matplotlib-users Mailing List for matplotlib

>>>>> "Randy" == Randy Heiland <he...@in...> writes:

    Randy> I'm just playing with matplotlib inside IPython for my
    Randy> first time and want to do the equivalent of:

    Randy> In [1]: subplot(211) In [2]: plot([1,2,3]) In [3]: show()
    -> displays as expected, in the upper half

Use ipython in the pylab mode -- it will detect your backend and do
the right thing if threading is required, and will set the matplotlib
interactive state to be True.

 > ipython --pylab

It will also set you to be in interactive mode, and no use of "show"
is required -- see http://matplotlib.sf.net/interactive.html and
http://matplotlib.sf.net/faq.html#SHOW.  

    Randy> Now I want to interactively display a 2nd plot in the
    Randy> bottom, but when I do: In [4]: subplot(212)

    Randy> I get a new, single empty plot displayed in the lower half.
    Randy> Is there some 'hold' cmd I'm missing to do this?

In matplotlibrc -- http://matplotlib.sf.net/.matplotlibrc -- there is
an axes.hold parameter.  matplotlib ships with axes.hold : True by
default, so this should be your setting unless you changed it in a
past life.  It may be that the strangeness you are seeing vis-a-vis
subplot(212) stems from not using interactive mode properly in
conjunction with unsupported use of show, which should not be used in
interactive mode as explained in the FAQ.

You can query the current hold state with "ishold" and set it with "hold".
Notice in the following pylab interactive session I don't need to
import pylab as ipython does it for me; just make sure you're using a
recent version of ipython and matplotlib.

John-Hunters-Computer:~> ipython --pylab
Python 2.3 (#1, Sep 13 2003, 00:49:11) 
Type "copyright", "credits" or "license" for more information.

IPython 0.6.6 -- An enhanced Interactive Python.
?       -> Introduction to IPython's features.
%magic  -> Information about IPython's 'magic' % functions.
help    -> Python's own help system.
object? -> Details about 'object'. ?object also works, ?? prints more.

  Welcome to pylab, a matplotlib-based Python environment
    help(matplotlib) -> generic matplotlib information
    help(pylab)     -> matlab-compatible commands from matplotlib
    help(plotting)   -> plotting commands

In [1]: ishold()
Out[1]: True

In [2]: hold(False)

In [3]: ishold()
Out[3]: False

In [4]: 

I'm just playing with matplotlib inside IPython for my first time and want
to do the equivalent of:

In [1]: subplot(211)
In [2]: plot([1,2,3])
In [3]: show()
-> displays as expected, in the upper half
Now I want to interactively display a 2nd plot in the bottom, but when I do:
In [4]: subplot(212)

I get a new, single empty plot displayed in the lower half.
Is there some 'hold' cmd I'm missing to do this?

--Randy

Hi:

The first version of a map plotting module is available at 
http://www.cdc.noaa.gov/people/jeffrey.s.whitaker/python/basemap/

(Source code
http://www.cdc.noaa.gov/people/jeffrey.s.whitaker/python/basemap/basemap-20041229.tar.gz)

It's not nearly as fully featured as the the matlab mapping toolbox, nor does it have the same API - but it does provide some useful functions for plotting images and contour plots on mercator, lat/lon, lambert conformal and stereographic projections.  Plotting coastlines, countries boundaries and state boundaries in the Americas is supported.  Filling continents is supported, but filling ocean areas is not (yet).  There are also functions to plot selected parallels and meridians, as well as an interpolation function to regrid regularly spaced lat/lon data to a native map projection grid.  

In addition to numarray and matplotlib, the module requires proj (from http://proj.maptools.org).

Hope someone finds it useful - comments and suggestions are welcome.

-Jeff

-- 
Jeffrey S. Whitaker         Phone : (303)497-6313
NOAA/OAR/CDC  R/CDC1        FAX   : (303)497-6449
325 Broadway                Web   : http://www.cdc.noaa.gov/~jsw
Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124

    Stephen> Sounds like an interesting project.  The first item I'd
    Stephen> have on a list of desired features is, surprise, the
    Stephen> ability to load multiple aligned images and blink between
    Stephen> them.

