matplotlib

>>>>> "Sascha" == Sascha  <saschagl@...> writes:

    Sascha> I am writing a web server app that creates charts among
    Sascha> other things. I am trying to get rid of the temporary file
    Sascha> that I use to transmit the figures created with matplotlib
    Sascha> to the actual web server. Although print_figure says "If
    Sascha> filename is a fileobject, write png to file object (thus
    Sascha> you can, for example, write the png to stdout)" I can't
    Sascha> successfully write anything to stdout. Anyone knows an
    Sascha> example or can give me some hint what I can do to get rid
    Sascha> of the tempfile?

Short answer: no known way to do this currently, though we'd like to
figure it out.  As far as I know (and could very well be wrong)
libpng requires a FILE*, which StringIO and cStringIO do not provide.

JDH

> As far as I know (and could very well be wrong)
> libpng requires a FILE*, which StringIO and cStringIO do not provide.

I think this is exactly the reason why printing to stdout fails.

Well, if anybody has any other idea I'd be very grateful.

Sascha

-- 
5 GB Mailbox, 50 FreeSMS http://www.gmx.net/de/go/promail
+++ GMX - die erste Adresse für Mail, Message, More +++

Aleksander Schwarzenberg-Czerny wrote:
> Dear matplotlib Pundits,
> I am trying to install matplotlib-0.83.2 on my FedoraCore3 linux.
> It seems I got installed all required packages (including pygtk2 and gtk2) 
> and the corresponding devel packages. After running
> python setup.py build
> I got many messages, terminated with the following error message:
> ++/3.4.4/functional:54,
>                  from src/_transforms.cpp:1:
> /usr/include/features.h:150:1: warning: this is the location of the 
> previous definition
> src/_transforms.cpp:8:34: Numeric/arrayobject.h: No such file or directory

> Do I miss some required package?

Numeric.

http://numeric.scipy.org

-- 
Robert Kern
rkern@...

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter

Aleksander Schwarzenberg-Czerny wrote:

> src/_transforms.cpp:8:34: Numeric/arrayobject.h: No such file or directory

> Do I miss some required package?

First, check whether you actually have numeric installed:

planck[python]> python -c 'import Numeric;print Numeric.__version__'
23.7

If that works, it means that you have installed the Numeric headers in some 
non-standard location.  On my system, they live in:

planck[python]> locate arrayobject.h
/usr/include/python2.3/Numeric/arrayobject.h


I believe by default, distutils adds automatically /path/to/include/python to 
the include file search path (via -I), but if you've installed Numeric in some 
non-standard location, that automatic search may fail.  I don't see 
immediately a way to tell distutils to add specific extra paths, but there may 
be one.  The cheap fix is to copy the Numeric/*.h directory over to the 
standard python location for headers in your system.

Cheers,

f

Hi!
Indeed, I misunderstood documentation and installed only one package
from two (Numeric and numarray). After installation of Numeric and 
devels for tcl and tk installation went smoothly.
Thank you

Alex Schwarzenberg-Czerny

On 
Tue, 30 Aug 2005, Fernando Perez wrote:

> Aleksander Schwarzenberg-Czerny wrote:
> 
> > src/_transforms.cpp:8:34: Numeric/arrayobject.h: No such file or directory
> 
> > Do I miss some required package?
> 
> First, check whether you actually have numeric installed:
> 
> planck[python]> python -c 'import Numeric;print Numeric.__version__'
> 23.7
> 
> If that works, it means that you have installed the Numeric headers in some 
> non-standard location.  On my system, they live in:
> 
> planck[python]> locate arrayobject.h
> /usr/include/python2.3/Numeric/arrayobject.h
> 
> 
> I believe by default, distutils adds automatically /path/to/include/python to 
> the include file search path (via -I), but if you've installed Numeric in some 
> non-standard location, that automatic search may fail.  I don't see 
> immediately a way to tell distutils to add specific extra paths, but there may 
> be one.  The cheap fix is to copy the Numeric/*.h directory over to the 
> standard python location for headers in your system.
> 
> Cheers,
> 
> f
>

>>>>> "David" == David  <david.trem@...> writes:

    David> Hi all, and thanks for amazing matplotlib.  I devellop my
    David> data processing / plotting software using glade, pygtk,
    David> scipy and matplotlib 0.83.2.

    David> I use the class library and I run into trouble with
    David> autoscaling.

    David> I expected that autoscale_view() function of axes class
    David> would do autoscaling. But if I update the data using
    David> set_data(), set_xdatat() or set_ydata() (of lines object)
    David> and I call autoscale_view() of the corresponding axes and
    David> then redraw the canvas, the scale is not changed and some
    David> data points are thus out of the picture.

