matplotlib-users Mailing List for matplotlib

Hi all,

I just got matplotlib working on my box, so I thought I'd post this, so 
that it will be in the archives, and because I'd like others to try it 
and let me know if it doesn't work for you.


One question for the knowledgeable folks here:

I have /usr/include/zlib.h on my system, but I have no idea where I got 
it. I do know that it isn't on other systems in my office without the 
dev tools. Did it come with the dev tools, or something else I installed?

Anyway, here's what worked for me:

Installing matplotlib on OS-X (10.3.7)

Here are my notes as to what it took to get matplotlib (0.71) installed 
and working on OS-X. I have so far kept a fink-free system, so that's 
what I've done here as well. I use it with the AGG back end for 
generating images for a web site, and hopefully with the wx backend for 
interactive use and embedding in wx Applications.

1) Requirements:
-------------------------------------------
According to the matplotlib install docs 
(http://matplotlib.sourceforge.net/installing.html), you need the following:

freetype (>= 2.1.7)
libpng
zlib

Personally, I've been avoiding Fink, as it doesn't seem to play well 
with the rest of OS-X, including the Apple supplied Python, so I've 
looked elsewhere for these libs.

a) Freetype:
I seem to have it in:

/usr/X11R6/include/freetype2

I don't think I installed it myself, so it probably came with Apple's 
X11, Which I did install.

However, I seemed to be having problems with that version, so I looked, 
and it seems to be:

libfreetype.6.3.dylib

Given that freetype2 is currently at version 2.1.9, I have no idea what 
to make of that! So off to sourceforge to get a new freetype:

http://freetype.sourceforge.net/index2.html

where I got:

freetype-2.1.9.tar.gz

Following the instructions in docs/INSTALL.UNX:

$ ./configure
$ make
$ sudo make install

That puts it in /usr/local/..., which is a good place for it.

b) zlib:
I have:
/usr/include/zlib.h

I don't know where I might have gotten it, but there it is.

NOTE: I checked on another system in my office, and it doesn't have 
zlib. It also doesn't have any developer tools installed, so you 
couldn't build anything on it anyway. If zlib comes with XCode tools, 
anyone who can build should have it. It also may have come with some 
other package I've installed, but I have no idea what. Let me know if 
you know.

c) libpng:
This, I didn't have on my system, except inside the wxWidgets source 
tree, so I went looking for it.

Note that you need zlib to compile libpng, so make sure you have that first.

I did a google search for "libpng OS-X". I found:

http://www.libpng.org/pub/png/pngcode.html

Which led me to the libpng sourceforge site.

 From there I downloaded:

libpng-1.2.8.tar.gz

unpacked it, and opened a terminal in the libpng-1.2.8 directory, and did:

$ cp scripts/makefile.darwin ./makefile

(note that according to the INSTALL, there is supposed to be a 
makefile.macosx, but it wasn't there)

I took a look in the makefile, and found:

ZLIBLIB=/usr/local/lib
ZLIBINC=/usr/local/include

Which is not where zlib is on my system. However, while I can find 
zlib.h, I couldn't fine the actual lib, so I tried make without changing 
anything.

$ make

Which seemed to work fine. zlib must be installed in a standard 
location, and gcc  found it.

$ sudo make install

to install the lib into /usr/local/ (this was specified in the makefile, 
and it's a good place for it)

$ make test
and
$ ./pngtest pngnow.png

Which both seemed to pass.

d) Numeric or numarray:

I already have Numeric installed, from Bob Ippolito's PIMP (MacPython 
Package Manager) repostitory. (www.undefined.org/python)

2) Building matplotlib
---------------------------------------------------------
First, I took a look at the matplotlib setup.py. It had "auto" for the 
back-end flags, so I thought I would give it a try that way:

$python setup.py build

That worked!
$ sudo python setup.py install

4) And now to test!

$python

 >>> import pylab
Could not open font file /Library/Fonts/NISC18030.ttf
No module named pygtk
PyGTK version 1.99.16 or greater is required to run the GTK Matplotlib 
backends

This turns out to be because the matlabrc file sets the GTKAgg back end 
as the default. You have two choices.

1) you can set the back end before importing pylab.
 >>> import matplotlib
 >>> matplotlib.use('Agg')
 >>> import pylab

This works fine

2) Edit the matplotlibrc file. I found it in:
/System/Library/Frameworks/Python.framework/Versions/2.3/share/matplotlib/.matplotlibrc

Change the line:
backend      : GTKAgg    # the default backend
to
backend      : Agg    # the default backend

And you're all set to make images for the web, etc.

I'm  going to leave getting it to work with wxPython for another day.

