matplotlib-devel Mailing List for matplotlib

I've run into a case where the rendering in a line plot is incomplete
and some lines are not drawn at all.  I submitted a question to
matplotlib-users with the same subject.  Eric Firing responded that
this is a manifestation of the "infamous path simplification" bug,
which should be fixed in svn.  However, the script below shows the
same problem for me using the current svn on CentOS-5 with python 2.6.
 For my purposes setting path.simplify to False is fine but Eric
requested that I provide an example, so here it is:

import numpy
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

y = numpy.array([
 4., 2., 2., 3., 3., 2., 2., 6., 6., 5., 5., 4., 4.,
 7., 7., 2., 2., 4., 4., 2., 2., 2., 2., 4., 4., 4.,
 4., 4., 4., 7., 7., 3., 3., 5., 5., 4., 4., 5., 5.,
 4., 4., 7., 7., 6., 6., 2., 2., 2., 2., 5., 5., 4.,
 4., 4., 4., 6., 6., 3., 3., 4., 4., 3., 3., 2., 2.,
 3., 3., 4., 4., 4., 4., 4., 4., 6., 6., 5., 5., 4.,
 4., 7., 7., 3., 3., 4., 4., 4., 4., 5., 5., 4., 4.,
 7., 7., 3., 3., 4., 4., 4., 4., 6., 6., 4., 4., 4.,
 4., 4., 4., 2., 2., 5., 5., 6., 6., 3., 3., 5., 5.,
 4., 4., 0., 0., 5., 5., 1., 1., 4., 4., 5., 5., 4.])

plt.figure(figsize=(7,4))
plt.plot(y)
plt.savefig('test.png')

plt.xlim(-12000, 8274)
plt.savefig('test_panned.png')


- Tom

Gökhan Sever wrote:

> On Sun, Apr 25, 2010 at 1:37 PM, Neal Becker
> <ndb...@gm...> wrote:
> 
>> Qt4Agg is the one that I used first - the one that doesn't work.
>>
> 
> Pardon my wrong guess :) Sometimes this Linux world goes beyond/behind my
> expectations...
> Can you run another Qt4 application properly?
> 

I'm sure I've used lots of PyQt4 apps before.

I just tried  hp-toolbox, which seems to be PyQt4, and AFAICT it is working.  
Of course, matplotlib partly works also.  Just the save dialog is broken.

On Sun, Apr 25, 2010 at 1:37 PM, Neal Becker <ndb...@gm...> wrote:

> Qt4Agg is the one that I used first - the one that doesn't work.
>

Pardon my wrong guess :) Sometimes this Linux world goes beyond/behind my
expectations...
Can you run another Qt4 application properly?

-- 
Gökhan

Gökhan Sever wrote:

> On Sun, Apr 25, 2010 at 1:28 PM, Neal Becker
> <ndb...@gm...> wrote:
> 
>> savefig works fine.
>>
>> Using 'gtk' backend I can save, but seems to have other problems.
>>
>> Using recommended TkAgg I get error that
>> ImportError: No module named backend_tkagg
>>
> 
> You will have to make some installations to bring TkAgg alive. I would
> suggest trying also the Qt4Agg backend since KDE based on Qt. It is always
> a good idea to use the package manager for these. Let me know if this
> helps.

Qt4Agg is the one that I used first - the one that doesn't work.

On Sun, Apr 25, 2010 at 1:28 PM, Neal Becker <ndb...@gm...> wrote:

> savefig works fine.
>
> Using 'gtk' backend I can save, but seems to have other problems.
>
> Using recommended TkAgg I get error that
> ImportError: No module named backend_tkagg
>

You will have to make some installations to bring TkAgg alive. I would
suggest trying also the Qt4Agg backend since KDE based on Qt. It is always a
good idea to use the package manager for these. Let me know if this helps.