With the new keypress event handling in matplotlib-0.70, and the
cleanup to make sure the visible property is respected, this is pretty
easy.  The example below is more complicated than you need for the
usual case, since it handles images of different pixel dimensions that
occupy the same physical dimensions, but it gives you the idea (btw,
this is now examples/toggle_images.py, which contains a bit more
information in the header)

What's the second item on the list :-)

JDH

from pylab import *

# two images x1 is initially visible, x2 is not
x1 = rand(100, 100)
x2 = rand(150, 175)

# arbitrary extent - both images must have same extent if you want
# them to be resampled into the same axes space
extent = (0,1,0,1) 
im1 = imshow(x1, extent=extent)
im2 = imshow(x2, extent=extent, hold=True)
im2.set_visible(False)

def toggle_images(event):
    'toggle the visible state of the two images'
    if event.key != 't': return
    b1 = im1.get_visible()
    b2 = im2.get_visible()
    im1.set_visible(not b1)
    im2.set_visible(not b2)
    draw()

connect('key_press_event', toggle_images)
#savefig('toggle_images')
show()

This is (hopefully) a final rollup of what is (should be) a fairly
stable tree.  Mainly I wanted to send a release notification out to
the larger python community (scipy/numpy/python-list) since I haven't
done that since 0.60 (July 2004).  

For the major releases, I tend to try and shy away from new features
preferring a stable src.  But for you hardened, grizzled matplotlib
veterans, there are a few enticements nonetheless

 - pie charts and new example/pie_demo.py.  See
   http://matplotlib.sf.net/screenshots.html#pie_demo

 - support for object picking - see examples/picker_demo.py.  As
   people test this out and we settle on an interface, this will
   probably become part of the core, as will other keypress
   functionality for navigation, grid toogle, zoom toggle etc.

 - Fixed coords notification in wx toolbar 

 - key press and release event supported across backends -- see
   examples/keypress_demo.py

 - added Shadow patch class to provide a shadow effect for polygons,
   legends, pie charts -
   http://matplotlib.sf.net/screenshots.html#legend_demo

 - new examples/text_rotation.py demonstrates how text rotations and
   alignment work in matplotlib.  

Bug fixes

  - Fixed PS mathtext bug where color was not set

  - Fixed an agg text rotation alignment bug, fixed some text kwarg
    processing bugs

  - Refactored event handling - multiple connects and disconnects now
    work across backends.  See examples/coords_demo.py, especially
    with test_disconnect

  - Fixed a tkagg interactive bug that caused segfaults in some
    conditions.

As usual, downloads at http://sourceforge.net/projects/matplotlib

JDH

>>>>> "seberino" == seberino  <seb...@sp...> writes:

    seberino> Normally x-axis (horizontal) values increase to right
    seberino> and y-axis (vertical) values increase up the page.

    seberino> Is it easy/possible to reverse this for the y-axis??
    seberino> (So that y-values **decrease** up the page for some
    seberino> range?)

  ylim(ymax, ymin)

OR

  xlim(xmax, xmin)

See, for example, http://matplotlib.sf.net/examples/invert_axes.py

>>>>> "Randy" == Randy Heiland <he...@in...> writes:

    Randy> Can matplotlib do image plots of a nonuniform mesh?  from
    Randy> matplotlib.matlab import *

    Randy> x=array([0.,0.5,0.51,4.])  y=array([0.,1.])  X,Y =
    Randy> meshgrid(x,y) Z=array([[0,1,1,2],[0,1,1,2]]) im =
    Randy> imshow(Z,cmap=cm.jet) show()

Yes, but note that the pcolor has numrows-1 rows and numcols-1 cols
since the endpoints must be specified in your X,Y arrays

    x = array([0.,0.5,0.6,4.])
    y = array([0.,1.,4])
    X,Y = meshgrid(x,y)
    Z = array([[0,1,2,2],[0,1,1,2],[2,0,1,2]])
    pcolor(X, Y, Z, cmap=cm.jet)

JDH

Can matplotlib do image plots of a nonuniform mesh?

from matplotlib.matlab import *

x=array([0.,0.5,0.51,4.])
y=array([0.,1.])
X,Y = meshgrid(x,y)
Z=array([[0,1,1,2],[0,1,1,2]])
im = imshow(Z,cmap=cm.jet)
show()

thanks, Randy

Normally x-axis (horizontal) values increase to right
and y-axis (vertical) values increase up the page.

Is it easy/possible to reverse this for the y-axis??
(So that y-values **decrease** up the page for some range?)

Chris
--
_______________________________________

Christian Seberino, Ph.D.
SPAWAR Systems Center San Diego
Code 2872
49258 Mills Street, Room 158
San Diego, CA 92152-5385
U.S.A.

