matplotlib-users Mailing List for matplotlib

John Hunter wrote:

> I do statically link png, freetype and zlib for win32 so that users
> don't have to install any of those prereqs.



Thanks John, I have a plan now.

I have a question, however. How do I get distutils to statically link a 
given lib? the only method I have now is to make sure that only the 
static lib is available. I can do that (for the moment) with libpng, but 
for freetype, I have a problem. In setupext.py, there is:

basedir = {
     'darwin' : ['/usr/local', '/usr', '/sw', '/usr/X11R6'],
}

There is a libfreetype in /usr/X11R6, but only if X11 is installed, so I 
don't want to use it. However, I imagine that /usr/X11R6 is required to 
the GTK build, so I can't take that out. Any ideas?

I notice that for win32, you have only:

basedir = {
     'win32'  : ['win32_static',],

Should I do that for darwin, and put my static libs in there?

I'm off to try some of this.


-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...

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

    Chris> The question is: Can I build a binary version that will
    Chris> work with all the back ends, if they are there, but also
    Chris> work on what is there if they aren't all. This comes down
    Chris> to when the linking happens, I guess. Experimentation will
    Chris> answer this for me, but I welcome and insight any of you
    Chris> may have.

    Chris> One issue that has come up is libfreetype. If you have GTK,
    Chris> you have X11, so you have libfreetype. Indeed, I don't have
    Chris> GTK, but I do have X11, and the X11 freetype got linked to
    Chris> my package. I can probably fix this by removing
    Chris> "/usr/X11R6" from the basedir list for darwin. Then I'll
    Chris> have libfreetype statically linked to matplotlib, but
    Chris> PyGTK, if the user has it, will have a different version
    Chris> dynamically linked. Is this going to cause problems?

Here is what I do for the win32 which has some related issues.

matplotlib needs Tkinter and pygtk to be installed on your build
system in order to build the matplotlib extension code for those
modules.  But you do not need to statically link in any gtk or tk
library code.

I do statically link png, freetype and zlib for win32 so that users
don't have to install any of those prereqs.

Then they get a package that has Agg, PS, SVG out of the box
regardless of their system.  If they have pygtk on their system, they
also get GTK and GTKAgg.  If they have Tkinter, they also get
TkAgg. If they have wxpython, they also get WX, WXAgg.

The key is, if you want your package to be maximally usable, you
should have pygtk and tk (and the requisite devel stuff) and should
probably statically link freetype, png and zlib.

Of course there is now also a Cairo and GTKCairo backend, which may
interest some OSX X11 users, but I haven't had to worry about this for
win32 users.

And FLTK and QT ..., but we can save that for another day.  I think if you
have the tkagg extension compiled in, most OSX people will be happy
with either that or wxagg.

JDH

Robert Kern wrote:
> Chris Barker wrote:
>> However, that doesn't seem to work if I have libfreetype.dylib 
>> somewhere standard, and I certainly don't want to remove it! (Maybe I 
>> could temporarily, but that's hardly the robust solution I'm looking for)
> 

> Change the paths that distutils will add to the link line. They're at 
> the top of setupext.py . Remove the ones you don't need.

Except that the *.a and *.dylib are put in the same place. Darn.

>> By the way, it would also be good to get this to work with TK and/or 
>> GTK. Has anyone done that successfully that would like to help out 
>> with this?

> With TclTkAqua, it Just Works.

Someone posted recently to the MacPython Mailing list with a series of 
questions that this was not the case, at least to for him. So I wasn't sure.

>> Also, as far as PyGTK is concerned. Can you run it without running 
>> Fink? That's the only way I've seen it done. If it is a Fink only 
>> option, then this is moot, as I'm looking for a Fink free approach, 
>> and someone else has put together a fink matplotlib package.

> You could try Darwinports. Set Darwinports' prefix to /usr/local, and 
> use port(1) to make Installer.app packages for GTK et al. Bundle them 
> with your bdist_mpkg metapackage.

Hmm. That sounds promising. However...

My immediate need to to get matplotlib installed on and OS-X Server box 
that someone else is administering. For this purpose, I only need the 
Agg back end. However, for my own purposes, I also want wx. The easiest 
way for me to accomplish my immediate goal is probably to just install 
all the libs needed on that server, but I'm trying to keep it simple for 
the admin, and I figured this could be a useful contribution to the 
MacPython community.

