matplotlib-users Mailing List for matplotlib

>>>>> "Nils" == Nils Wagner <nw...@me...> writes:
 Nils> Hi all, I am wondering, if there is a tutorial (pdf-format)
 Nils> illustrating all features of matplotlib.
There is nothing in pdf format, but there is a tutorial on the web
page
 http://matplotlib.sourceforge.net/tutorial.html
It's only a basic introduction, but is enough to get you started.
After that the best way to proceed is to look at the examples dir of
the source distribution (*.zip or *.tar.gz). Also, the library is
matlab compatible, and there is a wealth of documentation for matlab.
The reference for the matlab interface is 
 http://matplotlib.sourceforge.net/matplotlib.matlab.html
A read through this page will give you a reasonably complete overview.
Of course, you can also access these from the python shell
 >>> help(plot) or
 >>> help(text) or
 >>> help(set)
Note the above is just for the matlab interface, which is what most
people use. You can also embed matplotlib in GTK or WX applications.
Generating pdf from the html pydoc output is reasonably easy, by the
way, with a tool like htmldoc.
A PDF User's Guide is a good idea. It's officially on the lost of
things to do.
JDH

Hi all,
I am wondering, if there is a tutorial (pdf-format) illustrating all
features of matplotlib.
Nils

This is the alpha release of version 0.40. Lots of changes since
0.32, including some API changes which will require minor changes for
application writers but no changes in the matlab interface. Please
test this over the next couple of weeks so any bugs and changes can be
included in the real release
Here are the release notes
What's new
WX python backend
 Jeremy O'Donoghue has done an amazing job implementing the backend
 for wxpython, with all the features such as interactive navigation
 that are supported in the GTK backend. matplotlib should not be
 available with a GUI anywhere wxpython, Numeric and python work,
 including linux and friends, win32, and Mac OS X. See
 http://matplotlib.sourceforge.net/matplotlib.backends.backend_wx.html
 for a summary of known issues.
Pseudo color plots
 The pcolor command generates pseudo color plots. See
 http://matplotlib.sourceforge.net/screenshots.html#pcolor_demo and
 http://matplotlib.sourceforge.net/screenshots.html#mri_with_eeg for
 screenshots and example code
New time-series plotting functions
 
 * psd - plots the power spectral density of a time series
 * csd - plots the cross spectral density of two time series
 * cohere - plots the coherence
 See the examples dir in the src distro: psd_demo.py and csd_demo.py
Substantially improved layout
 The transform architecture was refactored, allowing much more
 precise layout. Lines, patches, text, etc... can now be placed and
 scaled in arbitrary units, relative axes units, or physical size.
 Applicaiton programmers who want to create lines, patches and text
 directly using the API should read the transform module docs for
 more info. See the text help for an example of how to specify text
 locations in axes coords (0,0 is lower left and 1,1 is upper right)
Expanded legend capabilities
 The legend class is improved, with a more sophisticated layout
 engine and the ability to accept lines and rectangle patches as an
 optional first argument to specify which lines/patches make up the
 legend. There are also additional legend placement locations, like
 'upper center'. See
 http://matplotlib.sourceforge.net/screenshots.html#legend_demo
Expanded errorbar capabilities
 Gary Ruben contributed some code to support x and y errorbars,
 either symmetrix or asymettric, in one enhanced function 'errorbar'.
 See the errorbar_demo.py for examples of all the wild and wonderful
 errorbar styles. Bar charts can now also display errorbars; see
 http://matplotlib.sourceforge.net/screenshots.html#barchart_demo
Figure size and DPI controllable from matlab interface
 The figure command now takes optional args figsize and dpi to set
 the figure size and DPI in the matlab interface. This change
 involved some changes in the way default sizes and resolutions were
 handled among the various backends, with the effect that figures
 generated by existing scripts may appear different, eg, in the
 relative size of text to the figure elements. The advantage is that
 the current implementation does a better job of computing true sizes
 with increased fidelity between backends. Apologies for any
 inconveniences!
API changes
 There have been some minor changes to the API for those using
 matplotlib embedded in GTK applications. 
 * If you instantiate an Axes of Subplot intace, the first arg to
 __init__ must be the figure that contains it.
 * If you instantiate any artists, eg, Line2D, Rectangle, or
 AxisTextGTk, you must initialize them with their dpi, bbox, and
 transforms. See the help for the transforms modules and the
 examples logo.py and mri_with_eeg.py, where the objects are
 explicity created using the new API.
 * The only change to the matlab interface is in the signature of the
 errorbar func, which breaks matlab compatibility for the enhanced
 ability to do x and y errorbars.
 * See matplotlib.axes.py for a complete list of API changes

>>>>> "Flavio" == Flavio C Coelho <fcc...@fi...> writes:
 Flavio> Hi, I accidentally deleted a message from Jeremy in which
 Flavio> he had code snippets that worked with the WX backend in
 Flavio> its current CVS incarnation.
Have you read http://matplotlib.sourceforge.net/backends.html?
There is nothing special about invoking the wx backend. The two ways
of invoking a given backend are described on that page.
If it's not clear after reading, please let me know so I can update
the documentation (or provide suggestions for what can be clearer).
Thanks!
JDH

Thanks for the help. Indeed I would like the toolbar eventually.
Please let me know when it is fixed.
jgw
> P.S. In the example I've commented out the toolbar as I made
> a mistake when I checked in __init__.py in the backends
> directory, and the toolbar doesn't work right now. It will
> be fixed tomorrow, if this matters.