Phone: (619) 553-9973
Fax  : (619) 553-6521
Email: seb...@sp...
_______________________________________

>>>>> "seberino" == seberino  <seb...@sp...> writes:

    seberino> It seems if your grid has 20 rows ( different y values)
    seberino> of points that you won't get 20 colored horizontal
    seberino> strips but rather *19*.  This is because we must lose
    seberino> one row to specify the top and/or bottom EDGE of the
    seberino> plot.... so XX rows means (XX-1) rows of colored
    seberino> squares.

yep that's right, a frequent source of confusion.  

    seberino> Please tell me if this is right but more importantly,
    seberino> how to most wisely remove the white horizontal strip
    seberino> from this pcolor plot.

Well it would help if you posted your code, but my guess is that you
need to set your axis limits to equal your ymin/ymax of the pcolor.

Perhaps this example will give you a hint

   >>> pcolor(rand(10,7))  # no white strip...
   >>> ylim(0,11)          # a white strip because the ylim is wrong

In essence, the axis autoscaler will round up, eg 4990 to 5000, to
make nice integer ticks.  With pcolors and images, you often don't
want this, so use the axis, xlim, and/or ylim commands to set the
limits properly.

Eg

  ymin = min(ravel(Y))
  ymax = max(ravel(Y))
  ylim(ymin, ymax)

Also, if your x and y grids are equally spaced, you'll get much better
performance for large arrays, as well as more interpolation options,
if you use imshow with the extent kwarg to set the extent of your
image data.

JDH

I'm getting great pcolor plots like attached one and learning lots
about it.  So far so good.  One minor question if you don't mind...

I'm getting an empty white strip across the top of attached plot
as you can see which led me deeper into workings of pcolor function.

It seems if your grid has 20 rows ( different y values) of points
that you won't get 20 colored horizontal strips but rather
*19*.  This is because we must lose one row to specify the top
and/or bottom EDGE of the plot.... so XX rows means (XX-1) rows
of colored squares.

Please tell me if this is right but more importantly, how to
most wisely remove the white horizontal strip from this
pcolor plot.

Thanks!

chris

>>>>> "Eric" == Eric Emsellem <ems...@ob...> writes:

    Eric> Am I doing something wrong there?

No, this is a bug in the tick labeling / extent setting for aspect
preserved images - thanks for reminding me.  I had managed to suppress
it from my active memory!

JDH

Hi,

I am trying to display an image with imshow. What I am doing now is
to write a small function which plot the image preserving the aspect
ratio of the pixels in X and Y, but scaling the size of the image of the 
window
but at the same time scaling the axes either to just surround the image
(default) or to be larger.

This should be easy  as I can use 'preserve'
for the 'aspect' of imshow: that takes care of the pixel scale on the 
figure.

However then it does have axis which show incorrect coordinates, namely:
- along the direction where the image is filling the figure window better
 it shows the right coordinates
- along the other direction, the image is not filling the entire length
(as expected if the image does not follow the window aspect ratio)
but the axis are STILL showing the limits of the image.

To be clear:
 for  a square figure window.
 If my image is 100 x 50,
 then with aspect='preserve' I have the right coordinates along X
(showing 0==> 100) but the wrong coordinates for Y (showing 0=>50,
and NOT 0=>100 as it should since it fills only half the figure...).

I hope this is clear.

Am I doing something wrong there?

Thanks,

Eric E.

-- 
===============================================================
Observatoire de Lyon                 ems...@ob...
9 av. Charles-Andre                      tel: +33 4 78 86 83 84
69561 Saint-Genis Laval Cedex            fax: +33 4 78 86 83 86
France               http://www-obs.univ-lyon1.fr/eric.emsellem
===============================================================

>>>>> "seberino" == seberino  <seb...@sp...> writes:

    seberino> I installed Numeric and I can import it into Python but
    seberino> newest stable Matplotlib is barfing on it.  I installed
    seberino> Matplotlib after I installed Numeric so I'm not sure
    seberino> reinstalling it will fix this problem.  How fix?

Just to be sure


  * rm -rf your "build" subdir in the matplotlib directory

  * make sure you have Numeric installed for the same python you are
    building with.  Eg, specify the full path to python when testing
    and building

    > /usr/bin/python2.3
    >>> import Numeric
    >>> CTRL-D
    > sudo /usr/bin/python2.3 setup.py install >& build.out


  * make sure you choose numerix : Numeric  in .matplotlibrc

If you encounter more problems, post build.out - the incantation for
capturing standard output and standard error may differ depending on
your shell
   
Most of the times problems like this occur from distutils not properly
rebuilding when it should and can be solved by erasing its memory, ie
the "build" directory.