Gökhan Sever wrote:

> On Sun, Apr 25, 2010 at 1:02 PM, Neal Becker
> <ndb...@gm...> wrote:
> 
>> Yes, I get this:
>>
>> kfilemodule(29192)/kio (KDirModel) KDirModelPrivate::_k_slotDeleteItems:
>> No node found for item that was just removed:
>> KUrl("file:///home/nbecker/test1.cs")
>> kfilemodule(29192): couldn't create slave: "Unable to create io-slave:
>> klauncher said: Unknown protocol ''.
>> "
>> kfilemodule(29192): couldn't create slave: "Unable to create io-slave:
>> klauncher said: Unknown protocol ''.
>> "
>>
> 
> Very specific KDE errors. Sorry can't comment on these errors.
> 
> 
>>
>> Have not tried savefig or other backend.  (Have to learn how to use
>> savefig first)
>>
> 
> These are easy. Within IPython -pylab run your script with run myscript.py
> . Then do a plt.savefig("myplot.png") or whatever extension available to
> you. To change the backend go into your .matplotlib directory and edit
> matplotlibrc file. Find backend keyword and experiment with the available
> backends to see if you can reproduce that error.
> 

savefig works fine.

Using 'gtk' backend I can save, but seems to have other problems.

Using recommended TkAgg I get error that 
ImportError: No module named backend_tkagg

Hello,

The amplitude of sharp peaks is not shown correctly when several plots
are stitched together and the x scale becomes very large.  I have
noticed this problem with the pdf and png backends in the attached
script.

Best regards,
Miguel

On Sun, Apr 25, 2010 at 1:02 PM, Neal Becker <ndb...@gm...> wrote:

> Yes, I get this:
>
> kfilemodule(29192)/kio (KDirModel) KDirModelPrivate::_k_slotDeleteItems: No
> node found for item that was just removed:
> KUrl("file:///home/nbecker/test1.cs")
> kfilemodule(29192): couldn't create slave: "Unable to create io-slave:
> klauncher said: Unknown protocol ''.
> "
> kfilemodule(29192): couldn't create slave: "Unable to create io-slave:
> klauncher said: Unknown protocol ''.
> "
>

Very specific KDE errors. Sorry can't comment on these errors.


>
> Have not tried savefig or other backend.  (Have to learn how to use savefig
> first)
>

These are easy. Within IPython -pylab run your script with run myscript.py .
Then do a plt.savefig("myplot.png") or whatever extension available to you.
To change the backend go into your .matplotlib directory and edit
matplotlibrc file. Find backend keyword and experiment with the available
backends to see if you can reproduce that error.

-- 
Gökhan

Gökhan Sever wrote:

> On Sun, Apr 25, 2010 at 12:52 PM, Neal Becker
> <ndb...@gm...> wrote:
> 
>> Yes, I meant the 'save' button in the upper right.
>>
>> It used to work, but some update seems to have broken it.  Strange, so
>> far the only application that shows a problem like this (and I do use a
>> lot of different apps!)
>>
> 
> Do you get any error messages behind the scenes?  Does it work when you
> save the plot using savefig? Is it same with different backends? I am on
> GNOME. You seem to like using KDE, right?
> 

Yes, I get this:

kfilemodule(29192)/kio (KDirModel) KDirModelPrivate::_k_slotDeleteItems: No 
node found for item that was just removed: 
KUrl("file:///home/nbecker/test1.cs") 
kfilemodule(29192): couldn't create slave: "Unable to create io-slave:
klauncher said: Unknown protocol ''.
" 
kfilemodule(29192): couldn't create slave: "Unable to create io-slave:
klauncher said: Unknown protocol ''.
"

Yes, I am on KDE.

Have not tried savefig or other backend.  (Have to learn how to use savefig 
first) 

On Sun, Apr 25, 2010 at 12:52 PM, Neal Becker <ndb...@gm...> wrote:

> Yes, I meant the 'save' button in the upper right.
>
> It used to work, but some update seems to have broken it.  Strange, so far
> the only application that shows a problem like this (and I do use a lot of
> different apps!)
>

Do you get any error messages behind the scenes?  Does it work when you save
the plot using savefig? Is it same with different backends? I am on GNOME.
You seem to like using KDE, right?

