-
-
Notifications
You must be signed in to change notification settings - Fork 7.9k
Make set_yscale("log") consistent with semilogy() #8537
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
clipping behavior as plt.semilogy(), plt.semilogx() plt.loglog() when handing non-positive data.
If you remove the re-generated image, does it pass? (I can't spot the differences in the github UI)
|
It seems to be the svg image (not the png one, which I updated) which is failing. It's not failing locally on my OSX machine, and isn't failing on the OSX build either, so there must be a difference between linux and OSX when it comes to svg. |
@dstansby maybe you need regenerate the SVG too. |
It's not failing on my machine, so I have no way to regenerate it! I presume if a linux box regenerated it the OSX build would fail? |
(note to myself to restart travis when #8672 gets merged) |
@dstansby done for you. |
Restarting travis didn't move the build to 14.04, so manually cycling the PR. |
Huh, still no luck on travis, will try and see what happens on a local linux box at some point. |
Can you check the history of why this was changed to clip? I have a vague memory that there was a long discussion at the time... |
It was originally changed to fix #163 (comment) - changing it back will break that problem, so I adding the code in that comment as a test would be a good idea. |
I do not see what changed about the test image... |
I think the basic idea of this--essentially reverting #2147 and thereby restoring problematic errorbar behavior (#163)--will have to be noted, and we will have to be prepared for someone to raise the issue again. Overall I think it is an improvement, though. I think that the errorbar problem could be fixed in the errorbar code in this way: for any errorbar covering a range from negative to positive values, insert a point in the Line2D at 1e-300. It would have no effect with a linear scale, and with a log scale it would allow a line to be plotted after the negative end-point is masked out. Worth the trouble? I'm not sure. |
I am in favor of:
|
This is an update of #8056, which fixes #8045 - I have now added the new test image, which should make the tests pass. Thanks to @jcalbert for the original fix.