matplotlib-users Mailing List for matplotlib

There is a minor bug in the object_picker.py example. On the 2nd and
subsequent times a line is "picked", if the marker style had been
changed previously, the original marker choice "None" no longer appears
as a choice in the marker menu.  I.e. there is no way to turn off a
previously selected marker.

Below is a (one line) patch that fixes this.

	-Al   
-- 
Al Schapira <a.d...@wo...>

 
*** /usr/local/matplotlib-0.51/examples/object_picker.py       
2004-02-26 15:22:58.000000000 -0500
--- object_picker.py    2004-03-09 16:50:36.000000000 -0500
***************
*** 151,157 ****
          marker = line.get_marker()
          if marker is None: marker = 'None'
          styles = [marker]
!         for key in lineMarkers.keys():
              if key == marker: continue
              styles.append(key)
   
--- 151,157 ----
          marker = line.get_marker()
          if marker is None: marker = 'None'
          styles = [marker]
!         for key in keys:
              if key == marker: continue
              styles.append(key)

>>>>> "Trevor" == Trevor Blackwell <tl...@an...> writes:

    Trevor> In backend_gtk.py, it does some antialiased rendering by
    Trevor> direct manipulation of pixel values which is only correct
    Trevor> for 24-bit TrueColor X11 visuals. I'm running a 16-bit (5
    Trevor> red, 6 green, 5 blue) visual, so this produces visual
    Trevor> junk.

    Trevor> I think the right way to do it is with a Pixbuf, which
    Trevor> handles alpha rendering. The code is much simpler and
    Trevor> faster too. Here is a patch:

Nice!!  Where did you learn that trick?  pb.pixel_array is not
documented at
http://www.gnome.org/~james/pygtk-docs/class-gdkpixbuf.html.

This is great to know because I can use the same approach to port
image handling to gtk once I get that up and running.

FYI, your patch was line wrapped but it was simple enough to apply "by
eye".

JDH

In backend_gtk.py, it does some antialiased rendering by direct
manipulation of pixel values which is only correct for 24-bit TrueColor
X11 visuals. I'm running a 16-bit (5 red, 6 green, 5 blue) visual, so
this produces visual junk.

I think the right way to do it is with a Pixbuf, which handles alpha
rendering. The code is much simpler and faster too. Here is a patch:

(This is from matplotlib-0.51.)

--- backends/backend_gtk.py~     Wed Mar  3 09:31:39 2004
+++ backends/backend_gtk.py      Tue Mar  9 12:00:37 2004

@@ -350,47 +351,20 @@
         xox = int(x+ox)
         yoy = int(y+oy)
 
-
-
         imw = min(imw, self.width-xox)
         imh = min(imh, yoy)
-        #print imw, imh, xox, yoy, self.width, self.height
-        image = self.gdkDrawable.get_image(xox, self.height-yoy, imw,
imh)
-        #return
-
-
-        tr = int(rgb[0]*255) # text red
-        tg = int(rgb[1]*255) # text green
-        tb = int(rgb[2]*255) # text blue
-
-        ind =  indices(Xs.shape)
-        numRows, numCols = Xs.shape
-        ind.shape = 2, numRows*numCols
-        Xs.shape = numRows*numCols,
-        visible = nonzero(Xs>0)
-        Xs.shape = numRows,numCols
-        for thisInd in visible:
-            j,i = ind[:,thisInd]
-            if i >= imw: continue
-            if j >= imh: continue
-            pixel = image.get_pixel(i, j)
-            br = (pixel >> 16)  & 0xff  # background red
-            bg = (pixel >> 8 ) & 0xff
-            bb = (pixel >> 0)  & 0xff
-            #print br, bg, bb, Xs[j,i]
-
-            alpha = int((255-Xs[j,i])*255)
-
-            nr = (((br - tr) * alpha) + (tr << 16)) >> 16
-            ng = (((bg - tg) * alpha) + (tg << 16)) >> 16
-            nb = (((bb - tb) * alpha) + (tb << 16)) >> 16
-            newpixel = (nr<<16) + (ng<<8) + (nb)
-            
-            image.put_pixel(i, j, newpixel)
 
