matplotlib-users Mailing List for matplotlib

 I  guess  it  was  a  FAQ  :) 

The  correct  way  -  after  experimenting  with  imshow  options  -  is  to  transpose  the  data  and  use  origin='lower'  option 

 

imshow(c.T,interpolation='nearest',origin='lower') 

 

Piotr 

 

 

----- Original Message ---- 

From: Piotr Zolnierczuk <zo...@ya...> 

To: mat...@li... 

Sent: Friday, February 22, 2008 2:31:08 PM 

Subject: [Matplotlib-users] imshow: axis not natural 

 

 Hi, 

it  is  probably  a  faq  -  but  I  could  not  find  it  anywhere. 

 

I  would  like  to  plot  a  2D  distribution  (2D  histogram)  of  some  data  using  a  2D  numpy.array  that  contains  my  data:  c[ix,iy]  

When  I  plot  it  via  imshow  the  x  axis  points  downwards  and  the  y  axis  point  to  the  left. 

What  I  want  is  that  the  x-axis  points  to  the  left  and  the  y  axis  points  upwards. 

 

So  I  ended  up  writing  a  small  script  twodimshow   (below)  where  I  transpose  data  (x<->y)  and  also  flip  the  sign  of  the  y-axis 

 

Now:  is  there  a  better  way  of  doing  it?  

 

 

See  also  CERN  Root  C++  package  (ftp://root.cern.ch/root/doc/3Histograms.pdf)  and/or  (SLAC  HIPPO  (http://www.slac.stanford.edu/grp/ek/hippodraw/) 

 

-------------------  CODE  BELOW  --------------------------- 

import  numpy 

from  pylab  import  * 

 

def  convertToImShow(data,copy=True): 

      '''Converts  data  so  that  imshow  displays  x,y  as  a  normal  

      physicist  expects  ie  x  pointing  left  and  y  pointing  up''' 

      data   =  numpy.array(data.T,copy=copy) 

      if  data.ndim  !=  2: 

            raise  TypeError,'data  must  be  a  two-dimensional  matrix' 

      nx,ny  =  data.shape 

      for  i  in  range(nx/2): 

            j=  nx  -  i  -  1 

            for  k  in  range(ny):  data[j,k],data[i,k]=data[i,k],data[j,k] 

      return  data 

 

def  twodimshow(data,extent=None,aspect=None): 

      r  =  convertToImShow(data) 

      if  not  aspect  and  extent: 

            if  len(extent)>=4: 

                  aspect  =  (extent[1]  -  extent[0])/float(extent[3]  -  extent[2]) 

      imshow(r,extent=extent,aspect=aspect,interpolation='nearest') 

 

if  __name__  ==  "__main__": 

      c  =  zeros((4,4)) 

      for  ix  in  range(4): 

            for  iy  in  range(ix): 

                  c[ix,iy]=ix+iy 

 

      figure() 

      subplot(221) 

      imshow(c     ,interpolation='nearest') 

      title('Axis  direction  not  natural') 

      xlabel('Y-Axis');  ylabel('X-Axis') 

 

      subplot(222) 

      imshow(c.T  ,interpolation='nearest') 

      title('Not  there  yet') 

      xlabel('X-Axis');  ylabel('Y-Axis') 

 

      subplot(223) 

      twodimshow(c,extent=(0,4,0,4)) 

      title('That  is  what  I  want') 

      xlabel('X-Axis');  ylabel('Y-Axis') 

      show() 

 

 

 

 

 

         ____________________________________________________________________________________ 

Never  miss  a  thing.   Make  Yahoo  your  home  page.  

http://www.yahoo.com/r/hs 

 

------------------------------------------------------------------------- 

This  SF.net  email  is  sponsored  by:  Microsoft 

Defy  all  challenges.  Microsoft(R)  Visual  Studio  2008. 

http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ 

_______________________________________________ 

Matplotlib-users  mailing  list 

Mat...@li... 

https://lists.sourceforge.net/lists/listinfo/matplotlib-users 

 

 







      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page. 
http://www.yahoo.com/r/hs

Hi,
it is probably a faq - but I could not find it anywhere.