-Chris


-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

John Hunter wrote:
>     Chris> How do I save an AGG figure with the OO interface?

> The distinction between the figure and
> the canvas was made to fully separate the front end from the backend.

Which is a good idea.

 > here the pure OO way:

>     from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
>     from matplotlib.figure import Figure
> 
>     fig = Figure()
>     canvas = FigureCanvas(fig)

So this is really the only extra call. not bad.

>     ax = fig.add_subplot(111)
>     ax.plot([1,2,3])
>     ax.set_title('hi mom')
>     ax.grid(True)
>     ax.set_xlabel('time')
>     ax.set_ylabel('volts')
>     canvas.print_figure('test')

> Hmm, I don't see the logic of being willing to use some pylab commands
> (pylab.figure) but not others (pylab.savefig).

Well, the figure constructor has to be in some namespace.

pylab.savefig, however, bothers me for two reasons:

1) It seems to me that it belongs as a figure (or canvas, I suppose) method.

2) Most critically, it uses the  "current figure". I can see where this 
makes sense with Matlab style interactive use, but in a program, I may 
be building more than one figure at a time, and it makes more sense to 
me to do something like:

fig1.save("plot1")
fig2.save("plot2")

than:
set_current_figure(1)
pylab.savefig("plot1")
set_current_figure(2)
pylab.savefig("plot2")

 > But, if you really
> hate savefig, you can call print_figure by getting your hands on the
> current canvas

>     manager = pylab.get_current_fig_manager()

Ah, the "current"  concept again!

> As we discussed earlier, it might be useful for the figure to store a
> ref to it's canvas, then we could define fig.savefig which forwards
> the call to canvas.print_figure.  Would make for a cleaner interface
> for folks who love '.'

Yes, I confess, I love ".". That does sound like a good idea to me.

Sorry to come off as being critical. I am very impressed with your work 
on matplotlib. There are been many promising starts to a python plotting 
package, and NONE of them have come even close to what you've done with 
matplotlib. That's why I'm using it. I also was a big fan and user of 
Matlab for years, I never would have gotten my dissertation done without 
it. However, I like Python as a language far more than Matlab, and with 
SciPy and matplotlib, we're getting close to having Matlab's wealth of 
functionality in Python. My ideal is that I have a plotting package that 
really feels Pythonesque. Matplotlib is actually pretty close. I think 
adding a few things, and making more docs and examples, and we'll have 
just what I've been wanting.

I hope as I get more familiar with matplotlib, I'll start to contribute 
code, rather than just questions and criticism.

John, do you want patches that add-to/improve the OO interface?

-Chris


-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

                Hi,

when I'm trying to play with the different sample script, I saw a bug 
with the polar plot. The zoom function provide by matplotlib is not 
working. When you are trying to zoom at a certain place, the result is 
in another place of the plot, probably not random but it seems at first.

Nicolas

>>>>> "Chris" == Chris Barker <Chr...@no...> writes:

    Chris> Thanks John and Andrew.  Now I've changed my matplotlibrc
    Chris> to use the Agg backend by default. It would be nice if the
    Chris> default could be set appropriately on install, but I have
    Chris> no idea how one would do that. However, I doubt that GTKAgg
    Chris> is the default on Windows, so maybe there is a way!

I just change the default .matplotlibrc value when I build the windows
installer to tkagg.  There's nothing automatic.

It would be nice if the setup.py script were smarter, and provided a
custom rc file depending on what it finds at build time (numeric vs
numarray, default backend) but there has been enough work to do just
getting core plotting functionality going (most recently contour and
polar) that configuration issues have been on the back burner.

JDH

>>>>> "Chris" == Chris Barker <Chr...@no...> writes:

    Chris> Hi all, There has been some recent discussion about perhaps
    Chris> improving the OO interface. So Here's a question:

    Chris> How do I save an AGG figure with the OO interface?

Hi Chris,

Here is a canonical script to create an Agg canvas/figure and save it
using the pure OO interface.  The distinction between the figure and
the canvas was made to fully separate the front end from the backend.
The Figure is the abstract object that stores all the information
about the figure, the canvas is where the ink goes.  I agree there
could be some changes to the interface to make this easier, but here
the pure OO way:

    from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
    from matplotlib.figure import Figure

    fig = Figure()
    canvas = FigureCanvas(fig)
    ax = fig.add_subplot(111)
    ax.plot([1,2,3])
    ax.set_title('hi mom')
    ax.grid(True)
    ax.set_xlabel('time')
    ax.set_ylabel('volts')
    canvas.print_figure('test')


    Chris> For example:
    >>>> fig = pylab.figure() ax = fig.add_subplot(1,1,1)
    >>>> ax.plot(range(10), pylab.sin(range(10)))

    Chris> Now how do I save it as a PNG? I can do:

    >>>> pylab.savefig("test.png")

    Chris> But then that is the interface I'm trying to get away from.