Let me know...

JDH

I installed Numeric and I can import it into Python but
newest stable Matplotlib is barfing on it.  I installed
Matplotlib after I installed Numeric so I'm not sure
reinstalling it will fix this problem.  How fix?

Chris

...Here is error output....:


The import of the numeric version of the _contour module,
_nc_contour, failed.  This is is either because numeric was
unavailable when matplotlib was compiled, because a dependency of
_nc_contour could not be satisfied, or because the build flag for
this module was turned off in setup.py.  If it appears that
_nc_contour was not built, make sure you have a working copy of
numeric and then re-install matplotlib. Otherwise, the following
traceback gives more details:

Traceback (most recent call last):
  File "test_campbell_diagram.py", line 10, in ?
    import campbell_diagram
  File
  "/home/seberino/MAIN/BusinessStuff/ThomasStuff/PlottingProject/CampbellDi
  agram/campbell_diagram.py", line 14, in ?
    import matplotlib.pylab
  File "/usr/lib/python2.3/site-packages/matplotlib/pylab.py", line
  184, in ?
    from axes import Axes, PolarAxes
  File "/usr/lib/python2.3/site-packages/matplotlib/axes.py", line 11,
  in ?
    import _contour
  File "/usr/lib/python2.3/site-packages/matplotlib/_contour.py", line
  11, in ?
    from matplotlib._nc_contour import *
ImportError: No module named _nc_contour

>>>>> "Eric" == Eric Emsellem <ems...@ob...> writes:

    Eric> Hi, working more and more with matplotlib and adapt my local
    Eric> tools I find that (for me) 2 plotting functionalities are
    Eric> still missing (but are in the list of the ''goals''):

    Eric> - plotting a surface either using a mesh/grid like plotwith
    Eric> colours or a smooth rendering - 

I am not familiar with "plotwith colours".  Is that a gnuplot command?

I assume you are referring to a 3D surf or mesh plot here, no?  The
official answer here has always been we're deferring any serious work
on 3D until we have a feature complete 2D lib in part because there
are already a number of excellent 3D tools for python.  There has been
some proof-of-concept work integrating VTK ( a premier 3D lib) with
matplotlib/agg -- you can find links on the goals page.

Although there are good tools already out there, I am receptive to the
argument that it's nice to have a core set of functionality under one
roof, and so would like to include support for basic 3D plots down the
road.  If some enterprising person wanted to take this on sooner, I
wouldn't be opposed.  But I can't give you an estimate on the time
frame right now.  Like much in the open source world, it's chronically
6 months away :-)

    Eric> have the connect function update the ''key'' field when
    Eric> using the keyboard

OK, to make up for the non-committal answer above, I implemented this
in CVS and it will be available in the next release.  Here's the
keypress_demo from CVS

from pylab import *

def press(event):
    if event.key=='g':
        grid()
        draw()
    
connect('key_press_event', press)

title('press g to toggle grid')
plot(rand(12), rand(12), 'go')
show()


In the process of getting this to work, I refactored the event
handling across backends in such a way that it is now robust and seems
to work well on all the backends, supporting multiple connects and
disconnects.  

JDH

>>>>> "Jeff" == Jeff Whitaker <js...@fa...> writes:

    Jeff> I'm currently working on a map plotting module.  I've added
    Jeff> the ability to plot filled continents on various map
    Jeff> projections, using the GSHHS coastline polygon dataset
    Jeff> (http://www.soest.hawaii.edu/wessel/gshhs/gshhs.html).  This
    Jeff> dataset provides polygon segments that define continental
    Jeff> outlines.  However, before I post it here I'd like to also
    Jeff> provide the capability to fill the 'wet' areas as well.  In
    Jeff> order to do this I need to figure out how to fill the area
    Jeff> between the polygons.  Looks like this module
    Jeff> (http://www.dezentral.de/soft/Polygon/index.html) would do
    Jeff> the trick, but it requires the generaly polygon clipping C
    Jeff> library.  Do you have any suggestions?

agg actually includes a wrapper for gpc, as well as the gpc code
itself, but gpc is GPLd and I haven't included it in matplotlib for
that reason.  Agg separately implements a "scanline boolean algebra"
which I haven't studied in detail but which I understand basically
does the same thing: allow clipping to arbitrary polygons, and
supports boolean operations on polygons --
http://www.antigrain.com/demo/index.html#PAGE_DEMO_scanline_boolean

We have to expose the functionality such that it is available for use
by other backends, much as we use agg for image across backends.

