matplotlib-users Mailing List for matplotlib

Good day all,

I've hit an infinite loop/hang using the delaunay package.   Unfortunately,
I haven't been able to reproduce the issue with a nice minimal dataset, so I
don't expect much help, but I did want to note it in case.

I've been working with the delaunay package, making contour plots from
mostly unstructured data.  Unfortunately, I cannot use the natgrid version
due to licensing issues, so I've been using the included delunay with mixed
results. 

I've attached a .txt file containing the data values and x,y arrays used to
create the triangulation as well as the x,y points for the grid to use for
the resulting interpolation/extrapolation results. Listing of pertinent code
is below. 

The hang occurs in the call:
        z = interp_extrap( x2, y2 )
     which, in this case is a call to the triangulations nn_interpolator(
x2, y2 )

If I use the nn_extrapolator, the hang doesn't happen, but there is a nasty
artifact.

The size of the resulting grid does not appear to matter ( so I made it
small for simplicity ).

I assume that I'm the one doing something wrong, so if there are better
approaches to this problem, don't hesitate to correct me (  please :)  ). 

--See output in attached text file.
CODE:
        print 'data values:'
        print list( data_values )
        print 'data x:'
        print list( x_data_points )
        print 'data y:'
        print list( y_data_points )

        print 'start tri'
        # triangulate data (beware this can get cranky if given concave
shape)
        try:
            ltri = Triangulation( array( x_data_points ), array(
y_data_points ) )
        except Exception, err:
            raise Exception, ( 'Triangulation failed ( %s )' % err )
        print 'end tri'

        print 'start interp/extrap'
        # it seems we want the extremes of the x and y, but double check
        bbox = ( x_model_points_sorted[ 0 ], x_model_points_sorted[ -1 ], \
         y_model_points_sorted[ 0 ], y_model_points_sorted[ -1 ] )

        print 'bounding box:'
        print bbox

        if extrapolate:
            # extrapolate
            print 'extrap'
            try:
                # bounding out to region maximums for extrapolation - then
nn interpolation is applied
                interp_extrap = ltri.nn_extrapolator( array( data_values ),
bbox = bbox )
            except Exception, err:
                raise Exception, ( 'Extrapolation/Interpolation failed ( %s
)' % err )
        else:
            # interpolate
            print 'interp'
            try:
                interp_extrap = ltri.nn_interpolator( array( data_values ) )
            except Exception, err:
                raise Exception, ( 'Interpolation failed ( %s )' % err )
        print 'end interp/extrap'

        # OVERRIDE SMOOTHNESS FOR TESTING:
        self.smoothness = 0
        num_grid_points = int( self.smoothness * 200 )
        if num_grid_points == 0:
            num_grid_points = 20

        if len( x_model_points_sorted ) > num_grid_points:
            x_model_points_sorted = linspace( x_model_points_sorted[ 0 ],
x_model_points_sorted[ -1 ], num_grid_points )

        if len( y_model_points_sorted ) > num_grid_points:
            y_model_points_sorted = linspace( y_model_points_sorted[ 0 ],
y_model_points_sorted[ -1 ], num_grid_points )

        x2,y2 = meshgrid( x_model_points_sorted, y_model_points_sorted ) #
model boundary region

        print 'do the i/e'
        print 'x2 shape and y2 shape:'
        print x2.shape, ' ', y2.shape
        print 'x, grid points:'
        print x_model_points_sorted
        print 'y, grid points:'
        print y_model_points_sorted
        # run the interp/extrap step out to the bounded region on the grid
        z = interp_extrap( x2, y2 ) # HANG ON nn_interp, but not nn_extrap!
        print 'done with i/e'
:END CODE


Python version: 2.6.4
MPL version: 0.99.1
Windows XP Pro SP3

Thanks for reading,
-Erik Schweller

http://old.nabble.com/file/p27026918/boom_mpl_delaunay_output.txt
boom_mpl_delaunay_output.txt 
-- 
View this message in context: http://old.nabble.com/Delaunay-nn_interpolator-hang-tp27026918p27026918.html
Sent from the matplotlib - users mailing list archive at Nabble.com.

I'm happy for it to remain just a suggestion and not a reality. I 
mentioned it in case it was easy to implement alongside the color cycle 
but it seems it is not. Thanks for considering it anyway Eric,

Gary

Eric Firing wrote:
> Dominik Szczerba wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Eric Firing wrote:
>>> Dominik Szczerba wrote:
>>> OK I started hacking and added a color_cycle property to matplotlibrc.
>>> Would you be so kind to add this fix to the official version? Thanks!
>>> Dominik

Dear all,

I have a little problem on understanding how to move an annotation on a  
plot.

My idea is to make a plot where some informations follow the cursor when  
moving on the data, like you can see in plots from piwik (  
http://piwik.org/demo )

In the script in attachment, I can follow the cursor coordinate, change the  
text of the annotation, but I cannot move it.

I tried several things, but none worked.

If you have any advices, I would be very happy.

=============================
Script : try_move_annotate.py
=============================
import numpy
import matplotlib
import matplotlib.pyplot as plt

def on_motion(event):
if event.inaxes:
print "before setting position : "
print this_annotation.get_position()
this_annotation.set_text('coordinate : ' + str(event.x) + ' ' +  
str(event.y))
this_annotation.set_position((event.x, event.y))
print "after setting position : "
print this_annotation.get_position()
ax.draw_artist(this_annotation)
plt.draw()

fig = plt.figure()
ax = fig.add_subplot(111)

ax.plot(range(10), range(10))
this_annotation = ax.annotate("coordinate : ",
xy = (100,100),
xycoords = 'figure pixels',
horizontalalignment = 'left',
verticalalignment = 'top',
fontsize = 20,
fontweight = 'bold',# animated=True,
bbox = dict(boxstyle="round", fc='black', ec="0.5", alpha=0.5)
)

fig.canvas.mpl_connect('motion_notify_event', on_motion)
plt.show()
==========
END Script
==========

Hi All,
      I'm looking for a solution to properly display histograms of data
that is neither logarithmic, nor linear. Due to the nature of my data, I
manually specify my bins (which are neither fixed width, nor
logarithmic) - what I'd like to have is for each of my bins to have
equal bar widths.

      Basically the old 'deprecated' fixed bar widths would be exactly
what I need. Is there any possibility to mimic the old behavior using
the new API? If not, it would be nice if someone re-introduced the old
fixed-width functionality or some other solution since I guess I'm not
the only one plotting statistical data with manually specified bins. In
my case relative bar widths are an eye-sore (especially since my bins
have large orders of magnitude differences, but exactly some of the
smaller ones are of interest).

      Cheers,
            Greg

Dominik Szczerba wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Eric Firing wrote:
>> Dominik Szczerba wrote:
>> OK I started hacking and added a color_cycle property to matplotlibrc.
>> Would you be so kind to add this fix to the official version? Thanks!
>> Dominik
>>
>>> Your basic idea--that the colorcycle should be settable in 
>>> rcParams--makes good sense, but the implementation needs some changes, 
>>> maybe including a bit of redesign of the color cycle handling.  I will 
>>> look into it.  A little discussion on the devel list may be required.  I 
>>> think we will want to completely decouple lines.color from a new 
>>> lines.colorcycle, but maybe there is some good reason, other than 
>>> history, for why they are coupled.
> 
> Hi Eric,
> That's great. I would be happy if a final conclusion from the devel list
> could be forwarded here as well.
> Thanks,
> Dominik

Dominik,

I added rcParams support for the color cycle to svn, but a line style 
cycle remains at the level of a suggestion.

Eric