-- 
Gökhan

Gökhan Sever wrote:

> On Sun, Apr 25, 2010 at 12:33 PM, Neal Becker
> <ndb...@gm...> wrote:
> 
>> python-matplotlib-0.99.1.2-1.fc12.x86_64
>>
>> When I hit 'print' button, I get a partially painted dialog.  Screenshot
>> attached.  This is repeatable on 2 different machines.
>>
> 
> I am on Fedora 12. Where is print button in matplotlib? You meant the save
> dialog? If that's the case I don't see any problem using the svn version
> of mpl.
> 
> I build sources manually ->
> http://matplotlib.sourceforge.net/faq/installing_faq.html#install-from-svn
> 
> Similarly I build numpy and scipy from sources too. If available I use
> easy_install for some packages, otherwise mostly manual builds. (Except
> the Python itself.) This might be confusing since I use the package
> manager to install dependencies most of the time. Once in a while I see
> some packages requiring older numpy builds, also when I try to uninstall
> some packages via package manager it tries to uninstall more packages (the
> critical ones) than what is installed. Not always required but manual
> builds give more control and easy to manage on some occurrences like in
> matplotlib build case.
> 

Yes, I meant the 'save' button in the upper right.

It used to work, but some update seems to have broken it.  Strange, so far 
the only application that shows a problem like this (and I do use a lot of 
different apps!)

On Sun, Apr 25, 2010 at 12:33 PM, Neal Becker <ndb...@gm...> wrote:

> python-matplotlib-0.99.1.2-1.fc12.x86_64
>
> When I hit 'print' button, I get a partially painted dialog.  Screenshot
> attached.  This is repeatable on 2 different machines.
>

I am on Fedora 12. Where is print button in matplotlib? You meant the save
dialog? If that's the case I don't see any problem using the svn version of
mpl.

I build sources manually ->
http://matplotlib.sourceforge.net/faq/installing_faq.html#install-from-svn

Similarly I build numpy and scipy from sources too. If available I use
easy_install for some packages, otherwise mostly manual builds. (Except the
Python itself.) This might be confusing since I use the package manager to
install dependencies most of the time. Once in a while I see some packages
requiring older numpy builds, also when I try to uninstall some packages via
package manager it tries to uninstall more packages (the critical ones) than
what is installed. Not always required but manual builds give more control
and easy to manage on some occurrences like in matplotlib build case.

-- 
Gökhan

python-matplotlib-0.99.1.2-1.fc12.x86_64

When I hit 'print' button, I get a partially painted dialog.  Screenshot 
attached.  This is repeatable on 2 different machines.

Eric, Tony, thanks for your replies. I'll look at SciPy then for this functionality.

Thanks,
--Michiel.

--- On Sat, 4/24/10, Eric Firing <ef...@ha...> wrote:

> From: Eric Firing <ef...@ha...>
> Subject: Re: [matplotlib-devel] Lowess smoothing
> To: "Michiel de Hoon" <mjl...@ya...>
> Cc: mat...@li...
> Date: Saturday, April 24, 2010, 2:29 PM
> Michiel de Hoon wrote:
> > Hi everybody,
> > 
> > A number of years ago I wrote a function to do Lowess
> smoothing to calculate a smooth curve through a scatter
> plot. I copied an example script below and attached the
> resulting figure to this mail.
> > I think that such a smoothing function would be a
> useful addition to matplotlib. Does anybody have any
> objections against me adding this to matplotlib? If not,
> what would be a suitable place to put this function?
> > 
> > --Michiel.
> > 
> 
> Michiel,
> 
> Certainly it would be good to have that function available,
> but I'm not in favor of putting it in mpl.  The trend
> has been to try to reduce overlap among mpl, numpy, and
> scipy.  To that end, scipy seems like the natural home
> for smoothing functions.
> 
> Eric
> 
> > 
> > from pylab import *
> > 
> > x = arange(0,10,0.01)
> > ytrue = exp(-x/5.0) + 2*sin(x/3.0)
> > 
> > # add random errors with a normal distribution 
>                
>     y = ytrue + normal(size=len(x))
> > scatter(x,y,color='cyan')
> > 
> > # calculate a smooth curve through the scatter plot
> > ys = smooth(x, y, 'lowess')
> > plot(x,ys,'red',linewidth=3)
> > 
> > # draw the true values for comparison
> > plot(x,ytrue,'green',linewidth=3)
> > 
> > 
> >       
> > 
> >
> ------------------------------------------------------------------------
> > 
> > 
> >
> ------------------------------------------------------------------------
> > 
> >
> ------------------------------------------------------------------------------
> > 
> > 
> >
> ------------------------------------------------------------------------
> > 
> > _______________________________________________
> > Matplotlib-devel mailing list
> > Mat...@li...
> > https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
> 
>