matplotlib-users Mailing List for matplotlib

On Wed, Jan 19, 2011 at 10:16 AM, Benjamin Root <ben...@ou...> wrote:
>
>
> On Tue, Jan 18, 2011 at 5:32 PM, Alex Liberzon <ale...@gm...>
> wrote:
>>
>> Hi,
>>
>> While moving from Matlab to Numpy/Scipy/Matplotlib I need sometimes to
>> work with Matlab figures. It would be nice if we could load Matlab figures
>> in Python, extract the data and some basic parameters of how it was looking
>> in Matlab and create its "clone" in matplotlib. At present the Matlab
>> figures are MAT files and in the near future it will be HDF5, both formats
>> are loadable. However, I have not found any reference for such attempts to
>> load and parse the FIG files. As a beginner I find it difficult to write a
>> large piece of code. However, if there are other  interested users that can
>> cooperate on such, I'd gladly contribute some hours for this task.
>> Meanwhile, to show the proof-of-concept attempt is attached below. All your
>> useful comments and suggestions are very welcome.
>>
>> Thank you,
>> Alex
>>
>
> Alex,
>
> That is very interesting.  I was not aware that matlab's figure files were
> simply .mat files in disguise.  I would presume that it would be feasible to
> produce some sort of importer in such a case (provided the documentation for
> Matlab's figure format is complete enough).
>
> I am wary of making such a function a core feature of Matplotlib, however,
> because it would require creating a dependency to the scipy.io package.
> However, I could see it being a toolkit package like Basemap.
>
> Would you mind creating a feature request ticket at:
>
> http://sourceforge.net/tracker/?group_id=80706
>
> If you include this example code, and maybe some links to some documentation
> on matlab's figure files, maybe something can grow from that.
>
> Thanks!
> Ben Root

I agree, but for a slightly different reason.  I think it would be
great if it would be possible to incorporate viewing matlab figures
into the rest of my system, so making it more isolated would help in
this regard.

-Todd

On Wed, Jan 19, 2011 at 10:20 AM, todd rme <tod...@gm...> wrote:

> On Wed, Jan 19, 2011 at 10:22 AM, Benjamin Root <ben...@ou...> wrote:
> >
> >
> > On Wed, Jan 19, 2011 at 4:48 AM, Jochen Deibele <joc...@nt...>
> > wrote:
> >>
> >> Hi!
> >>
> >> todd rme wrote on 19.01.2011 04:40:
> >>  > The class, on the other hand, seems to be an application-specific
> >>  > description of what the role is within the application.  I think this
> >>  > should probably be "figure".
> >>
> >> Just an idea: What about exposing this (or maybe all) of the values in
> >> the api to be set dynamically by the user? Of course providing a useful
> >> default value as well.
> >>
> >> The thought behind is that it's hard to guess in what context the user
> >> wants to use the figure. And perhaps the user has 21 plots of some type
> >> and one which is different and he wants to treat differently.
> >> But at the same time - you also could do this if you adjust the window
> >> title.
> >>
> >> Jochen
> >>
> >
> > I agree.  If we are going to do this, we might as well do it right.
> Anybody
> > who isn't using this feature will not see a difference, but those who do
> > manage their windows with rules can benefit significantly from just
> having
> > sensible default values and from having the ability to explicitly set the
> > value.  Therefore, this should probably be something that is added to the
> > rcParams and the user can specify a value by keyword argument to the
> Figure
> > __init__ method.
> >
> > Ben Root
>
> That is far outside the scope of my capabilities.  I already have wx
> and gtk backends set up (although it needs to be tested) with
> hard-coded values, and qt4 half-working (and tested, this doesn't
> appear to be possible in qt).  I can submit patches for that, and if
> someone else wants to make it configurable they can do so.  I'll set
> it up to use variables that you can then do what you want with.
>
> -Todd
>
>
That would be a good start.  I have no objections to starting off with that.