-        self.gdkDrawable.draw_image(gc.gdkGC, image, 0, 0,
-                                    xox, self.height-yoy, imw, imh)
+        pb=gtk.gdk.Pixbuf(gtk.gdk.COLORSPACE_RGB,
+                          has_alpha=1, bits_per_sample=8, width=imw,
height=imh)
+        pbpix=pb.pixel_array
+        pbpix[:,:,3]=Xs
+        pbpix[:,:,0]=int(rgb[0]*255)
+        pbpix[:,:,1]=int(rgb[1]*255)
+        pbpix[:,:,2]=int(rgb[2]*255)
 
+        pb.render_to_drawable(self.gdkDrawable, gc.gdkGC, 0, 0, xox,
self.height-yoy, imw, imh,
+                              gdk.RGB_DITHER_NONE, 0, 0)
+            
         if 0:
             self.gdkDrawable.draw_rectangle(
                 gc.gdkGC, 0, xox,


-- 
Trevor Blackwell      tl...@an...       (650) 210-9272

>>>>> "Kuzminski," == Kuzminski, Stefan R <SKu...@fa...> writes:

    Stephan> The GD output looks good when printed, maybe I will
    Stephan> switch between the 2 ( Agg for display, GD for
    Stephan> printing ).  The Agg un-aliased lines don't come out
    Stephan> quite as well, they seem to render more of the pixels
    Stephan> for each point on the line.  Nice to have the
    Stephan> different backend options.

This has to do with how agg handles subpixel positioning - I've
emailed the agg list and gotten some advice but haven't come up with a
good system to make the lines appear the same thickness in the aliased
and antialiased cases.  I'm still working on it.

In the meantime, here is a little backend magic that will make it
easier for you to print to your backend of choice. This example
displays the image in the default GUI (GTKAgg for me) and prints
with GD.  

    from matplotlib.backends.backend_gd import FigureCanvasGD
    from matplotlib.matlab import *

    plot([1,2,3])

    manager = get_current_fig_manager()
    canvasgd = manager.canvas.switch_backends(FigureCanvasGD)
    canvasgd.print_figure('gdfig')
    show()

print_figure takes the same args as savefig.

gd has a pesky color allocation bug that I haven't figured out that
you may bump into.

JDH

The GD output looks good when printed, maybe I will switch between the 2
( Agg for display, GD for printing ).  The Agg un-aliased lines don't
come out quite as well, they seem to render more of the pixels for each
point on the line.  Nice to have the different backend options.

S   =20

-----Original Message-----
From: John Hunter [mailto:jdh...@ni...]=20
Sent: Monday, March 08, 2004 8:37 AM
To: Kuzminski, Stefan R
Cc: mat...@li...
Subject: Re: [Matplotlib-users] simple api question

>>>>> "Kuzminski," =3D=3D Kuzminski, Stefan R <SKu...@fa...>
writes:

    Kuzminski> Thanks, that worked for the line being plotted,
    Kuzminski> although the legend box and axis are still aliased.
    Kuzminski> Part of my requirements include supporting
    Kuzminski> presentation quality printing.  I would like to just
    Kuzminski> use Agg as GD has dependency issues and keeps popping
    Kuzminski> up other problems ( not to mention how great the Agg
    Kuzminski> output looks ).  But that great looking anti-aliasing
    Kuzminski> doesn't print well, so ideally there would be a
    Kuzminski> 'global' level flag that controls aliasing ( or not )
    Kuzminski> for everything drawn.  I know when the image is being
    Kuzminski> created for viewing or for printing and so can set the
    Kuzminski> flag accordingly.

I'll work on getting the rest of the objects to respect the
antialiased flag.  You can control antialiasing for all lines globally
with rcParams

  http://matplotlib.sourceforge.net/faq.html#CUSTOM


JDH

>>>>> "Flavio" == Flavio Codeco Coelho <fcc...@ci...> writes:

    Flavio> I cant install matplotlib because it cant find some files
    Flavio> in the Font tools tree.


Note if you just want wx, set all the BUILD_* flags in setup.py to 0
and distutils won't compile anything; wx doesn't depend on any of the
extension code.  If you do want to build the extensions from CVS, read
on.

CVS doesn't have a complete version of the FontTools* and ttfquery
that are needed to build the extensions.  An increasing number of
matplotlib backends need font-finding capabilities which FontTools and
ttfquery provide.  However, they are a pain to install and Paul
Barrett has been working on a replacement.  At one point I added them
to CVS since I was distributing them with matplotlib but thought twice
about it and tried to remove them.  However despite multiple attempts
I have not been able to get them out of CVS.  No in a nutshell there
is an incomplete version of FontTools and ttfquery in CVS, and I'm not
too inclined to add them since they will be purged in short order in
any case with the new fontfinder.