John,
> John Wohlbier writes:
> I ended up downloading the latest CVS using the command
> cvs -z3 -d:pserver:anonymous@cv...:/cvsroot/matplotlib co
> matplotlib
> 
> and doing python setup.py install
> 
> Now when I run my application (which worked with version 0.32) I get
> this error on startup
[snip]
> ImportError: cannot import name FigureManagerBase
You're not doing anything wrong - there have been some recent changes to
the API which have moved FigureManagerBase. The example I posted just
now does things a different way (and does work with latest CVS, honest!)
 
> By the way, when I run the python setup.py install command will
> the script properly overwrite the existing version of the matplotlib
> library, or should I go in and delete the old library first?
It updates things very nicely - no need to delete anything.
Regards
Jeremy

>>>>> "John" == John Wohlbier <jd...@go...> writes:
 John> Am I doing something wrong? I guess I'll go back to version
 John> 0.32 for a while.
No, it appears your CVS mirror is behind. This happens all the time.
Here is a snapshot
 http://nitace.bsd.uchicago.edu:8080/files/share/matplotlib-0.40g.tar.gz
 John> By the way, when I run the python setup.py install command
 John> will the script properly overwrite the existing version of
 John> the matplotlib library, or should I go in and delete the old
 John> library first?
The installer will overwrite files but will not delete files that are
no longer used. There is no harm in deleting the old library and it
might do some good.
 John> Thanks!
No problem ...
JDH

I ended up downloading the latest CVS using the command
cvs -z3 -d:pserver:ano...@cv...:/cvsroot/matplotlib co
matplotlib
and doing python setup.py install
Now when I run my application (which worked with version 0.32) I get
this error on startup
Traceback (most recent call last):
 File "./mptwt.py", line 3, in ?
 from mp_win import *
 File "./mp_win.py", line 10, in ?
 from matplotlib.matlab import *
 File "/usr/lib/python2.2/site-packages/matplotlib/matlab.py", line 121,
in ?
 from axes import Subplot, Axes
 File "/usr/lib/python2.2/site-packages/matplotlib/axes.py", line 118, in ?
 import backends
 File "/usr/lib/python2.2/site-packages/matplotlib/backends/__init__.py",
line 21, in ?
 from backend_wx import AxisText, Figure, FigureManager, Gcf, \
 File
"/usr/lib/python2.2/site-packages/matplotlib/backends/backend_wx.py",
line 164, in ?
 from matplotlib._matlab_helpers import FigureManagerBase, GcfBase
ImportError: cannot import name FigureManagerBase
Am I doing something wrong? I guess I'll go back to version 0.32
for a while.
By the way, when I run the python setup.py install command will
the script properly overwrite the existing version of the matplotlib
library, or should I go in and delete the old library first?
Thanks!
jgw
>>>>>> "John" == John Wohlbier <jd...@go...> writes:
>
> John> Hi, I am building a wxPython app, and would like to try
> John> incorporating matplotlib, even though the wx backend is
> John> reportedly in its development state. Could someone please
> John> give a snippet of code using the wx backend? Also, do I need
> John> pygtk even though I intend on using wxPython?
>
> The matplotlib code you write is more or less backend independent
> (same for gtk, wx, ps, etc..). You do not need pygtk. You just need
> wxpython and Numeric - which are both included in the enthought
> edition of python for win32, by the way; http://enthought.com/python.
>
> If you just want to make some figure windows to explore the plotting
> commands, you have 2 choices
>
> import matplotlib
> matplotlib.use('WX')
> from matplotlib.matlab import *
> plot([1,2,3])
> show()
>
> or
> # myscript.py
> from matplotlib.matlab import *
> plot([1,2,3])
> show()
>
> > python myscript.py -dWX
>
> That is, you choose the wx backend either by doing 'use' or -dWX. See
> http://matplotlib.sourceforge.net/backends.html for more info.
>
> See the examples dir of the src distro for examples of how to make plots
> more sophisticated than [1,2,3].
>
> But I gather from your post that you want to see an example of
> embedding the plot in an application. Jeremy, the wx backend
> implemented, can probably help you more here. In the mean time, you
> might want to look at the FigureManagerWX class in the file
> matplotlib/backends/backend_wx.py to see how to embed a plot in a wx
> frame.
>
> I recommend you use the latest CVS, which has a number of improvements
> over the 0.32 release. If you don't have CVS access, let me know and
> I'll send you a snapshot (are you on windows).
>
> John Hunter