Ben Root

On Wed, Jan 19, 2011 at 10:22 AM, Benjamin Root <ben...@ou...> wrote:
>
>
> On Wed, Jan 19, 2011 at 4:48 AM, Jochen Deibele <joc...@nt...>
> wrote:
>>
>> Hi!
>>
>> todd rme wrote on 19.01.2011 04:40:
>>  > The class, on the other hand, seems to be an application-specific
>>  > description of what the role is within the application.  I think this
>>  > should probably be "figure".
>>
>> Just an idea: What about exposing this (or maybe all) of the values in
>> the api to be set dynamically by the user? Of course providing a useful
>> default value as well.
>>
>> The thought behind is that it's hard to guess in what context the user
>> wants to use the figure. And perhaps the user has 21 plots of some type
>> and one which is different and he wants to treat differently.
>> But at the same time - you also could do this if you adjust the window
>> title.
>>
>> Jochen
>>
>
> I agree.  If we are going to do this, we might as well do it right.  Anybody
> who isn't using this feature will not see a difference, but those who do
> manage their windows with rules can benefit significantly from just having
> sensible default values and from having the ability to explicitly set the
> value.  Therefore, this should probably be something that is added to the
> rcParams and the user can specify a value by keyword argument to the Figure
> __init__ method.
>
> Ben Root

That is far outside the scope of my capabilities.  I already have wx
and gtk backends set up (although it needs to be tested) with
hard-coded values, and qt4 half-working (and tested, this doesn't
appear to be possible in qt).  I can submit patches for that, and if
someone else wants to make it configurable they can do so.  I'll set
it up to use variables that you can then do what you want with.

-Todd

On Wed, Jan 19, 2011 at 9:16 AM, Nicholas Devenish <
mis...@gm... <misnomer%2Bm...@gm...>> wrote:

> Hi,
>
> I have a few questions on formatting axis labels, when using the
> ScalarFormatter class (which appears to be the basic standard).
>
> Firstly, is there a general method to change the font size of all
> labels on the axis, other than looping over them such as in:
>  >>> [x.set_fontsize(10) for x in plt.gca().xaxis.get_ticklabels()]
>

If you have version 1.0.0 or later, you can use tick_params():

http://matplotlib.sourceforge.net/api/pyplot_api.html#matplotlib.pyplot.tick_params
http://matplotlib.sourceforge.net/api/axes_api.html#matplotlib.axes.Axes.tick_params


>
> Secondly, when displaying the power limits with very large numbers on
> the axis, how do you change the font size of the displayed power
> limit? it does not seem to be tied into the axis ticklabels lists, and
> I do not know how to access it.
>
>
Most likely, you are talking about the "offset text", which you can get
access to its Text object using:
ax.xaxis.get_offset_text() or ax.yaxis.get_offset_text()



> Lastly, some questions on the defaults: The documentation mentions
> that the default power limits were changed from (-3, 4) to (-7, 7).
> I'm curious as to the reasoning for this, because 10^7 seems an
> excessively large number of digits and altogether a much worse
> default. In addition, the fact that using math text for the exponent
> is turned off by default (displaying 1e3 instead of the much more
> appropriate $10^3$) is also puzzling. In relation to this exponential
> display, is there a better way to turn it on, on an existing plot,
> other than the (seemingly undocumented):
>  >>> plt.gca().yaxis.get_major_formatter()._useMathText=True
>
>
I can't comment on the change of the default range of values (that happened
before I joined).  However, there are two ways to get the nice MathText
formatting of tick labels.  One, you can change the axis scale to log:

ax.set_xscale('log') or ax.set_yscale('log')

Or (if you don't want log scale), you can explicitly pass in a
ScalarFormatter object with MathText turned on:

http://matplotlib.sourceforge.net/api/ticker_api.html#matplotlib.ticker.ScalarFormatter

niceMathTextForm = ScalarFormatter(useMathText=True)
ax.xaxis.set_major_formatter( niceMathTextForm )


I hope that helps!
Ben Root

On Wed, Jan 19, 2011 at 4:48 AM, Jochen Deibele <joc...@nt...>wrote:

> Hi!
>
> todd rme wrote on 19.01.2011 04:40:
>  > The class, on the other hand, seems to be an application-specific
>  > description of what the role is within the application.  I think this
>  > should probably be "figure".
>
> Just an idea: What about exposing this (or maybe all) of the values in
> the api to be set dynamically by the user? Of course providing a useful
> default value as well.
>
> The thought behind is that it's hard to guess in what context the user
> wants to use the figure. And perhaps the user has 21 plots of some type
> and one which is different and he wants to treat differently.
> But at the same time - you also could do this if you adjust the window
> title.
>
> Jochen
>
>
I agree.  If we are going to do this, we might as well do it right.  Anybody
who isn't using this feature will not see a difference, but those who do
manage their windows with rules can benefit significantly from just having
sensible default values and from having the ability to explicitly set the
value.  Therefore, this should probably be something that is added to the
rcParams and the user can specify a value by keyword argument to the Figure
__init__ method.

Ben Root

Hi,

I have a few questions on formatting axis labels, when using the
ScalarFormatter class (which appears to be the basic standard).

Firstly, is there a general method to change the font size of all
labels on the axis, other than looping over them such as in:
 >>> [x.set_fontsize(10) for x in plt.gca().xaxis.get_ticklabels()]

Secondly, when displaying the power limits with very large numbers on
the axis, how do you change the font size of the displayed power
limit? it does not seem to be tied into the axis ticklabels lists, and
I do not know how to access it.

Lastly, some questions on the defaults: The documentation mentions
that the default power limits were changed from (-3, 4) to (-7, 7).
I'm curious as to the reasoning for this, because 10^7 seems an
excessively large number of digits and altogether a much worse
default. In addition, the fact that using math text for the exponent
is turned off by default (displaying 1e3 instead of the much more
appropriate $10^3$) is also puzzling. In relation to this exponential
display, is there a better way to turn it on, on an existing plot,
other than the (seemingly undocumented):
 >>> plt.gca().yaxis.get_major_formatter()._useMathText=True

Thanks!

Nick

On Tue, Jan 18, 2011 at 5:32 PM, Alex Liberzon <ale...@gm...>wrote:

> Hi,
>
> While moving from Matlab to Numpy/Scipy/Matplotlib I need sometimes to work
> with Matlab figures. It would be nice if we could load Matlab figures in
> Python, extract the data and some basic parameters of how it was looking in
> Matlab and create its "clone" in matplotlib. At present the Matlab figures
> are MAT files and in the near future it will be HDF5, both formats are
> loadable. However, I have not found any reference for such attempts to load
> and parse the FIG files. As a beginner I find it difficult to write a large
> piece of code. However, if there are other  interested users that can
> cooperate on such, I'd gladly contribute some hours for this task.
> Meanwhile, to show the proof-of-concept attempt is attached below. All your
> useful comments and suggestions are very welcome.
>
> Thank you,
> Alex
>
>
Alex,

That is very interesting.  I was not aware that matlab's figure files were
simply .mat files in disguise.  I would presume that it would be feasible to
produce some sort of importer in such a case (provided the documentation for
Matlab's figure format is complete enough).

I am wary of making such a function a core feature of Matplotlib, however,
because it would require creating a dependency to the scipy.io package.
However, I could see it being a toolkit package like Basemap.

Would you mind creating a feature request ticket at:

http://sourceforge.net/tracker/?group_id=80706

If you include this example code, and maybe some links to some documentation
on matlab's figure files, maybe something can grow from that.

Thanks!
Ben Root

Ryan May wrote on 19.01.2011 16:00:

> On Wed, Jan 19, 2011 at 4:48 AM, Jochen Deibele<joc...@nt...>  wrote:
>> todd rme wrote on 19.01.2011 04:40:
>>   >  The class, on the other hand, seems to be an application-specific
>>   >  description of what the role is within the application.  I think this
>>   >  should probably be "figure".
>>
>> Just an idea: What about exposing this (or maybe all) of the values in
>> the api to be set dynamically by the user? Of course providing a useful
>> default value as well.
>>
>> The thought behind is that it's hard to guess in what context the user
>> wants to use the figure. And perhaps the user has 21 plots of some type
>> and one which is different and he wants to treat differently.
>> But at the same time - you also could do this if you adjust the window
>> title.
>
> I'm not sure what the complete motivation behind all of this is, but
> if you just want to set the window title, that's already available:
>
> fig = plt.figure()
> fig.canvas.set_window_title("Look at me!")
>

Thanks, but that's not the intention.

