Bug in errorbar

Dear folks,

I have this problem consistently in 0.99 installed via package
respiratory and also 1.0.1 from source. This only occurs in linux.
Windows is OK. see attachments please. If I run flow-rate.py the output
of plt.show() is different from what is written in pdf. both are
attached. Very weird problem.

Looking forward to hearing from you,

System:Linux Amir-Ubuntu 2.6.38-10-generic #46-Ubuntu SMP Tue Jun 28
15:07:17 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

Other info:gcc version 4.5.2 (Ubuntu/Linaro 4.5.2-8ubuntu4)

image-linux.png

image-windows.png

matplotlibrc (19.1 KB)

flow-rate.py (1.03 KB)

Flow-rate.pdf (68.5 KB)

flow-rate-plt.show().png

Hi Amir,

It seems to me that the error is in the Windows output. In your code, you’ve set “mew” (i.e. marker edge width) to 0. So it makes sense that the markers don’t show. On my system (OS X), the pdf and png output are the same.

Best,

-Tony

···

On Fri, Aug 5, 2011 at 9:29 AM, Amir Jahanshahi <pajoohesh@…287…> wrote:

Dear folks,

I have this problem consistently in 0.99 installed via package

respiratory and also 1.0.1 from source. This only occurs in linux.

Windows is OK. see attachments please. If I run flow-rate.py the output

of plt.show() is different from what is written in pdf. both are

attached. Very weird problem.

Looking forward to hearing from you,

System:Linux Amir-Ubuntu 2.6.38-10-generic #46-Ubuntu SMP Tue Jun 28

15:07:17 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

Hi ,

The code for image-windows and image-linx comes from here:
http://matplotlib.sourceforge.net/examples/pylab_examples/errorbar_limits.html

That is why I think the problem is in linux. And the pdf which is attached along with the png image named flow-rate are executed in linux which are different. in windows both are the same.

Thanks,
Amir

···

On Fri, Aug 5, 2011 at 5:59 PM, Tony Yu <tsyu80@…287…> wrote:

On Fri, Aug 5, 2011 at 9:29 AM, Amir Jahanshahi <pajoohesh@…287…> wrote:

Dear folks,

I have this problem consistently in 0.99 installed via package

respiratory and also 1.0.1 from source. This only occurs in linux.

Windows is OK. see attachments please. If I run flow-rate.py the output

of plt.show() is different from what is written in pdf. both are

attached. Very weird problem.

Looking forward to hearing from you,

System:Linux Amir-Ubuntu 2.6.38-10-generic #46-Ubuntu SMP Tue Jun 28

15:07:17 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

Hi Amir,

It seems to me that the error is in the Windows output. In your code, you’ve set “mew” (i.e. marker edge width) to 0. So it makes sense that the markers don’t show. On my system (OS X), the pdf and png output are the same.

Best,

-Tony

Hi Amir,

So I guess there’s two separate issues: 1) pdf vs png output. 2) windows vs linux output.

  1. I have no idea what could be causing the difference in pdf vs png output. They’re rendered the same on my system, so I can’t be of much help.

  2. I still think the linux output is correct—assuming you’re using the matplotlibrc file you attached. Yes, the example on the matplotlib website has arrows on the error bars, but in your matplotlibrc file, you’ve set lines.markeredgewidth to 0. Thus, you should expect the arrow heads/tails to disappear. The default matplotlibrc has a nonzero markeredgewidth, which is why the heads/tails are visible on the website. So to reiterate, linux seems to be giving the correct behavior. Now, for the windows output, are you certain you’re using the same matplotlibrc? In other words, are you setting markeredgewidth to 0. If so, then I’d say there’s a bug in the windows output.

-Tony

···

On Fri, Aug 5, 2011 at 2:41 PM, Amir Jahanshahi <pajoohesh@…287…> wrote:

Hi ,

The code for image-windows and image-linx comes from here:
http://matplotlib.sourceforge.net/examples/pylab_examples/errorbar_limits.html

That is why I think the problem is in linux. And the pdf which is attached along with the png image named flow-rate are executed in linux which are different. in windows both are the same.

Thanks,
Amir

Thanks Tony for the answer. windows matplotlibrc is not this one, so no bug in windows.
I think I found the problem. Just look at the two screen shots, one in okular, one in foxitreader. this is the pdf which is attached in the first email. that is why I thought there is a problem with linux. In this day and age you can not trust your eyes :slight_smile:

So by putting mew > 0, the problem would be solved I suppose. will try it Monday.
Once more, thanks for help,
Have a nice weekend,
Greets,
Amir

···

On Fri, Aug 5, 2011 at 10:15 PM, Tony Yu <tsyu80@…287…> wrote:

On Fri, Aug 5, 2011 at 2:41 PM, Amir Jahanshahi <pajoohesh@…83…287…> wrote:

Hi ,

The code for image-windows and image-linx comes from here:
http://matplotlib.sourceforge.net/examples/pylab_examples/errorbar_limits.html

That is why I think the problem is in linux. And the pdf which is attached along with the png image named flow-rate are executed in linux which are different. in windows both are the same.

Thanks,
Amir

Hi Amir,

So I guess there’s two separate issues: 1) pdf vs png output. 2) windows vs linux output.

  1. I have no idea what could be causing the difference in pdf vs png output. They’re rendered the same on my system, so I can’t be of much help.
  1. I still think the linux output is correct—assuming you’re using the matplotlibrc file you attached. Yes, the example on the matplotlib website has arrows on the error bars, but in your matplotlibrc file, you’ve set lines.markeredgewidth to 0. Thus, you should expect the arrow heads/tails to disappear. The default matplotlibrc has a nonzero markeredgewidth, which is why the heads/tails are visible on the website. So to reiterate, linux seems to be giving the correct behavior. Now, for the windows output, are you certain you’re using the same matplotlibrc? In other words, are you setting markeredgewidth to 0. If so, then I’d say there’s a bug in the windows output.

-Tony