I would like to plot a 2D distribution (2D histogram) of some data using a 2D numpy.array that contains my data: c[ix,iy] 
When I plot it via imshow the x axis points downwards and the y axis point to the left.
What I want is that the x-axis points to the left and the y axis points upwards.

So I ended up writing a small script twodimshow  (below) where I transpose data (x<->y) and also flip the sign of the y-axis

Now: is there a better way of doing it? 


See also CERN Root C++ package (ftp://root.cern.ch/root/doc/3Histograms.pdf) and/or (SLAC HIPPO (http://www.slac.stanford.edu/grp/ek/hippodraw/)

------------------- CODE BELOW ---------------------------
import numpy
from pylab import *

def convertToImShow(data,copy=True):
    '''Converts data so that imshow displays x,y as a normal 
    physicist expects ie x pointing left and y pointing up'''
    data  = numpy.array(data.T,copy=copy)
    if data.ndim != 2:
        raise TypeError,'data must be a two-dimensional matrix'
    nx,ny = data.shape
    for i in range(nx/2):
        j= nx - i - 1
        for k in range(ny): data[j,k],data[i,k]=data[i,k],data[j,k]
    return data

def twodimshow(data,extent=None,aspect=None):
    r = convertToImShow(data)
    if not aspect and extent:
        if len(extent)>=4:
            aspect = (extent[1] - extent[0])/float(extent[3] - extent[2])
    imshow(r,extent=extent,aspect=aspect,interpolation='nearest')

if __name__ == "__main__":
    c = zeros((4,4))
    for ix in range(4):
        for iy in range(ix):
            c[ix,iy]=ix+iy

    figure()
    subplot(221)
    imshow(c   ,interpolation='nearest')
    title('Axis direction not natural')
    xlabel('Y-Axis'); ylabel('X-Axis')

    subplot(222)
    imshow(c.T ,interpolation='nearest')
    title('Not there yet')
    xlabel('X-Axis'); ylabel('Y-Axis')

    subplot(223)
    twodimshow(c,extent=(0,4,0,4))
    title('That is what I want')
    xlabel('X-Axis'); ylabel('Y-Axis')
    show()





      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page. 
http://www.yahoo.com/r/hs

Hi,

what is the best way to use a Formatter for contour labels (clabel)
instead of using simple formatting ('%g')?
Should we derive a class from ContourLabeler and do things by hand?

It would be useful for example for showing labels from
power(10.,level) and you plot logarithmic values.

Thanks for the help.

-- 
Stephane Raynaud

Hi all,

In my latest post, I wanted to use the mpl.hist() function in a different way, i.e.:

x = datalist
bins= 100
hist(x,bins,normed=0)     #returns a tupple (n,bins,patches)

Instead of ploting the number of counts n, I wanted to plot the relative percentage of counts, i.e. n/len(x). I can't really use the option normed=1 which returns n/(len(x)*dbin). In the axes.py module, this would simply mean adding an argument e.g. relpercent = 1. I added the code line to show how this could be done (in major cap). If this is useful, how could it be modified in the distribution ?


    def hist(self, x, bins=10, RELPERCENT = 1, normed=0, bottom=None,
             align='edge', orientation='vertical', width=None,
             log=False, **kwargs):
        """

        if not self._hold: self.cla()
        n, bins = npy.histogram(x, bins, range=None, normed=normed)
         IF NOT NORMED AND RELPERCENT: N = N/FLOAT(LEN(X))
        if width is None: width = 0.9*(bins[1]-bins[0])
        if orientation == 'horizontal':
            patches = self.barh(bins, n, height=width, left=bottom,
                                align=align, log=log)
        elif orientation == 'vertical':
            patches = self.bar(bins, n, width=width, bottom=bottom,
                                align=align, log=log)
        else:
            raise ValueError, 'invalid orientation: %s' % orientation
        for p in patches:
            p.update(kwargs)
        return n, bins, cbook.silent_list('Patch', patches)





      _____________________________________________________________________________ 
Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail http://mail.yahoo.fr