Some windowmanager on X11 systems provide the possibility to manage 
windows due to some criteria. You can f.ex. force all windows of your 
browser to open minimized or always on top  etc. As far as I understood 
the discussion it is about what information should be provided to make 
this kind of window-handling possible.

Jochen

On Wed, Jan 19, 2011 at 4:48 AM, Jochen Deibele <joc...@nt...> wrote:
> todd rme wrote on 19.01.2011 04:40:
>  > The class, on the other hand, seems to be an application-specific
>  > description of what the role is within the application.  I think this
>  > should probably be "figure".
>
> Just an idea: What about exposing this (or maybe all) of the values in
> the api to be set dynamically by the user? Of course providing a useful
> default value as well.
>
> The thought behind is that it's hard to guess in what context the user
> wants to use the figure. And perhaps the user has 21 plots of some type
> and one which is different and he wants to treat differently.
> But at the same time - you also could do this if you adjust the window
> title.

I'm not sure what the complete motivation behind all of this is, but
if you just want to set the window title, that's already available:

fig = plt.figure()
fig.canvas.set_window_title("Look at me!")

Ryan

-- 
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

Hi!

todd rme wrote on 19.01.2011 04:40:
 > The class, on the other hand, seems to be an application-specific
 > description of what the role is within the application.  I think this
 > should probably be "figure".

Just an idea: What about exposing this (or maybe all) of the values in 
the api to be set dynamically by the user? Of course providing a useful 
default value as well.

The thought behind is that it's hard to guess in what context the user 
wants to use the figure. And perhaps the user has 21 plots of some type 
and one which is different and he wants to treat differently.
But at the same time - you also could do this if you adjust the window 
title.

Jochen

Paul,

 

Thanks for your quick
response to help me handling this problem. As proposed, I removed the
Matplotlib module (and the /build) and reinstalled it. Unfortunately I still
obtain the same import error. I use the latest Numpy version 1.5.1. 

 

To compile Matplotlib
I use gcc. For instance to compile path.cpp:

gcc
-L/home/co200009/mylib/lib -fno-strict-aliasing -DNDEBUG -g -O3 -Wall -fPIC -I/home/co200009/PYTHON-2.5/lib/python2.5/site-packages/numpy/core/include
-I/usr/include -I/home/co200009/mylib/include -I.
-I/home/co200009/PYTHON-2.5/lib/python2.5/site-packages/numpy/core/include
-Isrc -Iagg24/include -I. -I/home/co200009/PYTHON-2.5/include/python2.5 -c
src/path.cpp -o build/temp.solaris-2.8-sun4u-2.5/src/path.o

 

Thanks in advance for
your further help,

Clément

 		 	   		  

  Hello everyone,

I'm trying to build a distribution directory for a wxpython 
app that integrates matplotlib.
I want to give the user all the packages that he/she needs 
without them having to install anything.
I was able to do this successfully on windows by using py2exe.
However, when on Unix I have ran into trouble.
When I try to include the numpy package, I get the following 
ImportError:

multiarray.so: undefined symbol: _Py_ascii_whitespace

It seems that this symbol is defined in unicodeobject.h but 
I could be wrong.
Is there I library I should include or am I missing 
something else.