Hmm, I don't see the logic of being willing to use some pylab commands
(pylab.figure) but not others (pylab.savefig).  But, if you really
hate savefig, you can call print_figure by getting your hands on the
current canvas

    import matplotlib
    matplotlib.use('Agg')
    import pylab

    fig = pylab.figure()
    ax = fig.add_subplot(111)
    ax.plot([1,2,3])

    manager = pylab.get_current_fig_manager()
    manager.canvas.print_figure('test')

Kind of awkward.  This is basically what savefig does.

    Chris> Another option would be:

    >>>> fig.savefig

    Chris> But that's not there either.

As we discussed earlier, it might be useful for the figure to store a
ref to it's canvas, then we could define fig.savefig which forwards
the call to canvas.print_figure.  Would make for a cleaner interface
for folks who love '.'

JDH

Thanks John and Andrew.

Now I've changed my matplotlibrc to use the Agg backend by default. It 
would be nice if the default could be set appropriately on install, but 
I have no idea how one would do that. However, I doubt that GTKAgg is 
the default on Windows, so maybe there is a way!

-Chris



-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Pressing "send" too soon...

To partially answer my own question, I looked in matplotlib.pylab.py, 
and found:

manager = get_current_fig_manager()
manager.canvas.print_figure(*args, **kwargs)

However, I can't see how to get the canvas from a figure instance. This 
looks like one place the OO interface could be improved.

-Chris





-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Hi all,

There has been some recent discussion about perhaps improving the OO 
interface. So Here's a question:

How do I save an AGG figure with the OO interface?

For example:
 >>> fig = pylab.figure()
 >>> ax = fig.add_subplot(1,1,1)
 >>> ax.plot(range(10), pylab.sin(range(10)))

Now how do I save it as a PNG? I can do:

 >>> pylab.savefig("test.png")

But then that is the interface I'm trying to get away from.

Looking at the docs, it looks like the agg back-end should have:
 >>> fig.print_figure("Test.png")
Traceback (most recent call last):
   File "<stdin>", line 1, in ?
AttributeError: Figure instance has no attribute 'print_figure'

But no such luck.

Another option would be:

 >>> fig.savefig

But that's not there either.

Thanks,
-Chris

-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

>>>>> "Andrew" == Andrew Straw <str...@as...> writes:

    Andrew> What does your .matplotlibrc say on the "backend" line?

To flesh out Andrew's answer a bit, see

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

and 

  http://matplotlib.sourceforge.net/backends.html

Cheers,
JDH

Chris Barker wrote:

> Hi all,
>
> I'm working on installing matplotlib on a fink-free OS-X. When I have 
> it all working, I'll post the methodology here.
>
> I've got it all compiled and installed, but when I run it, I get:
>
> Python 2.3 (#1, Sep 13 2003, 00:49:11)
> [GCC 3.3 20030304 (Apple Computer, Inc. build 1495)] on darwin
> Type "help", "copyright", "credits" or "license" for more information.
> >>> import pylab
> Could not open font file /Library/Fonts/NISC18030.ttf
> No module named pygtk
> PyGTK version 1.99.16 or greater is required to run the GTK Matplotlib 
> backends
>
> This makes sense, as I don't have PyGTK, However, I would have thought 
> that setup.py would have figured that out, and not tried to build it. 
> At this point, all I need is to generate images with the AGG back-end, 
> so why is it trying to run PyGTK on import?


What does your .matplotlibrc say on the "backend" line?

Hi all,

I'm working on installing matplotlib on a fink-free OS-X. When I have it 
all working, I'll post the methodology here.

I've got it all compiled and installed, but when I run it, I get:

Python 2.3 (#1, Sep 13 2003, 00:49:11)
[GCC 3.3 20030304 (Apple Computer, Inc. build 1495)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
 >>> import pylab
Could not open font file /Library/Fonts/NISC18030.ttf
No module named pygtk
PyGTK version 1.99.16 or greater is required to run the GTK Matplotlib 
backends

This makes sense, as I don't have PyGTK, However, I would have thought 
that setup.py would have figured that out, and not tried to build it. At 
this point, all I need is to generate images with the AGG back-end, so 
why is it trying to run PyGTK on import?

If I do:
 >>> import matplotlib
 >>> matplotlib.use('Agg')
 >>> import pylab

Then it works OK.

It would be nice to just do an "import pylab", however.

