matplotlib-users Mailing List for matplotlib (Page 7)

Rafael Perez Pascual wrote:

>I have a Debian  Ubuntu 5.10 Breeze system in an amd64  machine  with
>python2.4. I install numpy, scipy and matplotlib without troubles, but
>when I treat to import pylab  I get a segmentation fault  and python 
>breaks.
>I treat lots of thinks but nothing works, some one have an idea  about
>these?
>
>Thanks
>
>Rafael Perez
>  
>

>How did you install those packages? What versions? What code did you run that
>segfaulted? Can you isolate a small, self-contained example that segfaults so we
>can examine it? If it's a numpy or scipy problem (likely), we will be happy to
>answer your questions on the appropriate list to avoid cluttering the matplotlib
>list.
>
>-- 
>Robert Kern
>rob...@gm...
>
>"In the fields of hell where the grass grows high
> Are the graves of dreams allowed to die."
>  -- Richard Harter
>
>
>  
>
I got scipy-0.4.4.tar.gz from the scipy site.
I got numpy-0.9.2.tar.gz from the numpy site.
I got matplotlib-0.86.2.tar.gz  from the matplotlib site.
Python2.4 and all other libs from the ubuntu repository.
I did the usualy 'python  setup.py  build'  and  'python  setup.py 
install' as root  in the  scipy-0.4.4 , numpy-0.9.2 and
matplotlib-0.86.2 directories.

 When I try to import pylab I get:

rpp@rafael:~$ python
Python 2.4.2 (#2, Sep 30 2005, 22:19:27)
[GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu8)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from numpy import *
>>> from scipy import *
Overwriting fft=<function fft at 0x2aaaad8879b0> from
scipy.fftpack.basic (was <function fft at 0x2aaaad5f0758> from
numpy.dft.fftpack)
Overwriting ifft=<function ifft at 0x2aaaad887a28> from
scipy.fftpack.basic (was <function inverse_fft at 0x2aaaad5f07d0> from
numpy.dft.fftpack)
>>> from matplotlib import *
>>> import pylab
Segmentation fault
rpp@rafael:~$


Another try

rpp@rafael:~$ python
Python 2.4.2 (#2, Sep 30 2005, 22:19:27)
[GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu8)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import matplotlib
>>> import pylab
Overwriting fft=<function fft at 0x2aaaaf186320> from
scipy.fftpack.basic (was <function fft at 0x2aaaad972ed8> from
numpy.dft.fftpack)
Overwriting ifft=<function ifft at 0x2aaaaf186398> from
scipy.fftpack.basic (was <function inverse_fft at 0x2aaaad972f50> from
numpy.dft.fftpack)
Segmentation fault
rpp@rafael:~$
 
Scipy whithout ploting works fine and very fast, but allways I try to
import  pylab I get the segmentation fault.

Thak's a lot Robert

Rafael perez

Hi list,
Recently, a VISA library was released (http://pyvisa.sourceforge.net/),
and I would like to get it and matplotlib working nicely together.=20
VISA is a generic interface for data logging from scientific instruments.
Currently, I'm running these things from an ipython window by exec'ing a
file object, which refers to some data taking script.

Unfortuantely, there seem to be some circumstances under which the plot
window will not update. Often (awlays?) it has stopped updating after som=
e
ui event has occured in the plot window. I've checked the pages on
"interactive plotting", and tried wxagg and tkagg backends, but to no ava=
il.

Hopefully this is a stupid, easily answered question, as I'd rather get o=
n
with my experiments than trawl through code, but it would be great to get=
 this
working. I feel that python+numpy+matplot+VISA may be a far superior alte=
rnative
to the frequently used Labview commerical software, and any
hints that would save me from reinventing the wheel would be much appreci=
ated.

BTW: Developers, thank you indeed for matplotlib. It's one of the great
libraries for a great language.

Marc

Rafael Perez Pascual wrote:
> I have a Debian  Ubuntu 5.10 Breeze system in an amd64  machine  with
> python2.4. I install numpy, scipy and matplotlib without troubles, but
> when I treat to import pylab  I get a segmentation fault  and python 
> breaks.
> I treat lots of thinks but nothing works, some one have an idea  about
> these?

How did you install those packages? What versions? What code did you run that
segfaulted? Can you isolate a small, self-contained example that segfaults so we
can examine it? If it's a numpy or scipy problem (likely), we will be happy to
answer your questions on the appropriate list to avoid cluttering the matplotlib
list.

