matplotlib-users Mailing List for matplotlib

>>>>> "James" == James Boyle <bo...@ll...> writes:


    James> I have used nc = 3 or 4 if I use the plot command and all
    James> the plots are squished into the subregions for which I
    James> asked - even if things get ugly.

This is a one character bug!

I mainly use multiple rows in my own plots and never noticed it.  On
or around line 1492 in axes.Subplot.__init__

change
        figLeft = left + colNum*(figW + sepH)
                                           ^
to
        figLeft = left + colNum*(figW + sepW)
                                           ^

Thanks for the detailed bug report and example code.

JDH    

Dear John,

> Well, the behavior you are seeing is the correct, documented
> behavior. I think it would be easier for you to pass errorbar what it
> is expecting than for errorbar to handle this additional use case -
> it's already getting pretty hairy in there.
> 
> DO you have a problem doing
> 
> errorbar([1],[2],xerr=[.1])

Not really, it's just that I hate to think beforehand about whether a
routine eats everything or just arrays ;). The adjustments needed are
only minor I found out, after sending the previous mail. Here is the
diff against matplotlib-0.50e

<diff>

> 
> 7c7
<      log10, Float
---
> log10, Float, NewAxis
527a528,531
> if (len(x.shape) == 0): x=x[NewAxis,] if (len(y.shape) == 0): 
> y=y[NewAxis,]
533,534c537,544
<         if xerr is not None: xerr = asarray(xerr)
<         if yerr is not None: yerr = asarray(yerr)
---
> if xerr is not None: xerr = asarray(xerr) if (len(xerr.shape) == 0): 
> xerr=xerr[NewAxis,] if yerr is not None: yerr = asarray(yerr) if
> (len(yerr.shape) == 0): yerr=yerr[NewAxis,]

</diff>

> 
> You're not the first!  The temporary workaround is to use 
> plot(x,y,'o') which uses fixes sizes in points for the symbols (you
> can also use '*', '+', etc)  However, you cannot vary the size or
> color of the symbols so this may not be workable.

I'm perfectly happy with the plot command to do normal scatter plots. I
even think that it would be more logical (but perhaps not
matlab-compatible) to move all machinery of scatter to plot.

You asked for information of how people (would like to) use scatter
plots. Well, my wishlist for scatter would be more or less like (sorry
it's a long wishlist,  but see them as suggestions, not commands!):
* Scatter plots of symbols where the size of the symbols can be set per
   x-y pair. The use of this is that it depends on the number of points
   whether you want tiny dots, or large circles (the latter being nice
   when you have 10 points, but not when you have a million. For me,
   the size does not need to be specified in points, or another seemingly
   exact measure. I can live with a 'default size' (say 1/100 of the
   smallest of the axes) and a multiplier that can set by the user (like
   in Gnuplot).
* The symbols should be the same kind of symbols as used in plot (i.e.
   squares, triangles, circles etc.). The blob-symbols as plotted now
   should be a special case rather than the default. For myself, I only
   see a useful usage for *circular* (irrespective of axes) blobs  where
   the radius is set by the s-keyword. But  perhaps others like to have
   ellipsoid blobs that have a *two* axes that make  physical sense.
* It would be nice if all symbols -as far as applicable- come in 'open'
   and 'filled' versions (nice if you want to plot in one scatter plot
   data referring to the same physical quantity, but separating between
   daytime and nighttime for instance).
   For me, the number of symbols needed in one plot is rather limited: I
   hardly use more than 4 simultaneously in one scatter plot. On the
   other hand, in general x-y plots (lines, line-points, or points in
   Gnuplot terms) I would like to have at least say 10 symbols, since I
   usually use symbols to clarify the difference between lines, because
   that's easier than to ask the reader to differentiate between dash,
   dash-dot, dash-dash-dot etc.
* With respect to the coloring (c-keyword) it is perhaps easiest if this
   is applied to the symbols directly: just as one can say 'b>' in plot,
   it would be something like:
      scatter(x, y, '>', c=color_array)
* Assume that the new symbol for the 'blobs' would be say '@', then
      scatter(x, y, '@', s=size_array, c=color_array)
   gives colored blobs with unequal axes (as now),
      scatter(x, y, 'b@', s=size_array)
   gives blue blobs with unequal axes
      scatter(x, y, 'b@', s=size_array*-1)
   gives blue blobs with equal axes
   etc.