If this is an area that you want to dive into, by all means.
Otherwise, it's on the list of things to do (I just added it to the
goals page in my tree, so it will appear on the site docs in the
not-too-distant-future.

JDH

John Hunter wrote:

>>>>>>"Jeff" == Jeff Whitaker <js...@fa...> writes:
>>>>>>            
>>>>>>
>
>    Jeff> How do I suppress drawing a line around the polygon when
>    Jeff> using fill?  I've tried fill(x,y,'gray',linewidth=0), but I
>    Jeff> still get a little tiny line (which is especially noticeable
>    Jeff> when using the postscript backend).
>
>Just make the facecolor and edgecolor the same
>
> 
>
John:  Thanks - I figured that one out myself about 5 minutes after I 
sent the message.

I'm currently working on a map plotting module.  I've added the ability 
to plot filled continents on various map projections, using the GSHHS 
coastline polygon dataset 
(http://www.soest.hawaii.edu/wessel/gshhs/gshhs.html).  This dataset 
provides polygon segments that define continental outlines.  However, 
before I post it here I'd like to also provide the capability to fill 
the 'wet' areas as well.  In order to do this I need to figure out how 
to fill the area between the polygons.   Looks like this module 
(http://www.dezentral.de/soft/Polygon/index.html) would do the trick, 
but it requires the generaly polygon clipping C library.   Do you have 
any suggestions?

Thanks for you help,

-Jeff

-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/CDC  R/CDC1        Email  : Jef...@no...
325 Broadway                Web    : www.cdc.noaa.gov/~jsw
Boulder, CO, USA 80303-3328 Office : Skaggs Research Cntr 1D-124

>>>>>> "Jeff" == Jeff Whitaker <js...@fa...> writes:
>     Jeff> How do I suppress drawing a line around the polygon when
>     Jeff> using fill?  I've tried fill(x,y,'gray',linewidth=0), but I
>     Jeff> still get a little tiny line (which is especially noticeable
>     Jeff> when using the postscript backend).

On Tue, 28 Dec 2004, John Hunter apparently wrote:
> Just make the facecolor and edgecolor the same

I'll chime in here because I know someone is creating arrows
using polygons.  I have found that *any* width along the
edge is a problem for arrows: they must be *only* filled and
not stroked or they look like they point to the wrong point.
(Perhaps 0.3 point unscaled would not be a serious problem
for many uses, but remember there is a join at the tip!)
So: is it possible to fill a polygon without stroking the
edge (or to set the edge with to approximate 0)?

By way of comment: this was a real problem in gnuplot up to
version 3.8.  (I've been meaning to check whether it was
fixed in 4.0.)  The arrows were stroked as well as filled,
and they often looked very wrong.

fwiw,
Alan Isaac

Hi,

working more and more with matplotlib and adapt my local tools
I find that (for me) 2 plotting functionalities are still missing
(but are in the list of the ''goals''):

- plotting a surface either using a mesh/grid like plotwith colours  or 
a smooth rendering
- have the connect function update  the ''key'' field when using the 
keyboard

What is the present status/effort on these issues?

thanks!
Eric Emsellem

-- 
===============================================================
Observatoire de Lyon                 ems...@ob...
9 av. Charles-Andre                      tel: +33 4 78 86 83 84
69561 Saint-Genis Laval Cedex            fax: +33 4 78 86 83 86
France               http://www-obs.univ-lyon1.fr/eric.emsellem
===============================================================

>>>>> "Jeff" == Jeff Whitaker <js...@fa...> writes:

    Jeff> How do I suppress drawing a line around the polygon when
    Jeff> using fill?  I've tried fill(x,y,'gray',linewidth=0), but I
    Jeff> still get a little tiny line (which is especially noticeable
    Jeff> when using the postscript backend).