-- 
Robert Kern
rob...@gm...

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter

I have a Debian  Ubuntu 5.10 Breeze system in an amd64  machine  with
python2.4. I install numpy, scipy and matplotlib without troubles, but
when I treat to import pylab  I get a segmentation fault  and python 
breaks.
I treat lots of thinks but nothing works, some one have an idea  about
these?

Thanks

Rafael Perez

>>>>> "Charlie" == Charlie Moad <cw...@gm...> writes:

    Charlie>      You could use html image maps to accomplish this.
    Charlie> Obviously this would be in the context of a web page.  I
    Charlie> think frontpage or dreamweaver has some nice tools for
    Charlie> creating these.  Just google "html image maps" for more
    Charlie> info.

Andrew Dalke has a tutorial on creating html image maps with
matplotlib

http://www.dalkescientific.com/writings/diary/archive/2005/04/24/interactive_html.html

JDH

     You could use html image maps to accomplish this.  Obviously this
would be in the context of a web page.  I think frontpage or
dreamweaver has some nice tools for creating these.  Just google "html
image maps" for more info.

On 2/15/06, Julius Lucks <jul...@gm...> wrote:
> Is it possible to embed clickable URLs, associated with data points or
> plot features, into PNGs or JPEGs produced by matplotlib?
>
> Thank you for your help,
>
> Julius
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: Splunk Inc. Do you grep through log fi=
les
> for problems?  Stop!  Download the new AJAX search engine that makes
> searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
> http://sel.as-us.falkag.net/sel?cmdlnk&kid=103432&bid#0486&dat=121642
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>

Is it possible to embed clickable URLs, associated with data points or
plot features, into PNGs or JPEGs produced by matplotlib?

Thank you for your help,

Julius

On Feb 15, 2006, at 10:42 AM, J.B...@if... wrote:

> I 've to change line 736 in mplot3d.py too but then it works...
>
> Line 736:
>
> - if z:
> + if z is None: --> correct or not?

Your logic is backwards, I think.  This is what I think is correct:

     def auto_scale_xyz(self, X,Y,Z=None,had_data=None):
         x,y,z = X,Y,Z
         try:
             x,y = X.flat,Y.flat
             if Z is not None:
                 z = Z.flat
         except AttributeError:
             pass

         self.xy_dataLim.update_numerix(x, y, not had_data)
         if z is not None:
             self.zz_dataLim.update_numerix(z, z, not had_data)
         self.autoscale_view()

The PNG I get from python surface.py is attached.

Cheers,

-- Paul

--
Dr. Paul S. Ray              E-mail: Pau...@nr...
Naval Research Laboratory    WWW   : http://xweb.nrl.navy.mil/ 
personnel/paulr/
Code 7655                    Phone : (202) 404-1619
Washington, DC 20375         AIM   : NRLPSR

Hi,

>With that change, the demos both work for me.

I 've to change line 736 in mplot3d.py too but then it works...

Line 736:

- if z:
+ if z is None: --> correct or not?

>"python surface.py" makes a nice surface plot.

No, for me it looks rather strange...

Thanks

Jens

On Feb 15, 2006, at 7:12 AM, J.B...@if... wrote:

> for a better understanding please could someone explain this for  
> the actual
> problem?

Arrays are no longer allowed as truth values in numpy, because the  
"truth" of an array (except one with exactly one element) is ambiguous.

> It was about simple surface plots in matplotlib... I took the 3D  
> package
> from  http://matplotlib.sourceforge.net/mpl3d.zip and I changed the
> following in 'mplot3d.py':
>
> line 729:
>
> - if Z:
> + if Z is None: --> correct or not?

Yes, that is corect.

> Now I get this 'The truth value of an array with more than one  
> element is
>   ambiguous. Use a.any() or a.all()' ValueError in line 680 ...
>
> line 680:
>
> - M = M or self.M
> + ???

I changed it to this:
     def tunit_cube(self,vals=None,M=None):
         if M is None:
             M = self.M
         xyzs = self.unit_cube(vals)

With that change, the demos both work for me.
"python surface.py" makes a nice surface plot.

Now, my question is: Is "mplot3d" planned for inclusion in future  
matplotlib releases?  It does exactly what I need for 3d surface  
plots without the huge baggage and complexity of the VTK-based  
solutions that allow interactive 3D and all that stuff I don't need.