-Chris







-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

John Hunter wrote:
> Well, there are two exceptions to this: set operates on sequences or
> instances, and can handle multiple key/val pairs.  So you can do
> 
> set(labels, rotation=45, fontsize=14, color='red')
> 
> in place of 
> 
> for label in labels:
>    label.set_rotation(45)
>    label.set_fontsize(14)
>    label.set_color('red')
> 
> I prefer the former, particularly when working from the python shell.
> But it's a matter of taste...

That is handy. Actually, I really like it when you can work with sets of 
objects, rather than doing the explicit looping all the time. It just 
goes to show you that OO design is not always the best answer. Now the 
question is: What namespace should set() live in? pylab.set() works for me.

I'm now working on a Fink-free OS-X install, which is proving to be 
quite a chore. I'll report here when I'm done.

-Chris


-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

>>>>> "Delbert" == Delbert D Franz <dd...@lk...> writes:

    Delbert> I have made great progress with my GUI to plot
    Delbert> time-series files from my unsteady-flow modeling
    Delbert> software.  It works when I have one subplot per figure
    Delbert> but I have not been able to get two subplots (211 and
    Delbert> 212) to work.  Only the lower subplot appears and the
    Delbert> trace or line assigned to the upper subplot (211) appears
    Delbert> in the lower subplot.  The space for the upper subplot
    Delbert> appears as blank space in the figure.

You code looks correct on first glance (except did you mean tht title
to be in the loop?) .  My guess is your data structure has a bug in
it.  Add a print statement and make sure the loc code is as you think
it is.

        print fg.fgs[ifg].sp[isp].loc_code
        ax = f.add_subplot(fg.fgs[ifg].sp[isp].loc_code)  

If they are correct, you need to compose a minimum, free standing
script that replicates your problem and I can take a look.

Hope this helps,
JDH

I have made great progress with my GUI to plot time-series files from
my unsteady-flow modeling software.  It works when I have one subplot
per figure but I have not been able to get two subplots (211 and 212)
to work.  Only the lower subplot appears and the trace or line assigned to 
the upper subplot (211) appears in the lower subplot.  The space for the 
upper subplot appears as blank space in the figure.  

I'm a little unsure of how to embed multiple subplots using Tk.  I 
based what I did on examples using pylab and tried to translate them
to my case.  Here's  the code that does the work.  The fg.fgs..., is 
a data structure that holds various descriptive items for the figure,
subplot, and the various traces.  The case that fails has 2 subplots
with one trace per subplot. file_list holds various file-related objects
that get the data from one or more files when it is needed. 

    ifg = 0
    win = Toplevel(bd=5, relief=RIDGE)
    win.title(fg.fgs[ifg].name)

    # Try to setup useful time-series axis
    days = DayLocator(range(1,32))
    hours = HourLocator((6,12,18))
    timeformat = DateFormatter('%d %b %y')
    
    w = int(fg.fgs[ifg].width)
    h = int(fg.fgs[ifg].height)
    f = Figure(figsize=(w, h), dpi=120)
    nsp = len(fg.fgs[ifg].sp)
#    ax = []
    for isp in xrange(nsp):
        ax = f.add_subplot(fg.fgs[ifg].sp[isp].loc_code)  
        ntr = len(fg.fgs[ifg].sp[isp].tr)
        for itr in xrange(ntr):
            i = fg.fgs[ifg].sp[isp].tr[itr].file_index
            jt_0, f_0 = file_list[i].get_ts_segment()        
            jt_0 += 678576.0
            ax.plot_date(jt_0, f_0,  fg.fgs[ifg].sp[isp].tr[itr].plot_code)
    
        ax.xaxis.set_major_locator(days)
        ax.xaxis.set_minor_locator(hours)
        ax.autoscale_view()

        labels = ax.get_xticklabels()
        ax.xaxis.set_major_formatter(timeformat)
        set(labels, rotation=85)


        ax.set_xlabel(fg.fgs[ifg].sp[isp].xlabel)
        ax.set_ylabel(fg.fgs[ifg].sp[isp].ylabel)
        ax.grid(True)

    ax.set_title(fg.fgs[ifg].title)

    canvas = FigureCanvasTkAgg(f, master=win)
    canvas.show()
 
    canvas.get_tk_widget().pack(side=TOP, fill=BOTH, expand=1)
  
 
    toolbar = NavigationToolbar( canvas, win )
    toolbar.update()
    canvas._tkcanvas.pack(side=TOP, fill=BOTH, expand=1)
    canvas._tkcanvas.pack(side=TOP,  expand=1)

    canvas.print_figure('test.ps', dpi=300, orientation='landscape')


I'm finding that Python and matplotlib are fun to use but I'm still low 
on the learning curve.  However, I already have a usable package and can 
now start adding refinements and extensions. 

                             Thanks, 
                               Delbert Franz