I hope this helps. And perhaps I can help in some other ways.

With kind regards,

Arnold Moene


> 
> When I refactor the scatter / pcolor etc commands as discussed in my
> response to Dominique Orban above, I'll make sure that this case is
> handled as well.  I plan to support efficient handling of scatter
> with arbitrary (or at least several) shapes where the size can be in
> data (x,y) or user-defined coords, eg, physical sizes.  It would be
> helpful for me if people gave some information about how they would
> like scatter to behave, particularly in regard to the sizes since I
> am still working this out.
> 
> Would you like physical sizes, eg circles with radius in points, or
> what?
> 
> Should the shape depend on the aspect ratio of the plotting window.
> Eg, if it's a circle and you have a short, wide window, should it
> appear ellipsoidal?
> 
> What basic scatter symbols should be supported?
> 
> What are the most common uses: fixed size?  fixed color?  vary both?
> 
> JDH
> 


-- 
------------------------------------------------------------------------
Arnold F. Moene                          NEW tel: +31 (0)317 482604
Meteorology and Air Quality Group            fax: +31 (0)317 482811
Wageningen University                    e-mail: Arn...@wu...
Duivendaal 2                             url: http://www.met.wau.nl 6701
AP  Wageningen
The Netherlands
------------------------------------------------------------------------
Openoffice.org - Freedom at work
------------------------------------------------------------------------






-- 
------------------------------------------------------------------------
Arnold F. Moene                          NEW tel: +31 (0)317 482604
Meteorology and Air Quality Group            fax: +31 (0)317 482811
Wageningen University                    e-mail: Arn...@wu...
Duivendaal 2                             url: http://www.met.wau.nl
6701 AP  Wageningen
The Netherlands
------------------------------------------------------------------------
Openoffice.org - Freedom at work
------------------------------------------------------------------------

I was making subplots using the bar command and ran into a problem I do 
not understand.

nr = number of rows, nc = number of columns

I  want nr = 1, i.e. just one row of plots. If  nc is 1 or 2 all is 
well, but nc =3 the third plot is cutoff in the middle. If nc = 4, I 
get 3 plots and the fourth cannot be seen.
What seems to be happening is that the subplots are not scaled to fit 
the viewport dictated by suplot(nr,nc,plotNumber).

I have used  nc = 3 or 4 if I use the plot command and all the plots 
are squished into the subregions for which I asked - even if things get 
ugly.

I use the PS and Agg backends and both yield  similar results.

My code is below.

Thanks for any help.

Jim

The Code:
import matplotlib
matplotlib.use('PS')
#matplotlib.use('Agg')
from matplotlib.matlab import *
import showPS
import showAGG

nTypes = 3
ind = arange(nTypes)  # the x locations for the groups
width = .2       # the width of the bars
nr = 1
nc = 3
crap = []
for c in range(nc):
  for r in range(nr):
     subplotID = nc*r + c + 1
     crap.append(subplotID)
     ax = subplot(nr,nc,subplotID)
     frame = ax.get_frame()
     frame.set_linewidth(1.5)
     amip = (20, 35, 30)
     p1 = bar(ind, amip, width, color='r')
     climo = (25, 32, 34)
     p2 = bar(ind+width, climo, width, color='y')
     isccp = (5, 10, 60)
     p3 = bar(ind+2.0*width, isccp, width, color='g')
     atlas = (20, 40, 70)
     p4 = bar(ind+3.0*width, atlas, width, color='b')
     ylabel('Cloud Fraction (%)')
     title('Cloud' + str(subplotID)+'_'+str(c)+'_'+str(r) )
     set(gca(), 'xticks', ind+2*width)
     set(gca(), 'xticklabels', ('low', 'mid', 'high') )
     set(gca(), 'yticks', arange(0,100,10))
     legend( (p1[0], p2[0],p3[0],p4[0]), ('Amip', 
'climo','isccp','atlas') )

savefig('barchart_test')
showPS.showPS('barchart_test')

>>>>> "Arnold" == Arnold Moene <arn...@wu...> writes:

    Arnold> Dear all, Praise for this first serious python plotting
    Arnold> package!