Cheers,

-- Paul

--
Dr. Paul S. Ray              E-mail: Pau...@nr...
Naval Research Laboratory    WWW   : http://xweb.nrl.navy.mil/ 
personnel/paulr/
Code 7655                    Phone : (202) 404-1619
Washington, DC 20375         AIM   : NRLPSR

Hi,

for a better understanding please could someone explain this for the actual
problem?

It was about simple surface plots in matplotlib... I took the 3D package
from  http://matplotlib.sourceforge.net/mpl3d.zip and I changed the
following in 'mplot3d.py':

line 729:

- if Z:
+ if Z is None: --> correct or not?

Now I get this 'The truth value of an array with more than one element is
  ambiguous. Use a.any() or a.all()' ValueError in line 680 ...

line 680:

- M = M or self.M
+ ???

So, I do understand that the 'or' is the problem here but how would you
replace it???

Cheers

Jens

PS.: Maybe someone just send me a 'mplot3d.py' running under numpy... 

Curtis,

I have added support for the extent kwarg in the case with 
origin=None--that is, the case in which contour is being used alone, not 
superimposed on an image.

The intention is that extent = (x0,x1,y0,y1) with origin = None will 
make (x0,y0) the coordinates of Z[0,0] and (x1,y1) the coordinates of 
Z[-1,-1].  In other words, it provides an easy way to generate 
equally-spaced coordinates specified by the end points, and 
corresponding in a simple and unambiguous way to the Z array.  Beyond 
that, the orientation of the axes is determined in the usual way. This 
seems logical to me.



Curtis Cooper wrote:
>>But it does.  The X and Y arrays give the coordinates of the Z points.
>>If you want to reverse the Y axis, say to have Y increase downward, then
>>  you do something like this:
>>
>>a = gca()
>>a.set_ylim(a.get_ylim()[::-1])
>>
>>There are many other ways to do it, but the point is that the axes are
>>drawn with the first ylim value at the bottom and the second at the top,
>>  so to reverse the axis you simply reverse the order of the y limits.
>>Similarly for the x-axis.
>>
> 
> 
> Fair enough, but notice that if I use imshow(Z, extent=[xmin, xmax,
> ymax, min]), it displays the image with the ymax extent at the bottom and
> ymin at the top of the y-axis, as I intended.  For contour, the extent is
> always displayed in increasing order in each axis.
> 


With respect to the comparison to imshow: is there any circumstance 
under which the plotting the same data with the same origin and extent 
options on the same axes (as in contour_image.py) results in a 
mis-registration--any sort of mismatch between the image view and the 
contour view?

I still can't find anything wrong; but if you send me a script that 
illustrates incorrect behavior, I will certainly look at it.

Thanks.

Eric

>>>>> "Chris" == Chris Seberino <seb...@sp...> writes:

    Chris> I know this matplotlib code works.  I'm having trouble
    Chris> running it in a CGI script.  It appears some of modules
    Chris> matplotlib imports need certain X Window permissions or
    Chris> DISPLAY variable set?....

It looks like mpl is trying to load its default backend (eg GTKAgg)
and in CGI mode you'll likely want to use an image backend (eg Agg).
See the FAQ http://matplotlib.sourceforge.net/faq.html#APPSERVER and
links therein.  In particular, you'll likely want to set the backend
with

  backend : Agg

in your matplotlib rc file -- http://matplotlib.sf.net/matplotlibrc
or else in your CGI script with the *use* directive , eg

  import matplotlib
  matplotlib.use('Agg')
  import pylab

matplotlib.use must be called *before* importing pylab, and overrides
any rc or command line backend settings.

JDH

I know this matplotlib code works.  I'm having trouble
running it in a CGI script.  It appears some of modules
matplotlib imports need certain X Window permissions or=20
DISPLAY variable set?....

[seberino root /var/www] # tail /var/log/apache2/error.log
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1]
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1] import gtk; gdk =3D
gtk.gdk
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1]   File
"/usr/lib/python2.3/site-packages/gtk-2.0/gtk/__init__.py", line 37,
in ?
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1]
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1] from _gtk import *
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1] RuntimeError
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1] :
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1] could not open
display
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1]
[Tue Feb 14 17:07:30 2006] [error] [client 127.0.0.1] Premature end of
script headers: create_ssg.py

Robert Kern wrote:
> George Nurser wrote:
> 
> 
>>It still seems to me a real 'gotcha' that if Z: doesn't work. Is  there
>>any possibility of it being allowed to work in numpy?
> 

As Robert said, this was changed intentionally.  The construct if Z: is 
usually used to determine if an argument was passed in.  It is usually 
better to replace this construct with if Z is None: anyway.