If you want to build from CVS, copy the agg, FontTools* and ttfquery
dirs/files from the 0.51 src distro into the CVS tree and build from
there.  This will all be cleared up soon.

JDH

errata of my last message:

I meant to say that:

these folders did not come with cvs update -d


Summarizing,


I cant install matplotlib because it cant find some files in the Font
tools tree.

hi, I am trying to install the latest CVS and I am getting the following
error message:


root@iprocc1-164 matplotlib]# python setup.py install
running install
running build
running build_py
package init file 'ttfquery/__init__.py' not found (or not a regular
file)
package init file 'FontTools/__init__.py' not found (or not a regular
file)
package init file 'FontTools/fontTools/__init__.py' not found (or not a
regular file)
package init file 'FontTools/fontTools/encodings/__init__.py' not found
(or not a regular file)
error: package directory 'FontTools/fontTools/misc' does not exist
[root@iprocc1-164 matplotlib]# python setup.py install
running install
running build
running build_py
package init file 'ttfquery/__init__.py' not found (or not a regular
file)
package init file 'FontTools/__init__.py' not found (or not a regular
file)
package init file 'FontTools/fontTools/__init__.py' not found (or not a
regular file)
package init file 'FontTools/fontTools/encodings/__init__.py' not found
(or not a regular file)
error: package directory 'FontTools/fontTools/misc' does not exist

these folder did come with cvs update -d

what wrong here?

thanks for any help...

FLavio

HI everybody,

I dont know if any of you is aware of the boa-constructor python IDE and
wx gui builder.

I use it and subscribe to its mailing list. Recently, there was this
discussion about having some scientific plotting controls added to Boa.
I include below, a message from Boa's main developer, Ryaan Booysen,
where he gives some pointers to anyone that might be interested in
adding plotting controls to Boa. They are talking about Chaco, but as
far as I know, Chaco development is stalled and Matplotlib is far
superior (IMHO).

I believe that if anyone is interested in doing that should contact
Ryaan. He is a very nice guy. I also believe that it would greatly
improve the visibility of matplotlib since Boa has a very large user
base.

Well, its just an idea.

have fun

Flavio



-----Forwarded Message-----

From: Riaan Booysen <riaan@e.co.za>
To: Ricardo Henriques <pax...@sa...>
Cc: boa...@li... <boa...@li...>
Subject: Re: [Boa Constr] Fw: Any Chaco plugins?
Date: Tue, 09 Mar 2004 13:33:00 +0200

Hi Ricardo,

Ricardo Henriques wrote:
> Hi...
> I sucessfully used Boa to help me build scientific applications. I normally
> use wxPyPlot to plot graphics witch has a plug-in for Boa, it is quite
> alright, but sometimes I nead a plotting library with more features like
> Chaco found at www.scipy.org . Anyone knows any plug-in for this plotting
> library or any other than wxPyPlot?
> Where can I get some information about how to build a plug-in for Boa? 
> Tks...

You may look at the examples for adding a control in
Plug-ins/UserCompanions.py

I suggest you first try to use the Custom Classes feature to
use a Chaco Plot window in the Designer.
See Docs/boa/apphelp/MixingSource.html
This might be a simpler option.

Cheers,
Riaan.



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
Boa-constructor-users mailing list
Boa...@li...
https://lists.sourceforge.net/lists/listinfo/boa-constructor-users

>>>>> "Vincent" == Vincent BOYER <bo...@cl...> writes:
    Vincent> Is there a way to do that in Matplotlib?  Does the Matlab
    Vincent> command "hold" have a equivalent in Matplotlib? I didn't
    Vincent> find it, and if it exists, then I could plot the lines of
    Vincent> the matrix Y one by one.

Hold is on by default.  So you can do


for y in Y:
    plot(x,y)

To clear the axes between plot commands, use gca.

Hope this helps,
JDH

Hi.

I'm trying to plot a matrice against a vector using Matplotlib.
Something like we can do with Matlab :

plot(x,Y)

where
    x is an Numeric.array of shape (n,)
    Yis an Numeric.array of shape (m,n)

Is there a way to do that in Matplotlib?
Does the Matlab command "hold" have a equivalent in Matplotlib? I didn't 
find it, and if it exists, then I could plot the lines of the matrix Y 
one by one.

Thank you for any information.

Vincent