However, here is the complication. As long as I am doing this, I'd like 
to make it as useful as possible. That means it would be nice if I could 
come up with a single binary that works with all the available back ends 
on OS-X (Agg, wx, PyGTK, Tkinter). I also want people to be able to 
install it on a stock box and have it work. This doesn't mean that I 
want it to include PyGTK and TK, but it would be nice if it worked with 
those, if they were present.

The question is: Can I build a binary version that will work with all 
the back ends, if they are there, but also work on what is there if they 
aren't all. This comes down to when the linking happens, I guess. 
Experimentation will answer this for me, but I welcome and insight any 
of you may have.

One issue that has come up is libfreetype. If you have GTK, you have 
X11, so you have libfreetype. Indeed, I don't have GTK, but I do have 
X11, and the X11 freetype got linked to my package. I can probably fix 
this by removing "/usr/X11R6" from the basedir list for darwin. Then 
I'll have libfreetype statically linked to matplotlib, but PyGTK, if the 
user has it, will have a different version dynamically linked. Is this 
going to cause problems?

More as I make more progress...

-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...

>>>>> "Travis" == Travis Brady <td...@fa...> writes:

    Travis> Hi Everyone, I'm using plot_date to scatter a ton of data
    Travis> retrieved from a database and have encountered an error
    Travis> with the date ticks.

    Travis> I more or less copied the format of the example in
    Travis> http://matplotlib.sourceforge.net/examples/date_demo2.py
    Travis> and my ticks ended up being months in order but never
    Travis> making it past the year 2001 (they just recycle) though my
    Travis> data begin in 2000 and end in 2004.

    Travis> I took another look at the plot generated by date_demo2.py
    Travis> and it turns out that the same is true there.

    Travis> I'm thinking this is not intentional, but if so does
    Travis> anyone have any advice or example scripts that might help
    Travis> me fix my plot?

That's because the format string is printing 'month day' and not
'month year' and the day is always 01 in the example.  You might want

monthsFmt = DateFormatter("%b '%y")

Hope this helps,
JDH

> On Wed, 2005-02-02 at 10:14 -0800, matplotlib-users-
> re...@li... wrote:
> > On Wednesday 02 February 2005 04:51 am, Steve Chaplin wrote:
> > >
> > > When resizing the FigureCanvas on my system I notice:
> > > GTK   sets a FigureCanvas size that can only be increased.
> > > QtAgg sets a FigureCanvas size that can be increased or decreased
> > > TkAgg sets a FigureCanvas size that can be increased or decreased to a
> > > minimum (of perhaps 400 x 300 pixels)
> > > Wx I do not have, is it like Qt or Tk or is there a 4th variation?
> > 
> > on linux, wx lets me increase or decrease all the way. I wonder if it is 
> > different on windows.
> > 
> > >
> > > I looked at changing GTK so the FigureCanvas could be decreased a while
> > > ago but the problem is that while the axes scale/shrink nicely the text
> > > stays the same size and becomes misaligned or clipped and gives an odd
> > > looking graph. Perhaps thats why Tk allows you shrink the window but
> > > only to a certain size.
> > >
> > > Matplotlib currently has 3 (or more) ways of handling window resizing,
> > > which one of these behaviours is 'correct' or most desirable?
> > >
> > 
> > I think the most desirable for interactive use would be to allow scaling, but 
> > if it doesn't scale properly, I think the current behavior is appropriate. I 
> > guess I just didnt notice it before now.
> > 
> > Darren
I've changed the GTK backends (in cvs) so they will now allow the window
to be resized smaller than the original size. Text still does not scale
down when the window shrinks, but if the matplotlib user wants a smaller
window they now have the option to resize the window.

Steve

Hola

When i try to fit a polynomial of order 4 to 5 data
points, I get an answer far from what I would expect.
Try and run the attached script an see if you can fit
the datapoints in x and y better than me.
When I run the script, the fit doesn't go through any
of the points.
The problem is also present for higher order pol.

Kristen

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

And while we're on the subject of OS X builds, I have a request.

Can you add /usr/local/share/matplotlib to the list of default locations 
search for the data directory?

/System/Library/Frameworks/Python.framework/Versions/Current/share/matplotlib 
is not, I think a good place for it.

-- 
Robert Kern
rk...@uc...

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

Chris Barker wrote:

> However, that doesn't seem to work if I have libfreetype.dylib somewhere 
> standard, and I certainly don't want to remove it! (Maybe I could 
> temporarily, but that's hardly the robust solution I'm looking for)

Change the paths that distutils will add to the link line. They're at 
the top of setupext.py . Remove the ones you don't need.