>>>>> "John" == John Wohlbier <jd...@go...> writes:
 John> Thanks, that worked for being able to generate the [1,2,3]
 John> plot from within my application. Eventually I will want to
 John> embed into a wxFrame, but for now I think this should
 John> suffice.
I know Jeremy is planning on writing some example code to show how to
embed wx in an application with or without the navigation toolbar, and
possibly provide some utility classes to make it really easy.
If you have bug reports or suggestions for improvement regarding the
wx backend, the best place to send them is the matplotlib-dev mailing
list - http://sourceforge.net/mailarchive/forum.php?forum_id=36187.
There is a list of known bugs in header of the backend_wx.py code.
 John> I'm actually on linux, and am using 0.32 for now. I will get
 John> the latest CVS when I am up and running. When do you expect
 John> to release the next version?
We're shooting for early December.
JDH

John Wohlbier writes:
> Hi,
> 
> I am building a wxPython app, and would like to try
> incorporating matplotlib, even though the wx backend
> is reportedly in its development state.
> Could someone please give a snippet of code using
> the wx backend? Also, do I need pygtk even though I
> intend on using wxPython?
Hi John,
I'll answer your second question first: wxPython uses
the 'native' widget set of whichever platform you are using,
so if your platform is Windows or Mac you won't need gtk at
all. If your platform is Linux or other Unix/BSD variant,
I believe that wxPython depends on gtk, but not pyGtk
(gtk has been used as the 'native' Unix widget set...)
The answer to your first question is... it depends on
what you are trying to do.
I have modified the 'embedding_in_gtk.py' example to work
under wxPython - it hasn't yet made it into CVS (it will
shortly :-)
It's about as simple as things get. If you don't want a
navigation toolbar, simply leave out the calls to the
toolbar. I've put the Figure (and toolbar) into a Sizer,
as it's the easiest way to manage things.
Regards
Jeremy
P.S. In the example I've commented out the toolbar as I made
a mistake when I checked in __init__.py in the backends
directory, and the toolbar doesn't work right now. It will
be fixed tomorrow, if this matters.
==========================================================
import matplotlib
matplotlib.use('WX')
	
from matplotlib.backends import Figure
from matplotlib.axes import Subplot
import Numeric as numpy
	
from wxPython.wx import *
	
class PlotFigure(wxFrame):
	def __init__(self):
		wxFrame.__init__(self, None, -1, "Test embedded wxFigure")
		self.fig = Figure(self, -1, (5,4), 72)
		#self.toolbar = Toolbar(self.fig)
		sizer = wxBoxSizer(wxVERTICAL)
		sizer.Add(self.fig, 0, wxTOP)
		#sizer.Add(self.toolbar, 1, wxGROW)
		self.SetSizer(sizer)
		self.Fit()
			
	def plot_data(self):
		a = Subplot(self.fig, 111)
		t = numpy.arange(0.0,3.0,0.01)
		s = numpy.sin(2*numpy.pi*t)
		a.plot(t,s)
		self.fig.add_axis(a)
		self.fig.draw()
	
if __name__ == '__main__':
	app = wxPySimpleApp()
	frame = PlotFigure()
	frame.plot_data()
	frame.Show(true)
	app.MainLoop()
In a larger program, it is probably simplest to create
a class derived from wxPanel to encapsulate the plotting
functionality. The skeleton would be something like:
class PlottingPanel(wxPanel):
	def __init__(self, parent, id=-1)
		wxPanel.__init__(self, parent, id)
		self.fig = Figure(self, figsize=..., dpi=...)
		# Add a toolbar if you want...
		# Put the Figure into a sizer (toolbar as well, if needed)
		sizer = wxBoxSizer(...)
		sizer.Add(self.fig, ...)
		self.SetSizer(sizer)
		self.Fit()
		self.n_plot = 1
	def add_subplot(x_list, y_list):
		assert len(x_list) == len(y_list), "Data sets must be same length"
		a = Subplot(self.fig, n*111)
		a.plot(x_list, y_list)
		self.fig.add_axis(a)
For what it's worth, backend_wx is quite usable for basic plotting. There are a few
oddities in the operation of the toolbar, which I haven't quite sorted out, and there
is not yet the ability to save a plot as a graphic. I develop on wxPython for Win32
for a data analysis project which is the reason I started on backend_wx, but regression
test everything on a Debian box (hence gtk) at home. You will need to use either a
working snapshot (I believe John posted one a few days back), or keep a careful eye on
the CVS. John is making some very worthwhile improvements in basic functionality, but
I track these a day or so later in backend_wx (this makes my job much simpler as I
basically just copy what John has done and 'translate' it to wx!)
As of writing, the example above works with CVS. In a day or two it will appear as part
of the matplotlib examples (probably with more comments!). For now it should get you
going.