Thanks!

    Arnold> I have two minor comments:

    -> when I try to do an errorbar plot with a scalar, rather than an
    Arnold> array as the argument (i.e. errorbar(scalar_x, scalar_y,
    Arnold> xerr=scalar_xerr)) I get the following error message:

    Arnold> "/usr/lib/python2.3/site-packages/matplotlib/axes.py",
    Arnold> line 569, in errorbar right = x+xerr[1] IndexError: index
    Arnold> out of bounds

    Arnold> This is due to the fact that the length of xerr is tested
    Arnold> only to be 1 or greater than 1 (asymmetric
    Arnold> errors). However, the option that I would like to plot
    Arnold> just one point is not included. So an extra if-clause is
    Arnold> needed before the present code that checks whether x and
    Arnold> xerr are arrays or not.

Well, the behavior you are seeing is the correct, documented behavior.
I think it would be easier for you to pass errorbar what it is
expecting than for errorbar to handle this additional use case - it's
already getting pretty hairy in there.

DO you have a problem doing

  errorbar([1],[2],xerr=[.1])


    -> I very much like the option for scatter to color the blobs to
    -> be
    Arnold> plotted. However, I have an application where the shape of
    Arnold> the blobs is immaterial: I would rather have filled
    Arnold> circles (or even better: filled symbols, be it circles,
    Arnold> triangles ...). Anyhow: I would like them to be symmetric
    Arnold> irrespective of the range of the axes. One option would be
    Arnold> to allow the s-keyword to be allowed to be negative, where
    Arnold> negative means: symmetric and the value of the negative
    Arnold> number being the size relative to some default size.

You're not the first!  The temporary workaround is to use
plot(x,y,'o') which uses fixes sizes in points for the symbols (you
can also use '*', '+', etc)  However, you cannot vary the size or
color of the symbols so this may not be workable.  

When I refactor the scatter / pcolor etc commands as discussed in my
response to Dominique Orban above, I'll make sure that this case is
handled as well.  I plan to support efficient handling of scatter with
arbitrary (or at least several) shapes where the size can be in data
(x,y) or user-defined coords, eg, physical sizes.  It would be helpful
for me if people gave some information about how they would like
scatter to behave, particularly in regard to the sizes since I am
still working this out.  

Would you like physical sizes, eg circles with radius in points, or
what?

Should the shape depend on the aspect ratio of the plotting window.
Eg, if it's a circle and you have a short, wide window, should it
appear ellipsoidal?

What basic scatter symbols should be supported?

What are the most common uses: fixed size?  fixed color?  vary both?

JDH

>>>>> "Dominique" == Dominique Orban <do...@da...> writes:

    Dominique> Hello, I am having trouble viewing scatter plots of
    Dominique> large data sets with the GTK backend. The plots are
    Dominique> sparsity patterns of sparse matrices, much in the way
    Dominique> of Matlab's spy function.  First of all, there may be a
    Dominique> conflict somewhere with my version of GTK/PyGTK;
    Dominique> issuing a plot() or scatter() command often results in
    Dominique> the message 'None Active' being displayed.  After a
    Dominique> couple of seconds, the python prompt comes back, and
    Dominique> the 'scatter' command results in a huge number of
    Dominique> messages of the form

    Dominique>     <matplotlib.patches.Circle instance at 0x4531730c>

Hi Dominique, 

It looks like there is a "print" statement somewhere in your code.
It's possible that this was from a vestigial debug command I left in.
I don't get get it on my system.  Are you using matplotlib-0.50?  I
also don't get the "None Active" line.

I'm using this as a test script  

    from matplotlib.matlab import *

    x = 100*rand(100000)
    y = 100*rand(100000)
    s = rand(100000)

    scatter(x,y,s)
    #plot(x,y,'o')
    show()

Takes about 30s on my system.  Note that plot with circles can be must
faster that scatter if you don't need to vary the size or color of the
symbols.

