just have no lower bound for the value (ex neutrino mass)) so it might be worth doing the extra work to identify that.
Also see https://matplotlib.org/api/_as_gen/matplotlib.axes.Axes.set_yscale.html , if you use ax.set_yscale('log', nonposy='clip') it should clip the error bars to some very small number.
github-actions commented on Apr 17, 2023
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!