Thanks, that worked for being able to generate the [1,2,3] plot
from within my application. Eventually I will want to embed
into a wxFrame, but for now I think this should suffice.
I'm actually on linux, and am using 0.32 for now. I will
get the latest CVS when I am up and running. When do you
expect to release the next version?
Thanks again,
jgw
>>>>>> "John" == John Wohlbier <jd...@go...> writes:
>
> John> Hi, I am building a wxPython app, and would like to try
> John> incorporating matplotlib, even though the wx backend is
> John> reportedly in its development state. Could someone please
> John> give a snippet of code using the wx backend? Also, do I need
> John> pygtk even though I intend on using wxPython?
>
> The matplotlib code you write is more or less backend independent
> (same for gtk, wx, ps, etc..). You do not need pygtk. You just need
> wxpython and Numeric - which are both included in the enthought
> edition of python for win32, by the way; http://enthought.com/python.
>
> If you just want to make some figure windows to explore the plotting
> commands, you have 2 choices
>
> import matplotlib
> matplotlib.use('WX')
> from matplotlib.matlab import *
> plot([1,2,3])
> show()
>
> or
> # myscript.py
> from matplotlib.matlab import *
> plot([1,2,3])
> show()
>
> > python myscript.py -dWX
>
> That is, you choose the wx backend either by doing 'use' or -dWX. See
> http://matplotlib.sourceforge.net/backends.html for more info.
>
> See the examples dir of the src distro for examples of how to make plots
> more sophisticated than [1,2,3].
>
> But I gather from your post that you want to see an example of
> embedding the plot in an application. Jeremy, the wx backend
> implemented, can probably help you more here. In the mean time, you
> might want to look at the FigureManagerWX class in the file
> matplotlib/backends/backend_wx.py to see how to embed a plot in a wx
> frame.
>
> I recommend you use the latest CVS, which has a number of improvements
> over the 0.32 release. If you don't have CVS access, let me know and
> I'll send you a snapshot (are you on windows).
>
> John Hunter

>>>>> "John" == John Wohlbier <jd...@go...> writes:
 John> Hi, I am building a wxPython app, and would like to try
 John> incorporating matplotlib, even though the wx backend is
 John> reportedly in its development state. Could someone please
 John> give a snippet of code using the wx backend? Also, do I need
 John> pygtk even though I intend on using wxPython?
The matplotlib code you write is more or less backend independent
(same for gtk, wx, ps, etc..). You do not need pygtk. You just need
wxpython and Numeric - which are both included in the enthought
edition of python for win32, by the way; http://enthought.com/python.
If you just want to make some figure windows to explore the plotting
commands, you have 2 choices
 import matplotlib
 matplotlib.use('WX')
 from matplotlib.matlab import *
 plot([1,2,3]) 
 show()
or 
 # myscript.py
 from matplotlib.matlab import *
 plot([1,2,3]) 
 show()
 > python myscript.py -dWX
That is, you choose the wx backend either by doing 'use' or -dWX. See
http://matplotlib.sourceforge.net/backends.html for more info.
See the examples dir of the src distro for examples of how to make
plots more sophisticated than [1,2,3].
But I gather from your post that you want to see an example of
embedding the plot in an application. Jeremy, the wx backend
implemented, can probably help you more here. In the mean time, you
might want to look at the FigureManagerWX class in the file
matplotlib/backends/backend_wx.py to see how to embed a plot in a wx
frame.
I recommend you use the latest CVS, which has a number of improvements
over the 0.32 release. If you don't have CVS access, let me know and
I'll send you a snapshot (are you on windows).
John Hunter

Hi,
I am building a wxPython app, and would like to try
incorporating matplotlib, even though the wx backend
is reportedly in its development state.
Could someone please give a snippet of code using
the wx backend? Also, do I need pygtk even though I
intend on using wxPython?
Thanks!
jgw

>>>>> "Sandeep" == Sandeep Shetty <cab...@ya...> writes:
 John> According to what you wrote, you did everything correct.
 John> Many people run matplotlib under a variety of Microsoft
 John> environments, so you shouldn't have much trouble. My guess
 John> is that the PATH is not properly set. Which dir did you
 John> install the GTK runtime in? By default, I believe, it goes
 John> into a dir with spaces in the name, which are difficult to
 John> handle correctly in the PATH. For this reason, I always
 John> install the GTK runtime into C:\GTK and then add C:\GTK\lib
 John> and c:\GTK\bin to my PATH.
 Sandeep> Hi John, Thanks for getting back to me. I tried to start
 Sandeep> from scratch and got the same results. My path variable
 Sandeep> is set as follows: Path=c:\oracle\ora81\bin\;c:\Program
 Sandeep> Files\Oracle\jre1円.1.7\bin\;C:\WINNT\system32;C:\WINNT;C:\WINNT
 Sandeep> \System32\Wbem;d:\matlab6p5\bin\win32;d:\matlab\bin\win32;D:\Tcl\bin;c:\Program
 Sandeep> Files\ssh communicat ions security\ssh Secure
 Sandeep> Shell;c:\oracle\ora81\bin\;c:\Program
 Sandeep> Files\Oracle\jre1円.1.7\bin\;C:\WINNT\
 Sandeep> system32;C:\WINNT;C:\WINNT\System32\Wbem;d:\matlab6p5\bin\win32;d:\matlab\bin\win32;d:\Modeltech_5.5
 Sandeep> e\win32;d:\bin;d:\gnuwin32\b18\H-i386-cygwin32\bin;d:\PkZip;d:\Python23;C:\GTK\lib;C:\GTK\bin
I still wonder if there is a problem because of all those paths that
have spaces (eg, all those with "Program File" in them before the GTK
path entry.) For example, when I open a 'command' shell in windows XP
and do
 C:> set
to display my PATH, all the Program File entries are listed as
 C:\progra~1\somepath
with no spaces. When you do the same from the command shell, are any
spaces listed?
Please try the following. Open a command shell with
Start menu->Run->command ENTER
C:> set PATH=C:\GTK\bin;C:\GTK\lib
C:> d:\python23\python
>>> import pygtk
>>> pygtk.require('2.0')
>>> import gobject
and tell me what happens. I that works, the odds are that you have a
PATH problem. If not, its something else. If it does work, try
>>> from matplotlib.matlab import *
>>> plot([1,2,3,])
>>> show()
 Sandeep> Are are either of these the correct file for gobject? Or
 Sandeep> is gobject found some place else?
This looks good; I do have one additional gobject file, however,
 C:\Python23\Include\pygtk-2.0\pygobject
 Sandeep> If you think it is better I wait for the next rev of
 Sandeep> matlibplot, I can wait. Nonetheless, I would love to use
 Sandeep> your library as soon as possible.
The problem is that this is most likely not intrinsic to matplotlib,
but to your GTK settings. So you'll likely have the same problems in
the next release. There are many win32 users so don't despair, this
should be a relatively easy fix. Perhaps someone on the users list
has some input?
JDH