    Dominique> begin displayed; litterally hundreds of them---it takes
    Dominique> some 10 seconds. Finally, a show() opens up a Figure
    Dominique> window, but no plot. The matrix has 80519 nonzero
    Dominique> elements and is symmetric, so the scatter plot contains
    Dominique> roughly twice as many points. I have a Gnuplot
    Dominique> interface and a spy-like function which works just fine
    Dominique> and displays the pattern in a matter of a fraction of a
    Dominique> second.

There are several areas where matplotlib performance is subpar -
mostly for large numbers of patches (circles for scatter, rectangles
for pcolor).  Fixing this is a fairly high priority and I have a good
idea how to go about it - see
http://sourceforge.net/mailarchive/message.php?msg_id=7142332 for a
recent discussion.  I think in the next 3-4 weeks I can get this
fixed.  Basically, the plan is to set up an additional backend method
or two that the various backends may optionally implement in extension
code for performance.

    Dominique> Sometimes, a plot using the GTK backend gets stuck in
    Dominique> gtk.mainloop(), and i have to interrupt it with
    Dominique> Ctrl-C. I am using SuSE Linux 8.0. I have installed the
    Dominique> most recent versions of GTK and PyGTK.

Are you using the default GTK that comes with SuSE or did you upgrade?
I have gotten myself into a world of pain before trying to upgraded
GTK libs on a linux box.  It does look like you are getting some
unusual behavior.  Make sure you are using the latest matplotlib and
try running the test script I posted above.  If you still get the same
errors, something is whacked with your install or paths.  Otherwise,
stay tuned for performance enhancements coming soon to theaters
everywhere.

JDH

  - I have 2 Suplots on top of each other, but when I select only one
    axis on the drop-down button I get an error: 

This was a python2.2 specific bug relating to nested scopes.  It's
fixed for the next release.

  - I want to systematically disable the vertical zoom/move on the
    second subplot, but not the first. How can I do that ?

Well, when the menu is working properly that is how you do it.  By
"systematically", do you mean by default so it's disabled when the
plot comes up?

If so, this depends on how you are building your GUI window.
Basically you need to get your hands on the gtk.Toolbar instance,
which contains an omenu (gtk.OptionMenu) attribute (wait until the
next release, a couple of days, where I've cleaned this up a bit).
Eg, if you are using the matlab interface

  manager = get_current_fig_manager()
  items = manager.toolbar.omenuItems
  items[-1].set_active(False)

Turns off navigation for the last subplot.  items[0] would turn off
navigation for the first subplot, and so on.

Here are the docs for gtk.MenuItem -
http://www.gnome.org/~james/pygtk-docs/class-gtkmenuitem.html

  - How can I activate a tooltip on top of my plots ?

You'll have to read up on gtk tooltips.  Depends on what elements you
want to add tips to.  Give me a little more info.  DO you want to add
tips to the navigation buttons or lines on your plot or text elements
or what?

http://www.gnome.org/~james/pygtk-docs/class-gtktooltips.html

Also check out the pygtk FAQ http://www.async.com.br/faq/pygtk and
tutorial
http://www.moeraki.com/pygtktutorial/pygtk2tutorial/index.html.


  - I cannot zoom or move an plot I set the axis ticks to , ie
    ax.set_xticks([])

This is a bug - thanks for reporting it.  I use the tick width as the
default interval in pan.  When there are no ticks, I divided by the
number of ticks and got a division by zero error.  I now check for 0
and default to 20% of the plot if there are no ticks.

  - How could I add a button to directly print out the picture in the toolbar ?

Print as in to a printer?  No support for that yet and am not sure I
want to go there right now (cross platform printing would might take
lot of work).  Best I can offer you currently is the savefig button.
I think gnome-print has made a fair amount of progress since I first
wrote matplotlib so if you want to be on the vanguard and explore this
option I would be happy.

  - Is there any density plot available ?

Do you mean hist?

http://matplotlib.sourceforge.net/screenshots.html#histogram_demo
http://matplotlib.sourceforge.net/matplotlib.matlab.html#-hist

  - Can I change the size taken by each of the Suplots ? 