The major problem is that the 'and' and the 'or' operations do "truth" 
value testing not "element-by-element" testing.  This is going to cause 
an algorithmic error (but not raise an exception) over and over again. 
It is better to tell the person they are being ambiguous and move on.

I know it can be a bit of a pain to convert code that uses this 
construct, sorry about that.  In the long run I think it will be better 
than having code that doesn't work because you forgot that 'and' and '&' 
  do two entirely different things.

-Travis

Darren,

I did a little more googling after my initial post and found that if  
I set:

font.latex.package : type1cm
text.usetex : true
text.tex.engine: latex

in my .matplotlibrc file the problem goes away. So I guess this is a  
false alarm, although I don't remember having problems with fonts in  
Latex in the past. Still, I think I've found my workaround. Thanks  
for your quick response, and apologies for not having done more  
research before posting.

Rick


On Feb 14, 2006, at 2:00 PM, Darren Dale wrote:

> Hi Rick,
>
>> Forgive me if this is a FAQ. I'm having trouble outputting fonts into
>> EPS files from matplotlib. If I plot a figure and then include the
>> figure in a LaTeX file, I get the figure minus the text in the xlabel
>> and ylabel fields. If I use Preview (I'm on a Mac) to show the
>> figure, the captions are there.
>>
>> Any idea what's wrong? Is there an easy workaround?
>
> I'm not sure. Maybe on of the Mac users can speak up, but in the  
> meantime,
> please make a barebones test script that will reproduce the  
> problem, run
> python test.py --verbose-debug, and post the output along with the  
> script
> and the resulting eps file. Also, please note any changes you have  
> made to
> your rc settings (for example, are you using the usetex option,  
> have you
> enabled ps.useafm, etc), and what version of mpl you are using.
>
>

Rick Muller
rm...@sa...

Hi Rick,

> Forgive me if this is a FAQ. I'm having trouble outputting fonts into
> EPS files from matplotlib. If I plot a figure and then include the
> figure in a LaTeX file, I get the figure minus the text in the xlabel
> and ylabel fields. If I use Preview (I'm on a Mac) to show the
> figure, the captions are there.
>
> Any idea what's wrong? Is there an easy workaround?

I'm not sure. Maybe on of the Mac users can speak up, but in the meantime,
please make a barebones test script that will reproduce the problem, run
python test.py --verbose-debug, and post the output along with the script
and the resulting eps file. Also, please note any changes you have made to
your rc settings (for example, are you using the usetex option, have you
enabled ps.useafm, etc), and what version of mpl you are using.

Forgive me if this is a FAQ. I'm having trouble outputting fonts into  
EPS files from matplotlib. If I plot a figure and then include the  
figure in a LaTeX file, I get the figure minus the text in the xlabel  
and ylabel fields. If I use Preview (I'm on a Mac) to show the  
figure, the captions are there.

Any idea what's wrong? Is there an easy workaround?

Thanks in advance,

Rick

Rick Muller
rm...@sa...

> But it does.  The X and Y arrays give the coordinates of the Z points.
> If you want to reverse the Y axis, say to have Y increase downward, then
>   you do something like this:
>
> a = gca()
> a.set_ylim(a.get_ylim()[::-1])
>
> There are many other ways to do it, but the point is that the axes are
> drawn with the first ylim value at the bottom and the second at the top,
>   so to reverse the axis you simply reverse the order of the y limits.
> Similarly for the x-axis.
>

Fair enough, but notice that if I use imshow(Z, extent=[xmin, xmax,
ymax, min]), it displays the image with the ymax extent at the bottom and
ymin at the top of the y-axis, as I intended.  For contour, the extent is
always displayed in increasing order in each axis.

Hanno,

See examples/image_masked.py.  I think it does what you want.  If your 
version of mpl is recent enough to include this demo, then you are all set.

