matplotlib-users Mailing List for matplotlib

Hello NG,

    I am probably missing something here, so sorry for the possible stupid
question. I have tried to use axes legend to create a legend on my plot, but
this doesn't work very well if I have 2 y-axis (as in two_scales.py demo),
because the legend is created as:

MyLeftAxis.legend()

And it doesn't see the properties plotted on the second axis. If I do:

MyRightAxis.legend()

I get 2 different legends, which I don't like very much. So I tried to use
figlegend, using MyFigure.legend(MyLines, MyLabel, Location), but I am
having 2 problems:

1) The use may decide to hide/show the legend as he/she wants, and for this
I am using legend.set_visible(), and it works almost smoothly;
2) I don't know how to update the legend as new plots are added to the
figure;
3) I don't know if there is a way to destroy a legend - completely delete it
from the figure.

Thank you very much for every hint.

Andrea.

"Imagination Is The Only Weapon In The War Against Reality."
http://xoomer.virgilio.it/infinity77

Hello John,

On Sun, Jan 29, 2006 at 10:58:50AM -0600, John Hunter wrote:
> Yes, that is annoying.  I just made some changes to CVS to fix it.  The
> root of the problem is that Agg does subpixel rendering.  I have some
> hacks now to make it work right for grid lines, ...
Sorry for the delay.  I tried it now and it does NOT work :-(
The diagram on the web page is now generated with the current
CVS version of matplotlib, but the problem still persists.

I uncommented the

    std::cout << "snapto "<<thisx<<" "<<thisy<<std::endl;

line in _backend_agg.cpp and found that this bit of code is only
called 142 times, which is significantly less than the number of
grid line fragments.

I tried to change (in RendererAgg::draw_lines) the test

    snapto =3D (x0=3D=3Dx1) || (y0=3D=3Dy1);

(looked dangerous to me, these are doubles) into

    snapto =3D fabs(x0-x1)<1e-6 || fabs(y0-y1)<1e-6 ;

but this did not help.

Any more ideas?

Many thanks,
Jochen
--=20
http://seehuhn.de/

Tom Denniston <tom...@al...> writes:

> Is there any easy way to make an area chart like the one attached in
> matplotlib without manually manipulating polygons?

Here's a snippet that makes a plot like that (but I haven't figured
out how to create the legend). To test it in ipython -pylab, type
something like

figure()
plot_cumulative(gca(), arange(20), rand(20,20))
show()

Perhaps a more refined version of this would be a useful addition to
matplotlib?

import Numeric as N
import matplotlib

# Colors from http://geography.uoregon.edu/datagraphics/color_scales.htm
colors = [
 [1.000, 0.500, 0.000],
 [1.000, 0.750, 0.500],
 [1.000, 1.000, 0.200],
 [1.000, 1.000, 0.600],
 [0.200, 1.000, 0.000],
 [0.700, 1.000, 0.550],
 [0.100, 0.700, 1.000],
 [0.650, 0.930, 1.000],
 [0.400, 0.300, 1.000],
 [0.800, 0.750, 1.000],
 [0.900, 0.100, 0.200],
 [1.000, 0.600, 0.750],
]

def plot_cumulative(ax, x, array):
    x, array = N.asarray(x), N.asarray(array)
    # Change e.g. array=[[1,2,3],[4,5,6]] to cs=[[0,0,0],[1,2,3],[5,7,9]]]
    cs = N.concatenate((N.zeros((1,len(x))), N.cumsum(array)))
    # Change e.g. x=[1,2,3] to xx=[1,2,3,3,2,1]
    xx = N.concatenate((x, x[::-1]))
    for i in range(0, N.size(cs,0)-1):
        # Define y coordinates of polygon as cs[i,:] and cs[i+1,:]
        yy = N.concatenate((cs[i,:], cs[i+1,::-1]))
        c = colors[i % len(colors)]
        ax.fill(xx, yy, facecolor=c)

-- 
Jouni

I did a fresh install on a new computer this week, and installed JDH's
Ubuntu packages from uchicago, and got the following error upon "from
pylab import *":

  File "/usr/lib/python2.4/site-packages/numpy/linalg/__init__.py",
line 4, in ?    from linalg import *
  File "/home/jdhunter/debs/numpy/usr/lib/python2.4/site-packages/numpy/lin=
alg/linalg.py",
line 16, in ?
ImportError: liblapack.so.2: cannot open shared object file: No such
file or directory

I do, however, have liblapack.so.3 available in /usr/lib. I tried
uninstalling and installing python-matplotlib-jdh and python-scipy
(because that had solved a different installation problem before) in
different orders, but no luck. So, I symlinked liblapack.so.3 to
liblapack.so.2, and then had to do the same to:
libatlas.so.3
libcblas.so.3
libf77blas.so.3

And "from pylab import *" works fine, and plotting seems to work fine.
However, it seems like unless version three of these libraries are
100% compatible, that this is problematic.

This is on a fresh install, and I'm not smart enough to figure out
anything else substantive, but wanted to let Matplotlib
maintainers/etc. know.

-Chris

I'd guess you'll want to plot against times converted using 
the datetime module
http://ftp.python.org/doc/lib/datetime-time.html
and then place your time strings as ticks.

Cheers,
Alan Isaac

I have used the plot function before to plot points for over a year lets 
say, now I have to write something that makes a plot with the current 
time of the day on the x-axis.  The data I am getting is all in military 
time, so it starts at 0700 and ends at 2200.  The data points are all in 
15 minute increments, so it goes 0700, 0715, 0730, 0745 etc.

I am having trouble using the multiple locator in this case.  I can't 
simply use 15, there would be a created gap in the graph.  For example, 
if I do a bar graph with bar(), i would see my data points and then big 
gaps.

So my question is, how can I have my X axis go in 15 minute increments 
stopping at 59?  ie, minutes in an hour only go from 0 to 59?  I would 
like to make a bar graph with 15 minute intervals, and also a plot graph 
with 15 minute intervals.

Thanks much,
mohan