Just use axes.  Subplot derives from Axes.  With axes, you can set the
size of your subplot with a rect [left, bottom, width, height]

http://matplotlib.sourceforge.net/matplotlib.matlab.html#-axes
http://matplotlib.sourceforge.net/matplotlib.axes.html
http://matplotlib.sourceforge.net/screenshots.html#axes_demo

- I have MANY small plots on my graph. Is there a cost effective way
  to filter which plots to show ?

I don't know what this means.  Wouldn't you want to show all the plots
on your figure?  Why else would you put them there?  Please elaborate.

Cheers,
JDH

>>>>> "matthew" == matthew arnison <ma...@ca...> writes:

    matthew> Specifically, what about making the default behaviour to
    matthew> clip the data at the first point which is equal or
    matthew> greater than the axis range? That should maintain the
    matthew> efficiency gains of clipping, while still keeping the
    matthew> scientific plotting behaviour that I think most users are
    matthew> accustomed to.

Hi Matthew -

That's a good idea - the data clipping feature was a gotcha for many.
For the next release, I turned data clipping off by default since I
don't think it's the typical use case and implemented the "first point
which is equal or greater" suggestion.  However, I also have created a
config file, so you can set the default as you like.

JDH 

I upgraded to GD v2 and gdmodule v0.52 and I noticed that line widths
don't work correctly.  If the line width is >1 the line comes out as
'stripes'.  This is on Linux.  The GD lib < 2.0 didn't have line width
support at all so maybe I've never seen GD create variable width lines
on linux?  I wonder if this is a GD problem.

=20

I tried the new Agg backend, very nice.  I'm all set to jettison GD
altogether and go with Agg except that the anti-aliased graphs that look
so great, print poorly.. :-(  Is there a way to turn off the
anti-aliasing?  It would be *great* to be able to drop that GD
dependency.

=20

thanks,

Stefan =20

Dear all,

Praise for this first serious python plotting package!

I have two minor comments:

-> when I try to do an errorbar plot with a scalar, rather than an array 
as the argument (i.e. errorbar(scalar_x, scalar_y, xerr=scalar_xerr)) I 
get the following error message:

"/usr/lib/python2.3/site-packages/matplotlib/axes.py", line 569, in errorbar
     right = x+xerr[1]
IndexError: index out of bounds

This is due to the fact that the length of xerr is tested only to be 1 
or greater than 1 (asymmetric errors). However, the option that I would 
like to plot just one point is not included. So an extra if-clause is 
needed before the present code that checks whether x and xerr are arrays 
or not.

-> I very much like the option for scatter to color the blobs to be 
plotted. However, I have an application where the shape of the blobs is 
immaterial: I would rather have filled circles (or even better: filled 
symbols, be it circles, triangles ...). Anyhow: I would like them to be 
symmetric irrespective of the range of the axes. One option would be to 
allow the s-keyword to be allowed to be negative, where negative means: 
symmetric and the value of the negative number being the size relative 
to some default size.

Thanks and keep up the good work,

Arnold Moene

Hello,

I am having trouble viewing scatter plots of large data sets
with the GTK backend. The plots are sparsity patterns of
sparse matrices, much in the way of Matlab's spy function.
First of all, there may be a conflict somewhere with my
version of GTK/PyGTK; issuing a plot() or scatter() command
often results in the message 'None Active' being displayed.
After a couple of seconds, the python prompt comes back, and
the 'scatter' command results in a huge number of messages
of the form 

    <matplotlib.patches.Circle instance at 0x4531730c>

begin displayed; litterally hundreds of them---it takes some
10 seconds. Finally, a show() opens up a Figure window, but
no plot. The matrix has 80519  nonzero elements and is
symmetric, so the scatter plot contains roughly twice as
many points. I have a Gnuplot interface and a spy-like
function which works just fine and displays the pattern in a
matter of a fraction of a second.

Sometimes, a plot using the GTK backend gets stuck in
gtk.mainloop(), and i have to interrupt it with Ctrl-C. I am
using SuSE Linux 8.0. I have installed the most recent
versions of GTK and PyGTK.

Thanks in advance for any suggestions.
Dominique