    David> Do I miss or misunderstand something ?


The axes keeps a copy of the "dataLim" and won't automatically know if
you change the line's data.  You need to call

  ax.update_datalim_numerix(x, y, ignore)

where x and y are the numerix arrays you pass to line.set_data.

If ignore is True, the previous data passed to the axes will be
ignored in computing the new datalim.  If False, the datalim will
include the previous data and the current data you are passing it.

After this, a call to 

 ax.autoscale_view()

should work as you expect.

JDH

Hello,

 No actually it doesn't work exactly as I expected.

I have different lines (at least two) that I would like to
update and also hide/show and I expected that autoscaling will take all
these aspects (updated data, hiden data) to scale the graph to all visibl=
e data.

The main problem is that old data limit is kept or totaly lost. The axes
scaling doesn't seem to work with the list of lines,
but is updated each time you add a line to the axes (as far as
I can undestand from the source code). So the scale can only be increased
but never decreased. Am I write or wrong ?

Furthermore I use the sharey stuff that complicated a bit the story.
So I don't think I can easelly do what I want using the axes class.
Maybe I should calculate the scaling in my software and set the min/max v=
alue.

If one of you have another idea please let me know.

Thanks for your help,

David



Selon John Hunter <jdhunter@...>:

> >>>>> "David" =3D=3D David  <david.trem@...> writes:
>     David> I use the class library and I run into trouble with
>     David> autoscaling.
>
>     David> I expected that autoscale_view() function of axes class
>     David> would do autoscaling. But if I update the data using
>     David> set_data(), set_xdatat() or set_ydata() (of lines object)
>     David> and I call autoscale_view() of the corresponding axes and
>     David> then redraw the canvas, the scale is not changed and some
>     David> data points are thus out of the picture.
>
> The axes keeps a copy of the "dataLim" and won't automatically know if
> you change the line's data.  You need to call
>
>   ax.update_datalim_numerix(x, y, ignore)
>
> where x and y are the numerix arrays you pass to line.set_data.
>
> If ignore is True, the previous data passed to the axes will be
> ignored in computing the new datalim.  If False, the datalim will
> include the previous data and the current data you are passing it.
>
> After this, a call to
>
>  ax.autoscale_view()
>
> should work as you expect.
>
> JDH
>

david> I have different lines (at least two) that I would like to
     david> update and also hide/show and I expected that autoscaling
     david> will take all these aspects (updated data, hiden data) to
     david> scale the graph to all visible data.
     david> Furthermore I use the sharey stuff that complicated a bit the=
 story

Thank you very much John. It works !
Here is the final version of the algorithm:
line_list is the list of line

    >ignore =3D True
    >for line in (line_list):
    >   if not line.get_visible(): continue
    >   self.ax.dataLim.update_numerix( line.get_xdata(),
    >                                   line.get_ydata(),
    >                                   ignore )
    >   ignore =3D False
    >self.ax.autoscale_view()

Note: we cannot use update_datalim_numerix() because "ignore" is not an a=
rgument
      but using dataLim.update_numerix is ok.

This version allows to avoid problem if the fist line is
not visible (compared to your original version)

Same function applied to the second axes which share the y axis
and everything is perfect :-)

What about adding this function to the axes class using the line list ava=
ilable?
Should give something like:

  def autoscale_visible_lines(self):
    ignore =3D True
    for line in (self.lines):
       if not line.get_visible(): continue
       self.dataLim.update_numerix( line.get_xdata(),
                                    line.get_ydata(),
                                    ignore )
       ignore =3D False
    self.autoscale_view()

Thanks for your help and your amazing matplotlib !

Best regards,

David


Selon John Hunter <jdhunter@...>:
> If you have a list of lines that you want to pass to the autoscaler,
> and have it operate on that list only (ignoring previous settings) do
> the following
>
> First update your line data and then create a list of lines that you
> want to autoscale for.  The ignore setting will be True for the first
> line in the list which will cause it to ignore it's history.  Thus the
> datalim will be updated to bound only the lines in the list "lines"
>
> for i,line in enumerate(lines):
>    ignore =3D i=3D=3D0
>    if not line.get_visible(): continue
>    x =3D asarray(line.get_xdata())
>    y =3D asarray(line.get_ydata())
>    ax.update_datalim_numerix(x, y, ignore)
> ax.autoscale_view()
>
> You should be able to do this separately for each of your two axes
> (even if they are sharing the x axes, their y axes is independent).
>
> If you are still having trouble after trying this, please post a
> complete example.
>
> JDH
>

>>>>> "Schindler" == Schindler Benjamin <benjamin.schindler@...> writes:

    Schindler> Hi In my program, I do a lot of axis changes and
    Schindler> redraws / resizes. The redraw speed of matplotlib seems
    Schindler> very bad at best - is there anything I can do to make
    Schindler> it go faster?