Just make the facecolor and edgecolor the same

  >>> fill(x,y, edgecolor='gray', facecolor='gray)

or whatever color you want them to be.  You can also use aliases

  >>> fill(x,y, ec='gray', fc='gray)

FYI, the new set/get introspection is designed to help you find these
things, by printing property names and the values they accept

In [3]: p, = fill(x,y)

In [4]: p
Out[4]: <matplotlib.patches.Polygon instance at 0x3b5ec60>

In [5]: set(p)
    alpha: float
    antialiased or aa: [True | False]
    clip_box: a matplotlib.transform.Bbox instance
    clip_on: [True | False]
    edgecolor or ec: any matplotlib color - see help(colors)
    facecolor or fc: any matplotlib color - see help(colors)
    figure: a matplotlib.figure.Figure instance
    fill: [True | False]
    label: any string
    linewidth or lw: float
    lod: [True | False]
    transform: a matplotlib.transform transformation instance
    visible: [True | False]
    zorder: any number


Hope this helps,
JDH

How do I suppress drawing a line around the polygon when using fill?

I've tried  fill(x,y,'gray',linewidth=0), but I still get a little tiny 
line (which is especially noticeable when using the postscript backend).

-Jeff

-- 
Jeffrey S. Whitaker         Phone : (303)497-6313
NOAA/OAR/CDC  R/CDC1        FAX   : (303)497-6449
325 Broadway                Web   : http://www.cdc.noaa.gov/~jsw
Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124

On Fri, 2004-12-24 at 09:21 -0500, Paul Barrett wrote:
> I therefore propose that we start developing a Python version of DS9. 
> The benefits of a Python version based on matplotlib are TrueType fonts 
> (with arbitrary text rotation), alpha blending, and direct support for 
> numarray.

Sounds like an interesting project.  The first item I'd have on a list
of desired features is, surprise, the ability to load multiple aligned
images and blink between them.

>>>>> "Haibao" == Haibao Tang <ba...@ug...> writes:

    Haibao> Hi, in some analysis, I really think it useful if you can
    Haibao> add a crosshair feature so I can visually align the
    Haibao> position of peaks and falls (like the stock market), and
    Haibao> may be not hard to include, too. Basically, if the
    Haibao> "crosshair" checkbox checked, a horizontal and a vertical
    Haibao> line will appear with the movement of the mouse.

It would be nice to have this feature efficiently across backends.
There is an example which shows you how to do this for an arbitrary
backend

  http://matplotlib.sf.net/examples/cursor_demo.py

but it can be slow because it redraws the entire figure every time you
move the cursor.  This is a general problem with animations in
matplotlib that we are thinking about - see for example
http://sourceforge.net/mailarchive/message.php?msg_id=10330431 .

There is also an example showing how to do this for WX/WXAgg (if
memory serves, you are using wx because it supports chinese
characters).  This uses WX to do the cursoring, which is much more
efficient because you don't have to redraw the figure

  http://matplotlib.sf.net/examples/wxcursor_demo.py

When time permits, it would be nice to add cursoring on each backend
with native drawing of the cursor for efficiency...

JDH

>>>>> "Istvan" == Istvan Szapudi <sz...@If...> writes:

    Istvan> Hi All, I am fairly new to matplotlib, and I am impressed
    Istvan> with its capabilities.

    Istvan> I have trouble plotting loglog plots with errorbars. I
    Istvan> have the following, program, a slightly modified version
    Istvan> of one given earlier by JDH on 9-28-2004 in this mailing
    Istvan> list:

This is definitely a gotcha that matplotlib needs to be smarted about
handling.  The log zero error is occurring when the transformation is
made on the axes limits and not on the data.  The autoscaler picks the
best min/max for the data coordinates, and will round down or up to
facilitate nice integer ticking.  When you plot with linear coords,
the autoscaler makes it's pick according to linear scaling, and then
when you change scales the old scaling is in effect and the log
transform fails when converting the viewport.

Solution: rescale the axes after changing coords, 
either manually

  ax.set_yscale("log")
  ax.set_xscale("log")
  axis([0.5*min(x), 2*max(x), 0.5*min(y-err), 2*max(y+err)])

or use the autoscaler

  ax.set_yscale("log")
  ax.set_xscale("log")
  ax.autoscale_view()

or set your log coords *before* calling plot

  ax = gca()
  ax.set_yscale("log")
  ax.set_xscale("log")
  errorbar(x,y,err,fmt='o')
  show()

and then the errorbar command will pick a "locator" to handle ticking
and viewport scaling appropriately from the outset.  This is the
approach taken in http://matplotlib.sf.net/examples/log_bar.py .

Until I get this fixed to work automagically, I'll make it a FAQ.

JDH