Thanks for you help,
Sandy

PS. I'm setting up the distribution on an Ubuntu 10.10 machine.
I have Python2.6 installed as well as the latest versions of 
all packages (matplotlib,numpy).
The distribution package I'm building should run with python2.5

On Tue, Jan 18, 2011 at 8:58 PM, Benjamin Root <ben...@ou...> wrote:
> On Tuesday, January 18, 2011, todd rme <tod...@gm...> wrote:
>> I think I've found it.  But now I need to know what the values should
>> be.  I assume the application name should be "matplotlib" (no quotes,
>> with no capitalization).  What about the window role?  I am think
>> either "figure" or "plot" would be good, probably the former (once
>> again without quotes and with no capitalization).  Is that reasonable?
>>  It is easy to change later, although the change should be consistent
>> across all backends.
>>
>> The other issue is whether I should use the same window role for all
>> backends, or separate roles for each backend (such as "qt4 figure").
>> Does anyone have an opinion on this?
>>
>> -Todd
>>
>
> My guess is that they should all have the same name.  Any given run of
> matplotlib can only use one backend at a time.  I don't see a value in
> distinguishing the source of the figures.
>
> Ben Root

That is what I thought as well.

There are three properties to deal with.  First the window class.
This contains two strings.  One is simply the application name.  The
other, however, is a class.  I can't find a good explanation of what
this is supposed to be.  For Firefox, this is "Navigator" for the
browser and "Places" for the bookmarks manager.  For libreoffice
writer, it is VCLSalFrame.DocumentWindow.

The third is the window role.  For the firefox browser, this is
"browser".  For the bookmark manager, it is "organizer".

So as best as I can tell, the role is a general description of the
sort of application it is, such as a web browser.  For matplotlib, I
think this should probably be "plot" or "graph" or something along
those lines.

The class, on the other hand, seems to be an application-specific
description of what the role is within the application.  I think this
should probably be "figure".

Does that sound reasonable?

-Todd

On Tuesday, January 18, 2011, todd rme <tod...@gm...> wrote:
> I think I've found it.  But now I need to know what the values should
> be.  I assume the application name should be "matplotlib" (no quotes,
> with no capitalization).  What about the window role?  I am think
> either "figure" or "plot" would be good, probably the former (once
> again without quotes and with no capitalization).  Is that reasonable?
>  It is easy to change later, although the change should be consistent
> across all backends.
>
> The other issue is whether I should use the same window role for all
> backends, or separate roles for each backend (such as "qt4 figure").
> Does anyone have an opinion on this?
>
> -Todd
>

My guess is that they should all have the same name.  Any given run of
matplotlib can only use one backend at a time.  I don't see a value in
distinguishing the source of the figures.

Ben Root



> On Tue, Jan 18, 2011 at 12:54 PM, todd rme <tod...@gm...> wrote:
>> Maybe, I am not that familiar with X11 programming.  Do you happen to
>> know which source files I should change?
>>
>> -Todd
>>
>> On Tue, Jan 18, 2011 at 8:39 AM, Michael Droettboom <md...@st...> wrote:
>>> You're right -- matplotlib doesn't set any of these attributes.  Fixing
>>> this problem would involve adding a few lines of code to set them in
>>> each GUI backend (since each GUI toolkit, gtk, qt, tk etc. presumably
>>> handles them differently).  Is this something you'd be interested in
>>> providing a patch for?
>>>
>>> Mike
>>>
>>> On 01/15/2011 03:06 PM, todd rme wrote:
>>>> On X11 systems, windows generally provide information about themselves
>>>> to the system.  These include things like the window title, window
>>>> class, window role, and window type.  These allow window managers to
>>>> properly handle the windows, and in some window managers (like KDE's
>>>> kwin), lets you set rules for windows of certain types.
>>>>
>>>> As best as I can tell, by default matplotlib does not provide most of
>>>> this information.  It provides the window title and the window type,
>>>> but it doesn't seem to provide the window class or the window role.
>>>> These two properties, however, are exactly the properties someone
>>>> would want to use when trying to set up rules for particular windows.
>>>> This makes it impossible to set window-specific rules for the plots.
>>>>
>>>> Is there something in the matplotlib configuration that would let
>>>> someone set these properties, and if not is there any way matplotlib
>>>> could set them itself?
>>>>
>>>> I'm sorry if this is a common question.
>>>>
>>>> -Todd
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Protect Your Site and Customers from Malware Attacks
>>>> Learn about various malware tactics and how to avoid them. Understand
>>>> malware threats, the impact they can have on your business, and how you
>>>> can protect your company and customers by using code signing.
>>>> http://p.sf.net/sfu/oracle-sfdevnl
>>>> _______________________________________________
>>>> Matplotlib-users mailing list
>>>> Mat...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Protect Your Site and Customers from Malware Attacks
>>> Learn about various malware tactics and how to avoid them. Understand
>>> malware threats, the impact they can have on your business, and how you
>>> can protect your company and customers by using code signing.
>>> http://p.sf.net/sfu/oracle-sfdevnl
>>> _______________________________________________
>>> Matplotlib-users mailing list
>>> Mat...@li...
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>>
>>
>
> ------------------------------------------------------------------------------
> Protect Your Site and Customers from Malware Attacks
> Learn about various malware tactics and how to avoid them. Understand
> malware threats, the impact they can have on your business, and how you
> can protect your company and customers by using code signing.
> http://p.sf.net/sfu/oracle-sfdevnl
> _______________________________________________
> Matplotlib-users mailing list
>