>>>>> "Flavio" == Flavio C Coelho <fcc...@fi...> writes:
 Flavio> Hi john, I've been away for a couple of weeks, and I just
 Flavio> did a "cvs update" to test the new features especially the
 Flavio> WX backend which I was very anxious to test.
I guess this is why they call CVS bleeding edge, eh? Yes, bug crept
in last night and is fixed in CVS as of early this morning. Note to
self: don't do commits over an xterm past 11PM.
However, the wx backend has not yet caught up with the latest changes
since Jeremy is out of town until Friday. I expect he'll do an update
then.
I've uploaded a Nov 15th snapshot which has a working wx backend to a
temporary web site. There are still a number of known bugs (eg,
figure placement and size) that Jeremy is systematically squashing but
it won't be long (target date early December). Have fun testing.
Please let us know if you find any bugs not in the known bugs section
of backends/backend_wx.py.
http://nitace.bsd.uchicago.edu:8080/files/share/matplotlib-0.40d.tar.gz
JDH

>>>>> "Charles" == Charles R Twardy <cha...@in...> writes:
 Charles> I might even suggest replacing the "pan" and "zoom"
 Charles> buttons by just using left-click and right-click for the
 Charles> drag. That way I can change tasks without mousing
 Charles> around. Zoom-rectangle could even be middle-click.
This is the way VTK does it and it is great. All three mouse buttons
are used for pan, zoom and rotate. The only problem is for
application developers using matplotlib, who will probably want to
reserve some of the mouse buttons over figures for context menus or
measurement tools, etc... I have done both in my applications. Its
possible to do both via different toolbars, but there are only so many
hours in the day. I'll let it simmer.
JDH

Hi John,
New navbar sounds good, esp. replacing arrows with hand. Default to hand
mode (maybe even revert after completing an operation with other modes).
I'm not so happy with all the modifier keys.
I'd have one modifier key, Shift, that constrains the motion, and
everything be drag-based:
	* In pan mode, Shift drag restricts motion to whichever the
	 principle mouse direction is. I can keep it within 45 degrees
	 myself, and this allows me to easily change x, then y.
	* In zoom mode, Shift drag zooms along principle mouse
	 direction. Otherwise zooms as radial distance of drag.
	 Mousewheel zoom zooms both axes.
	* Zoom rectangle: Shift-drag maintains axis ratio in
	 drawing the rectangle.
I might even suggest replacing the "pan" and "zoom" buttons by just using
left-click and right-click for the drag. That way I can change tasks
without mousing around. Zoom-rectangle could even be middle-click.
-Charles
-- 
Charles R. Twardy www.csse.monash.edu.au/~ctwardy
Monash University sarbayes.org
Computer Sci. & Software Eng.
+61(3) 9905 5823 (w) 5146 (fax)