Start here : http://matplotlib.sourceforge.net/faq.html#SLOW

The best thing you can do is create a sample application that you feel
is slow that can be run on our machines -- eg create a dummy data
server to replace for your real data server.  Then run your script
with some timing information to get an estimate of how fast the draw
is on your system (eg frames/second, see
examples/dynamic_image_gtkagg.py for one example which uses timing).

That way we can see if we are getting comparable performance numbers
and more importantly, may be able to suggest some improvements.

Generally the three things that make drawing slow are

 * creating too many objects (use collections instead).

 * using a backend which has a slow Agg->GUI widget transfer speed.  I
   just checked in Ken's patch which fixes this for WXAgg.

 * redrawing the entire canvas when you only need to redraw a selected
   artist or two.  See
   http://www.scipy.org/wikis/topical_software/Animations -- Ken's
   patch in CVS also added support for this kind of redrawing to
   WXAgg.

Hope this helps,
JDH

On Tuesday 30 August 2005 1:37 am, sipior@... wrote:
> > Try increasing the tick.major.pad in your rc settings
>
> Much obliged, Darren, that was just what I needed. Another small
> imposition, if I may: I'm using twinx() to get a second y-axis for a plot.
> It works fine, except that the tick marks from the first axis are also
> present on the right-hand side, causing a bit of a mess. Is there a way to
> suppress printing the tick marks just for the right plot side? All the
> reading I've done leads me to believe that this is an all-or-nothing
> affair, as I can only suppress tick marks by setting yticks to the empty
> set, and this of course removes all the tick marks on that axis. Am I
> missing something obvious here? I noticed that the two_scales.py demo
> exhibits the same behaviour (juxtaposing the two sets of ticks on the right
> side).

try this:

plot([1,2,3,4])
gca().yaxis.tick_left()
twinx()
plot([11,12,13,14])

Darren

I have sortof solved my problem. The main problem was, that =
autoscrolling
doesn't seem to work the way it should - now paired with a little bug =
that
got hidden made these problems for me. Since I have to control my y and =
x
axis closely anyway, There is no real problem anymore.=20
Thanks for all the help.=20

-----Urspr=FCngliche Nachricht-----
Von: John Hunter [mailto:jdhunter@...
Gesendet: Dienstag, 30. August 2005 00:22
An: Benjamin Schindler
Cc: 'Werner F. Bruhin'; matplotlib-users@...
Betreff: Re: [Matplotlib-users] Re: Line2D and wx Backend

>>>>> "Benjamin" =3D=3D Benjamin Schindler <bschindler@...> =
writes:

    Benjamin> Hi Sortof, but not really. I've just discovered
    Benjamin> something reeeally interesting.

    Benjamin> I've produced a sample app that exhibits this
    Benjamin> behaviour. However, monitor.py (in plugins/wx) still
    Benjamin> depends on twisted.spread.pb. If I comment out that
    Benjamin> import, it all runs fine. (you can add data with the a
    Benjamin> button and the slider in the left pane) It's hard for me
    Benjamin> guess who is causing the troubles - whether it's
    Benjamin> matplotlib not beeing able to handle the twisted
    Benjamin> threading or twisted breaking matplotlib - I don't know

Yep, this looks like a threading problem.  Could be a tough nut to
crack.  Sorry, but I don't have any concrete suggestions for you.
Perhaps someone on the wx list would have some idea how to make wx
events play nicely with twisted threads.

JDH

John Hunter wrote:

>>>>>>"Vidar" == Vidar Gundersen <vidar+list@...> writes:
>>>>>>            
>>>>>>
>
>    Vidar> thanks for providing this, i hope this will be included in
>    Vidar> the matplotlib distribution, along with your script for
>    Vidar> previewing and listing all colormaps?  (see modification of
>    Vidar> original code below.)
>
>I haven't been able to find any licensing information for the color
>scales at colorbrewer or http://geography.uoregon.edu/datagraphics .
>Does anyone know what their license is, if any?
>
>JDH
>
>  
>
John: Colorbrewer uses an apache-style license 
(http://www.personal.psu.edu/faculty/c/a/cab38/ColorBrewer/ColorBrewer_updates.html).  
I don't know about the DataGraphics color schemes.

-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