> By the way, it would also be good to get this to work with TK and/or 
> GTK. Has anyone done that successfully that would like to help out with 
> this?

With TclTkAqua, it Just Works.

> Also, as far as PyGTK is concerned. Can you run it without running Fink? 
> That's the only way I've seen it done. If it is a Fink only option, then 
> this is moot, as I'm looking for a Fink free approach, and someone else 
> has put together a fink matplotlib package.

You could try Darwinports. Set Darwinports' prefix to /usr/local, and 
use port(1) to make Installer.app packages for GTK et al. Bundle them 
with your bdist_mpkg metapackage.

My strategy for building matplotlib (and I've done it *a lot* in the 
past few weeks) is as follows:

I have Darwinports with a prefix in a GNU Stow repository. What Stow 
does is it allows you to install stuff into it's own directory 
(/usr/local/stow/darwinports, which has bin/, lib/, share/ et al.) and 
then makes symlinks such that everything appears to be installed to 
/usr/local. So I have Darwinports install libpng and libfreetype. I have 
a script that will remove the symlinks to the dylibs for libpng, 
libfreetype, and libz (I could probably resolve this by changing the 
order of search). I build matplotlib and double-check the dylib 
dependencies with "otool -L". I do not bother with GTK at this time.

-- 
Robert Kern
rk...@uc...

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

On Feb 3, 2005, at 19:07, Chris Barker wrote:

> I posted a note here the other day about a matplotlib binary. I've 
> discovered a problem with it. Apparently, it has linked against the 
> libfreetype in /usr/X11. That means it will work fine on a system with 
> Apple's X11 installed (like mine or the first system I tested it on).

.. and the system that I tested it on :)

> So, the question remains: How to I get distutils to statically link 
> selected libs? Bob had suggested:
>
> """Make sure you don't have libpng.dylib or libfreetype.dylib sitting 
> around on your link paths...
>
> Build them like this:
> ./configure --disable-shared --enable static
>
> Make sure you only have libpng.a and libfreetype.a on your link paths. 
> It will link statically.
> """
>
> However, that doesn't seem to work if I have libfreetype.dylib 
> somewhere standard, and I certainly don't want to remove it! (Maybe I 
> could temporarily, but that's hardly the robust solution I'm looking 
> for)

You have three options:
- make sure the dylibs aren't sitting around in the same location as 
the .a, and make sure the .a's path is preferred (you could make a 
local dir, -L to it, symlink the archives in)
- change the setup.py relatively significantly and specify them as 
object files with their full path (I can't tell you off the top of my 
head what the distutils incantation would be for this, but you 
definitely can not use it's "libraries" option, or whatever that is 
called)
- don't link statically and install the dylibs too, somehow (possibly 
by building+installing them with 
--prefix=/Library/Python/2.3/matplotlib/lib-dynload/ and -L to that 
when linking.. note that this would require a bdist_mpkg subclass to 
make a .pkg out of it)

-bob

Hi all,

I posted a note here the other day about a matplotlib binary. I've 
discovered a problem with it. Apparently, it has linked against the 
libfreetype in /usr/X11. That means it will work fine on a system with 
Apple's X11 installed (like mine or the first system I tested it on).

So, the question remains: How to I get distutils to statically link 
selected libs? Bob had suggested:

"""Make sure you don't have libpng.dylib or libfreetype.dylib sitting 
around on your link paths...

Build them like this:
./configure --disable-shared --enable static

Make sure you only have libpng.a and libfreetype.a on your link paths. 
It will link statically.
"""

However, that doesn't seem to work if I have libfreetype.dylib somewhere 
standard, and I certainly don't want to remove it! (Maybe I could 
temporarily, but that's hardly the robust solution I'm looking for)

By the way, it would also be good to get this to work with TK and/or 
GTK. Has anyone done that successfully that would like to help out with 
this?

Also, as far as PyGTK is concerned. Can you run it without running Fink? 
That's the only way I've seen it done. If it is a Fink only option, then 
this is moot, as I'm looking for a Fink free approach, and someone else 
has put together a fink matplotlib package.

By the way the package, as it stands, can be found at:

http://undefined.org/python/matplotlib-0.71-py2.3-macosx10.3.zip

thanks to Bob Ippolito

-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...