On Fri, Feb 22, 2008 at 7:52 AM, Auré Gourrier
<aur...@ya...> wrote:

> Rather trivial... but instead of the plotting the counts n, I'd like to plot
> the realtive percentage counts, i.e. n/len(x). I can't really use the option
> normed = 1 which plots n/(len(x)*dbins). I guess the simplest way would be
> to simply change the yticklabels (by dividing them by len(x)). The thing is
> that I simply cannot find out how to do this...
>
> I tried using the axes.set_yticklabels() but doesn't work. I've also tried
> to find the child containing the label but couldn't find it (not in Axes,
> nor in YAxis etc...). I guess it must be a Text instance.

You can set your own custom tick formatter:

import matplotlib.ticker as ticker

N = len(x)
def fmt_percent(x, pos=None):
    return '%1.2f'%(float(x)/N)

ax.xaxis.set_major_formatter(ticker.FuncFormatter(fmt_percent)).  See
http://matplotlib.sourceforge.net/examples/custom_ticker1.py for a
complete example.

JDH

Thanks guys.

The env variable method doesn't work, but I can use the file in current dir
for now (will just symlink from the file i want).

Best Regards T.

On Fri, Feb 22, 2008 at 2:39 PM, Darren Dale <dar...@co...>
wrote:

> On Friday 22 February 2008 04:45:46 am Troels Kofoed Jacobsen wrote:
> > Hey everybody
> >
> > Is there at way to make matplotlib use a different matplotlibrc than the
> > one in $HOME/.matplotlib. This is because I want to be able to use
> > different settings for different plots, e.g one for publishable plots,
> and
> > one for daily plotting.
> >
> > Best regards
> >
> > Troels Kofoed Jacobsen
>
> Try saving a matplotlibrc file in your current working directory.
> Matplotlib
> looks there first.
>
> Darren
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>



-- 
Med Venlig Hilsen

Troels Kofoed Jacobsen

Hi Aure,

You could try:

ax.set_yticklabels(ax.get_yticks()/len(x))


/Antonio



Auré Gourrier wrote:
> Hi all,
> 
> I'm plotting the histogram of a data set:
> 
> x = datalist
> bins= 100
> hist(x,bins,normed=0)    #returns a tupple (n,bins,patches)
> 
> Rather trivial... but instead of the plotting the counts n, I'd like to 
> plot the realtive percentage counts, i.e. n/len(x). I can't really use 
> the option normed = 1 which plots n/(len(x)*dbins). I guess the simplest 
> way would be to simply change the yticklabels (by dividing them by 
> len(x)). The thing is that I simply cannot find out how to do this...
> 
> I tried using the axes.set_yticklabels() but doesn't work. I've also 
> tried to find the child containing the label but couldn't find it (not 
> in Axes, nor in YAxis etc...). I guess it must be a Text instance.
> 
> Can anyone give me a hint,
> 
> Cheers,
> 
> Aure
> 
> ------------------------------------------------------------------------
> Ne gardez plus qu'une seule adresse mail ! Copiez vos mails 
> <http://fr.rd.yahoo.com/mail/mail_taglines/trueswitch/SIG=11gshn0bu/**http%3A%2F%2Fwww.trueswitch.com%2Fyahoo-fr%2F> 
> vers Yahoo! Mail
> 
> 
> ------------------------------------------------------------------------
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Hi all,

I'm plotting the histogram of a data set:

x = datalist
bins= 100
hist(x,bins,normed=0)     #returns a tupple (n,bins,patches)

Rather trivial... but instead of the plotting the counts n, I'd like to plot the realtive percentage counts, i.e. n/len(x). I can't really use the option normed = 1 which plots n/(len(x)*dbins). I guess the simplest way would be to simply change the yticklabels (by dividing them by len(x)). The thing is that I simply cannot find out how to do this...

I tried using the axes.set_yticklabels() but doesn't work. I've also tried to find the child containing the label but couldn't find it (not in Axes, nor in YAxis etc...). I guess it must be a Text instance.