Eric

Hanno Klemm wrote:
> Hi there,
> 
> is there a possibility to use imshow() with masked arrays? The
> problem: I have a large matrix where no_data_values are denominated by
> a certain number. It now would be very easy to mask these values and
> let imshow for example display them in another color but not use them
> in e.g. normalising the color scale. 
> 
> I think I saw some other mechansm to do that in the cookbook but I'm
> not sure if it can do what I want and at the moment the cookbook (or
> rather the scipy server) seems to be down.
> 
> Best regards,
> Hanno
> 
> 

Curtis,

Curtis Cooper wrote:
>>Now I see what you mean.  The problem is that the extent kwarg works
>>only if the origin kwarg is also set and is not None.  I have changed
>>the docstring accordingly.  There was a reason I made that requirement,
>>but possibly it could be relaxed.  The alternative is to simply generate
>>X and Y arrays that specify the coordinates of your Z points, and then
>>call contour(X, Y, Z, ...).
>>
> 
> 
> I think extent should work even if origin is not set, but at least if the
> documentation is clear about this, users can figure out the right thing to
> do easily enough.

OK, probably I will make the change (to allow use of extent without 
origin) soon, but I did not want to do it in too much of a hurry.

> 
> The alternative you mention does not allow plotting of the y-axis in
> reverse.  For example, I often put pressure decreasing upward on the
> y-axis of my plots.  It would be nice if contour(X, Y, Z, ...) respected
> my choice of ordering.

But it does.  The X and Y arrays give the coordinates of the Z points. 
If you want to reverse the Y axis, say to have Y increase downward, then 
  you do something like this:

a = gca()
a.set_ylim(a.get_ylim()[::-1])

There are many other ways to do it, but the point is that the axes are 
drawn with the first ylim value at the bottom and the second at the top, 
  so to reverse the axis you simply reverse the order of the y limits. 
Similarly for the x-axis.

Eric

Ryan Krauss wrote:
> This is a common problem I have with switching code over to numpy.  If
> you are in the habit of using if myMatrix: as a way to check for an
> empty matrix, numpy now throws this error.  It could be replaced with
> myMatrix.any(), but then I think there are problems if you use the
> same code under Numeric or numarray.

sometrue() exists as a ufunc in both.

-- 
Robert Kern
rob...@gm...

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter

George Nurser wrote:

> It still seems to me a real 'gotcha' that if Z: doesn't work. Is  there
> any possibility of it being allowed to work in numpy?

We've had a long discussion about this. No.

http://www.scipy.net/pipermail/scipy-dev/2005-November/thread.html#3799

In short, that "if Z:" worked in Numeric was a *huge* gotcha that got almost
everyone who ever used that feature.

-- 
Robert Kern
rob...@gm...

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter

On 14 Feb 2006, at 16:59, Ted Drain wrote:

> FYI: I believe the best way to check for None is to use:
>
> if Z is None
>
> or
>
> if Z is not None
>
> This eliminates the possibility of Z implementing the __cmp__  
> method which would get called if you use "if Z == None".  I only  
> know this after wasting several hours trying to figure out why some  
> code that looked perfectly reasonable wouldn't work after I


I've fallen into this trap too. Using if Z is None seems to be the  
*only* sensible way with numpy; it was explained in a mail of Robert  
Kern's last week.

It still seems to me a real 'gotcha' that if Z: doesn't work. Is  
there any possibility of it being allowed to work in numpy?

--George.

Dan Christensen <jd...@uw...> writes:

> I'd like to plot a large number of points (say 10**6 or 10**7) with
> different colours.  The colours will not naturally be parametrized by
> a single float, but rather the rgb components will all vary
> independently, so I don't think I can use a colormap.

I ended up not using matplotlib for this, but for the sake of anyone
with a similar question, I'll briefly say what I did:  I had my python
program write out raw postscript.  This turned out to be extremely
efficient in both time and space, and allows very high-resolution
printouts from a relatively small postscript file.  For example, one
plot with 13.5 million points is stored in an uncompressed postscript
file which is only 24MB, less than 2 bytes per point (and the points
vary in x, y coordinates, R,G,B colours and radii).  I'm able to do
this by taking into account (non-linear) symmetries of the plot
(making the postscript code do some of the calculations) and by taking
into account patterns in the colours and radii.

I'd still love it if matplotlib's scatter command could handle small
round dots more efficiently, since I use it for most of my other work,
but for now the above is working for me.

Dan