>>>>> "Nelson" == Nelson Minar <ne...@mo...> writes:
 Nelson> John Hunter <jdh...@ac...> writes:
 >> There are a lot of good plotting libraries out there. I'll
 >> pitch mine. http://matplotlib.sourceforge.net
 Nelson> Thanks for posting this! I've been using gdchart for a
 Nelson> long time because it's simple and the graphs aren't too
 Nelson> bad. But it's very limited, in particular your X axis has
 Nelson> to be equally spaced.
Hi Nelson, thanks for the email and the detailed comments. It would
be great to have you join the mailing list - other users would benefit
from your comments -
http://lists.sourceforge.net/mailman/listinfo/matplotlib-users
 Nelson> I spent about an hour with matplotlib. It's clear you put
 Nelson> a lot of work into this, and I think this could be a
 Nelson> really great library. Some thoughts for you from a casual
 Nelson> user:
 Nelson> The installation prerequisites are a bit daunting but on a
 Nelson> Debian system it's not too bad. Debian has many of the
 Nelson> packages you need: apt-get install python2.3-numeric
 Nelson> python2.3-numarray python2.3-numeric-ext apt-get install
 Nelson> fonttools libgd2-dev python2.3-gtk2 python2.3-numeric-ext
 Nelson> is easy to forget: you need that for FFT and MLab.
I'll update the web page with this information for debian users.
Debian is great; unfortunately, I don't get much opportunity to use it
so I haven't gone through the process of installing matplotlib on
debian. There was some interest on the mailing list earlier about
getting matplotlib packaged for debian, and an RFP was made, but I
don't there has been any any progress on this front. Charles, Nathan,
any comments here?
 Nelson> The only things I had to install from sources were
 Nelson> TTFQuery and gdmodule, and both of those were easy.
 Nelson> I'd like to use this module as a simple thing to create
 Nelson> PNGs of graphs from data to put on web pages. The GD
 Nelson> driver is most of what I need, and the idea that I could
 Nelson> also do postscript output is great. But the packaging of
 Nelson> matplotlib is a bit funky. The most confusing thing was
 Nelson> the way you read sys.argv - that's probably not good for a
 Nelson> resuable library.
I agree to a certain extent. But I don't modify the args and they
will be without effect if you use the incantation, as you did
 import matplotlib 
 matplotlib.use('GD')
Mainly, I did it because I wanted an easy way to switch between
backends without switching scripts. There are other ways to do this
however. Unless the presence of the sys.argv is causing a problem, I
am hesitant to remove it to maintain backwards compatibility. If you
can think of a problem, please let me know.
 Nelson> I was also mystified as to how to set the plotsize using
 Nelson> the matlab module. I gave up and ended up doing something
 Nelson> else, but it was so much work I felt like I probably did
 Nelson> something wrong:
 Nelson> import matplotlib
 Nelson> matplotlib.use('GD')
 Nelson> from matplotlib.backends import Figure
 Nelson> from matplotlib.axes import Subplot
 Nelson> f = Figure(figsize=(3, 2), dpi=100)
 Nelson> a = Subplot(111)
 Nelson> a.plot([1,2,3,4], [1,4,9,16])
 Nelson> f.add_axis(a)
 Nelson> f.print_figure("nelson2.png", 100)
You're right on the money here. The ability to control the dpi,
figsize and other attributes from the matlab interface is limited. I
think I'll add kwargs to the fig command to enable this, and plan to
go to a backend specific config file for the next release -- a
.matplotlibrc file containing a dictionary that provided default font,
figsize, dpi, etc, for each backend. But beyond this, it is important
to be able to control these attributes at the script level using the
matlab interface.
I'm glad you figured out the workaround by using the API.
JDH

Sounds awesomely powerful - not since the development of John Harrison's clocks will navigation have taken such a great step forward :-)
----- Original Message -----
From: John Hunter <jdh...@ac...>
Date: 2003年11月16日 07:50:53 -0600
To: "Gary Ruben" <ga...@em...>
Subject: Re: [Matplotlib-users] Navigation toolbar redraw button
> >>>>> "Gary" == Gary Ruben <ga...@em...> writes:
> 
> Gary> I think it's important to always have an easy way of zooming
> Gary> out to show the full plot, so if you implement your saved
> Gary> view idea and don't want to add a 'fit to full data' button,
> Gary> I think having a 'fit to full data' checkpoint automatically
> Gary> added is important. It might be even be worth looking into
> Gary> cacheing the full view to speed its rendering. That way, you
> Gary> could quickly navigate by zooming to full and then using a
> Gary> rectangle zoom to your area of interest.
> 
> Combining my original thoughts with yours and Charles, then. How
> about the following? Allow the user to select the old style toolbar
> from a config file, but the new toolbar would
> 
> 1) lose all the left, right, up down buttons, in favor of a single
> 'hand' icon that allows you to move the view limits by 'grabbing
> them in the axes and moving them around. With x or y key pressed
> the pan motion will be contained to the respective axis. The
> mouse button which activates this in the axes should be easily
> configurable so as to not class with application developers.
> 
> 2) have a zoom in and zoom out button that work on both axis
> symmetrically unless the 'x' or 'y' modifier keys are pressed in
> which case work only on that axis. Allow similar functionality
> with the mouse in the axes, eg wheel mouse, as Charles suggested
> or mouse-3 as vtk interactor windows use. Then you could pan and
> zoom in an axes without ever removing your mouse from the axes
> lim. 
> 
> 3) have a zoom rectangle tool that allows you to zoom to a selected
> rectangle, constrained to maintain axis ratio if CTRL is pressed
> 
> 4) have a zoom to data button that changes viewlim to incorporate
> entire dataset
> 
> 5) use the 'views' navigation with forward / mark / and back buttons
> that allows you to navigate between views
> 
> 6) lose the redraw button, obsolete
> 
> 7) have a savefig button which allows you to set the dpi, filename
> and backend, so you can output with ps, gd, or gtk backend
> 
> 8) close window unchanged.
> 
> Further comments?
> 
> JDH
-- 
__________________________________________________________
Sign-up for your own personalized E-mail at Mail.com
http://www.mail.com/?sr=signup
Search Smarter - get the new eXact Search Bar for free!
http://www.exactsearchbar.com/