Can anyone give me a hint,

Cheers,

Aure




      _____________________________________________________________________________ 
Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail http://mail.yahoo.fr

On Friday 22 February 2008 04:45:46 am Troels Kofoed Jacobsen wrote:
> Hey everybody
>
> Is there at way to make matplotlib use a different matplotlibrc than the
> one in $HOME/.matplotlib. This is because I want to be able to use
> different settings for different plots, e.g one for publishable plots, and
> one for daily plotting.
>
> Best regards
>
> Troels Kofoed Jacobsen

Try saving a matplotlibrc file in your current working directory. Matplotlib 
looks there first.

Darren

Hey everybody

Is there at way to make matplotlib use a different matplotlibrc than the one
in $HOME/.matplotlib. This is because I want to be able to use different
settings for different plots, e.g one for publishable plots, and one for
daily plotting.

Best regards

Troels Kofoed Jacobsen

On Thu, 21 Feb 2008, John Hunter apparently wrote:
> you should not be using pylab, since you are embedding mpl in a 
> GTK GUI. 

Given the frequency of this question, maybe it would be 
a good idea on the main project page
<URL:http://matplotlib.sourceforge.net/>
to augment the text

        The plotting functions in the pylab interface have 
        a high degree of Matlab® compatibility.

to include a caveat:

        The plotting functions in the pylab interface have 
        a high degree of Matlab® compatibility.  (These are
        suitable for plotting scripts but NOT for GUI 
        embeddding, which should use the matplotlib API.)

fwiw,
Alan Isaac

On Thu, Feb 21, 2008 at 2:33 AM, Wolfgang Kerzendorf
<wke...@go...> wrote:
> Dear all,
>  I use the scisoft package on leopard 10.5.2 (it is a package that
>  creates its own python framework and delivers astronomical tools). I
>  have recompiled tcl tkk 8.4 and 8.5 (which in hindsight was probably

You probably have a few choices:

  * revert to a standard tk package to use the standard scisoft
compiled against it,

  * use a different backend for matplotlib -- eg download wxpython for
os x and then use the wxagg backend in matplotlib by setting the
backend variable in your rc file:
http://matplotlib.sf.net/matplotlibrc

  * compile matplotlib yourself on your platform against your new tk, see
      http://matplotlib.sourceforge.net/installing.html
      http://ipython.scipy.org/moin/Py4Science/InstallationOSX
      http://ipython.scipy.org/moin/MatplotlibOSXBuildNotes

Hope this helps,
JDH

On Thu, Feb 21, 2008 at 3:07 PM, Kevin Milner <km...@us...> wrote:
> Hi everyone,
>
>  I am a developer on a PyGTK application for Earth Science research and
>  education called SEATREE ( http://geosys.usc.edu/projects/seatree/wiki )
>  and we are trying to integrate some pylab plots.
>
>  I can successfully create and use a pylab frame/window inside of our GTK
>  application using the following sequence (note that all of this is
>  called AFTER the application has launched and the GTK main loop has
>  started):
>
>  -------------------------------------
>  import math
>  import pylab as p
>  import matplotlib
>  from matplotlib.backends.backend_gtk import FigureCanvasGTK,
>  NavigationToolbar
>  matplotlib.use('GTK')

For the "use" directive to be honored, it must be called before you
import pylab.  If it is after, as in your example code, pylab will
respect the setting in your matplotlibrc file -- see
http://matplotlib.sf.net/matplotlibrc (this is likely why you are
getting unexplained references to tk code).

But you should not be using pylab, since you are embedding mpl in a
GTK GUI.  Instead, you should be using the matplotlob API,
specifically the idioms for embedding mpl in gtk -- it takes a little
more work up front but it is the only acceptable way to use mpl in a
GTK app

http://matplotlib.sf.net/examples/embedding_in_gtk.py
http://matplotlib.sourceforge.net/examples/embedding_in_gtk2.py

See also the OO FAQ and the API tutorial for more insight into using
the API directly rather than the pylab interface

http://matplotlib.sf.net/faq.html#OO
http://matplotlib.sf.net/leftwich_tut.txt

JDH

JDH