I think I've found it.  But now I need to know what the values should
be.  I assume the application name should be "matplotlib" (no quotes,
with no capitalization).  What about the window role?  I am think
either "figure" or "plot" would be good, probably the former (once
again without quotes and with no capitalization).  Is that reasonable?
 It is easy to change later, although the change should be consistent
across all backends.

The other issue is whether I should use the same window role for all
backends, or separate roles for each backend (such as "qt4 figure").
Does anyone have an opinion on this?

-Todd

On Tue, Jan 18, 2011 at 12:54 PM, todd rme <tod...@gm...> wrote:
> Maybe, I am not that familiar with X11 programming.  Do you happen to
> know which source files I should change?
>
> -Todd
>
> On Tue, Jan 18, 2011 at 8:39 AM, Michael Droettboom <md...@st...> wrote:
>> You're right -- matplotlib doesn't set any of these attributes.  Fixing
>> this problem would involve adding a few lines of code to set them in
>> each GUI backend (since each GUI toolkit, gtk, qt, tk etc. presumably
>> handles them differently).  Is this something you'd be interested in
>> providing a patch for?
>>
>> Mike
>>
>> On 01/15/2011 03:06 PM, todd rme wrote:
>>> On X11 systems, windows generally provide information about themselves
>>> to the system.  These include things like the window title, window
>>> class, window role, and window type.  These allow window managers to
>>> properly handle the windows, and in some window managers (like KDE's
>>> kwin), lets you set rules for windows of certain types.
>>>
>>> As best as I can tell, by default matplotlib does not provide most of
>>> this information.  It provides the window title and the window type,
>>> but it doesn't seem to provide the window class or the window role.
>>> These two properties, however, are exactly the properties someone
>>> would want to use when trying to set up rules for particular windows.
>>> This makes it impossible to set window-specific rules for the plots.
>>>
>>> Is there something in the matplotlib configuration that would let
>>> someone set these properties, and if not is there any way matplotlib
>>> could set them itself?
>>>
>>> I'm sorry if this is a common question.
>>>
>>> -Todd
>>>
>>> ------------------------------------------------------------------------------
>>> Protect Your Site and Customers from Malware Attacks
>>> Learn about various malware tactics and how to avoid them. Understand
>>> malware threats, the impact they can have on your business, and how you
>>> can protect your company and customers by using code signing.
>>> http://p.sf.net/sfu/oracle-sfdevnl
>>> _______________________________________________
>>> Matplotlib-users mailing list
>>> Mat...@li...
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>
>>
>> ------------------------------------------------------------------------------
>> Protect Your Site and Customers from Malware Attacks
>> Learn about various malware tactics and how to avoid them. Understand
>> malware threats, the impact they can have on your business, and how you
>> can protect your company and customers by using code signing.
>> http://p.sf.net/sfu/oracle-sfdevnl
>> _______________________________________________
>> Matplotlib-users mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>
>