>>>>> "Jay" == Jay Eichelberger <mje...@ea...> writes:
 Jay> Greetings, Thank you for the reply.!!
Hi Jay, I suggest you join the matplotlib mailing list
http://lists.sourceforge.net/mailman/listinfo/matplotlib-users
 Jay> I started using Cygwin at work, where it is used to port Java
 Jay> code between Linux, solaris, and windows. ((
 Jay> http://www.onesaf.org/ )) I installed it at home on my laptop
 Jay> - which runs windows - wasn't able to find all of the Linux
 Jay> device drivers. Cyqwin has GCC compilers - which I used to
 Jay> write C program that called a public server that generated
 Jay> html and jpgs with charts.
 Jay> Alas, the service is no more.
 Jay> May I ask your opinion? Is there a simple way to do this?
 Jay> :-) After 5 hours of surfing and digging, it seems like I
 Jay> would need: - data for the charts - an api for a
 Jay> chart-generator - an output method - portable output
If I understand you correctly, yes this is simple. First, though, I
am not clear if you want to generate the charts and look at them
immediately, or if you want to run the script in batch mode and
produce figures which are not displayed, but are perhaps embedded in
some html for use with a web server. Both are possible but the way
you go about it is a little different. I'm going to assume the former
for now.
 Jay> Graph paper and a compass are starting to look good.. :-)
 Jay> Seriously, if I install Python, MatLab and GTK on windows,
 Jay> what would the chart output be??
I think you are referring to matplotlib, not matlab. matplotlib is an
open source python library for producing graphics with a matlab-like
syntax. matlab is a proprietary application developed by The
Mathworks.
If you install python, Numeric, pygtk, GTK and matplotlib following
the instructions on the matplotlib web site, it should be fairly easy
to do statistical charting with or without cygwin. That seems like a
lot of packages but it is easy since they all have friendly windows
installers.
If you have a fast internet connection and can do a 60MB download, I
recommend the enthought addition of python
http://enthought.com/python/ since it comes Numeric, scipy, wxpython
and a lot of extra goodies you'll find useful down the road. To
install GTK and pygtk, it is important that you follow the
instructions at http://matplotlib.sourceforge.net/installing.html;
basically you have to set a PATH.
Your prototype matplotlib script would be
 from matplotlib.matlab import *
 x, y = ... your data here ...
 plot(x,y)
 show()
This would open a GTK window which you can use to view, interact with,
and save your output. I recommend saving figures as PNG, which is a
nice, portable image format, if you want to put them on a web page or
send them to a colleague as an attachment, or PS for publication
quality.
Let me know in a little more detail what it is you are trying to
achieve and I (or someone else on the matplotlib-users mailing list)
can probably offer some more help.
JDH

>>>>> "Gary" == Gary Ruben <ga...@em...> writes:
 Gary> I think it's important to always have an easy way of zooming
 Gary> out to show the full plot, so if you implement your saved
 Gary> view idea and don't want to add a 'fit to full data' button,
 Gary> I think having a 'fit to full data' checkpoint automatically
 Gary> added is important. It might be even be worth looking into
 Gary> cacheing the full view to speed its rendering. That way, you
 Gary> could quickly navigate by zooming to full and then using a
 Gary> rectangle zoom to your area of interest.
Combining my original thoughts with yours and Charles, then. How
about the following? Allow the user to select the old style toolbar
from a config file, but the new toolbar would
 1) lose all the left, right, up down buttons, in favor of a single
 'hand' icon that allows you to move the view limits by 'grabbing
 them in the axes and moving them around. With x or y key pressed
 the pan motion will be contained to the respective axis. The
 mouse button which activates this in the axes should be easily
 configurable so as to not class with application developers.
 2) have a zoom in and zoom out button that work on both axis
 symmetrically unless the 'x' or 'y' modifier keys are pressed in
 which case work only on that axis. Allow similar functionality
 with the mouse in the axes, eg wheel mouse, as Charles suggested
 or mouse-3 as vtk interactor windows use. Then you could pan and
 zoom in an axes without ever removing your mouse from the axes
 lim. 
 3) have a zoom rectangle tool that allows you to zoom to a selected
 rectangle, constrained to maintain axis ratio if CTRL is pressed
 4) have a zoom to data button that changes viewlim to incorporate
 entire dataset
 5) use the 'views' navigation with forward / mark / and back buttons
 that allows you to navigate between views
 6) lose the redraw button, obsolete
 7) have a savefig button which allows you to set the dpi, filename
 and backend, so you can output with ps, gd, or gtk backend
 8) close window unchanged.
Further comments?
JDH

>>>>> "Gary" == Gary Ruben <ga...@em...> writes:
 Gary> Have you thought about whether it's possible to use the
 Gary> interactive gtk (or wx) viewer to zoom to a region of
 Gary> interest then change the backend to PS to save the current
 Gary> view as a PS or EPS? I have no idea whether this is
 Gary> possible.
I have thought about this. It would be nice to work in gtk/wx and then
be able to 'save as' postscript. This shouldn't be too hard. It's
basically a matter of initializing the postscript figure with the
axes contained in the gtk figure.
Thinking aloud in pseudo-py-code
 .. in the gtk print function ...
 if ext matches postscript:
 psfig = FigurePS(args)
 psfig._axes = self._axes
 psfig._text = self._text
 psfig._figurePatch = self._figurePatch
 psfig.print_figure(filename)
 return
I'll look into it. I think it is eminently doable. Barring any major
roadblocks I'll post a patch today. This reminds me: it would be nice
to be able to set the dpi from the print_figure dialog in the GUI as
you can with the savefig commmand.
BTW Gary, in CVS I have reworked the transform architecture and now
you can create lines (and patches) in data units (as before), relative
axes units (0-1) or physical units (points, inches, centimeters). So
I will make the necessary changes to the errorbar code to get the cap
lines specified in the appropriate units. My feeling is that they
should be in points (physical size) but the alternative is in
fractional axes size. The main difference between the 2 is that in
the latter case they change if you are, for example, subplot(221)
versus subplot(111), or if you resize the figure window, and in the
former case they do not. Thoughts?
JDH 

Have you thought about whether it's possible to use the interactive gtk (or wx) viewer to zoom to a region of interest then change the backend to PS to save the current view as a PS or EPS?
I have no idea whether this is possible.
Gary
-- 
__________________________________________________________
Sign-up for your own personalized E-mail at Mail.com
http://www.mail.com/?sr=signup
Search Smarter - get the new eXact Search Bar for free!
http://www.exactsearchbar.com/

My thoughts on navigation:
How about a button that toggles between 1x, 2x, 4x which affects the scaling factor of all zoom and pan buttons?
I think a combination of this, a 'fit to full data' button and a rectangle zoom tool would be most intuitive to me.
I think the zoom in and out should default to affecting both axes equally and have zooming axes indepedently as a special case.
If you implement a zoom rectangle, there should be a way of locking its width-height ratio to be the same as the current display, perhaps by holding CTRL down.
I think it's important to always have an easy way of zooming out to show the full plot, so if you implement your saved view idea and don't want to add a 'fit to full data' button, I think having a 'fit to full data' checkpoint automatically added is important. It might be even be worth looking into cacheing the full view to speed its rendering. That way, you could quickly navigate by zooming to full and then using a rectangle zoom to your area of interest.
Gary
-- 
__________________________________________________________
Sign-up for your own personalized E-mail at Mail.com
http://www.mail.com/?sr=signup
Search Smarter - get the new eXact Search Bar for free!
http://www.exactsearchbar.com/

>>>>> "Steve" == Steve Chaplin <ste...@ya...> writes:
 Steve> What is the "Redraw the figure" Navigation toolbar button
 Steve> used for? When I press it nothing (visible) happens. I
 Steve> guess it may be for redrawing the figure (as it says!), but
 Steve> thats not needed since (in GTK+) you setup an expose event
 Steve> which redraws the window automatically when required.
In earlier versions of matplotlib, the screen could get in an
inconsistent state following a pan or zoom (some residual effects of
ticklabels not properly erased, etc...). That has all been cured,
thankfully, and the redraw tool is a vestigial organ. Slated for
destruction.
 Steve> It would be nice to have a button to redraw the figure AND
 Steve> undo any zooming and panning you may have done. Which
 Steve> commands would I need to call to do that?
I've been planning on making some changes to the toolbar. How would
you feel about a 'save excursion / restore excursion'? As you
navigate through, you could click 'save excursion' and continue
navigating. At any point, when you click restore excursion it would
return you to the save point, or the original view if none clicked,
Ideally, you could click save excursion multiple times, and repeated
restores would step you back trough those points, to the beginning.
This is easy to implement.
Alternatively, there could be three buttons: back view, mark view, and
forward view, and as you navigate you could mark as many views as you
want, and then use forward and back to go between them. Eg, for the
stock_demo, where I have 60 days of stock prices for IBM and Apple,
you could mark several days or hours in that whole 60 days, and then
use the forward and back tools to switch between them. If nothing is
marked, back would simply revert you to the original view. I think
this latter scheme sounds more useful, but potentially adds too many
buttons.
Other tools I've been considering
 -- add a key modifier to the pan tool so CTRL pans an entire view.
 Again, suppose for the stock demo you had 1 day in the view, with
 ticks every hour. Use pan right to pan hour by hour and
 CTRL-right to pan day by day
 -- add a key modifier so a zoom affects both axes, eg, SHIFT-zoom
 zooms both x and y
-- add a rectangle zoom tool, so you can select a rectangular region
 with your mouse and set that to be the x and y view limits.
The primary application I use matplotlib in is an EEG viewer, so I
have a lot of navigation requirements.
If anyone has any special navigation requests, weigh in, because I'll
probably try and get some or all of these features before the next
major (0.4x) release early next month.
JDH