matplotlib-users Mailing List for matplotlib

Eli Brosh wrote:
> Hello,
> I am trying to extract the coordinates of contour lines.
> I tried the following:
>
> cs = *contour*(Z)
> for lev, col in zip(cs.levels, cs.collections):
>      s = col._segments
>
> that I found in a previous post (title "contouring", by Jose 
> Gómez-Dans-2 <http://www.nabble.com/user/UserProfile.jtp?user=30071> 
> Nov 30, 2007; 07:47am ) .
>
> I hoped that s will be a list of numpy arrays, each containing the 
> (x,y) vertices
> defining a contour line at level lev.
> However, I got an error message:
> AttributeError: 'LineCollection' object has no attribute '_segments'
>
>
> How is it possible to get coordinates of the contours, similar to the 
> MATLAB command
>  [C,H] = *CONTOUR*(...)
> where the result in C is the coordinates of the contours.
>
> A similar question appeared in a post "contour data" (by Albert Swart 
> <http://www.nabble.com/user/UserProfile.jtp?user=382945> May 17, 2006; 
> 09:42am) but I could not understand the answer.
> Is it possible to get more specific directions with a simple example ?
>
>
> Thanks
> Eli
Eli:  Calling get_paths() on each line collection in CS.collections will 
return a list of Path objects.  From the Path objects, you can get a Nx2 
array of vertices from the "vertices" attribute.  There are no examples 
that I know of, but if you get it to do what you want to do, it would be 
great if you could contribute an example.  As you noted, this question 
has come up several times before.

-Jeff

-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/PSD  R/PSD1        Email  : Jef...@no...
325 Broadway                Office : Skaggs Research Cntr 1D-113
Boulder, CO, USA 80303-3328 Web    : http://tinyurl.com/5telg

Hello,
I am trying to extract the coordinates of contour lines.
I tried the following:

cs = *contour*(Z)
for lev, col in zip(cs.levels, cs.collections):
     s = col._segments

that I found in a previous post (title "contouring", by Jose
Gómez-Dans-2<http://www.nabble.com/user/UserProfile.jtp?user=30071>
Nov
30, 2007; 07:47am ) .

I hoped that s will be a list of numpy arrays, each containing the (x,y)
vertices
defining a contour line at level lev.
However, I got an error message:
AttributeError: 'LineCollection' object has no attribute '_segments'


How is it possible to get coordinates of the contours, similar to the MATLAB
command
 [C,H] = *CONTOUR*(...)
where the result in C is the coordinates of the contours.

A similar question appeared in a post "contour data" (by Albert
Swart<http://www.nabble.com/user/UserProfile.jtp?user=382945> May
17, 2006; 09:42am) but I could not understand the answer.
Is it possible to get more specific directions with a simple example ?


Thanks
Eli

> The problem I have is then when I issue the following
>
> >>> from pylab import *
>
> Traceback (most recent call last):
>   File "<stdin>", line 1, in <module>
>   File "/usr/local/lib/python2.6/site-packages/pylab.py", line 1, in
> <module> File "/usr/local/lib/python2.6/site-packages/matplotlib/pylab.py",
> line 206, in <module> File
> "/usr/local/lib/python2.6/site-packages/matplotlib/mpl.py", line 1, in
> <module> File
> "/usr/local/lib/python2.6/site-packages/matplotlib/artist.py", line 5, in
> <module> File
> "/usr/local/lib/python2.6/site-packages/matplotlib/transforms.py", line 34,
> in <module> ImportError: No module named _path


I have resolved it. The problem was the protection on the .so files. They were 
installed with -rwxrwx--- which I changed to -rwxrwxrwx.

I can now plot.

Peter

It just occurred to me that another option might be to simply add a new 
colour option "line" for mec and mfc which would instruct them to pick 
up the current line colour.

Gary

At http://www.scipy.org/Cookbook/Matplotlib/SigmoidalFunctions I read 
that matplotlib supports drawing arbitrary paths with splines, but the 
functionality hasn't been exposed to the user. Can anyone tell me how I 
can use this functionality? I tried searching the source code for 
"spline," but I couldn't find anything useful.

(Sorry if this has been already brought up. I tried to search the 
archives but, weirdly enough, I couldn't find any search functionality. 
Perhaps it hasn't been exposed to the user, yet!)

Norbert Nemec wrote:
> Before my work in 2004, the colors were not following the line color at 
> all, which was clearly bad behavior.
> 
> Now, there are two categories: filled markers (with edge color black and 
> filling following the line color) and non-filled markers (with edge 
> color following line color).
> 
> The black edge of filled markers is a matter of style which I personally 
> like and would not want to change.
> 
> The thing that was up for dispute was only about what the edge color of 
> filled markers should do when the filling is switched off. I see three 
> ways to solve this:
> 
> a) Leave it black. (current behavior)
> b) Switch mec to line color if mfc is either "none" or "white".
> c) Switch mec to line color if mfc is not "auto"

For (b), please don't consider "white" as equivalent to "none"--they are 
completely different.  The most logical thing, with minimum surprise and 
maximum ease of use, *might* be to consider filled markers with the 
filling turned off as exactly equivalent to unfilled markers.  Setting 
mfc to "none" is what turns off the filling.

I don't see any docstring explanation of the "auto" settings.

Eric


> 
> b) or c) might be what people would expect and prefer, but I feared that 
> it would be one step too many in built-in intelligence. But then - maybe 
> c) would be ok? After all, switching from c) to a) by an explicit 
> mec="k" is simple and obvious, the other way around takes a bit more.
> 
> Greetings,
> Norbert
> 
> 
> 
> Gary Ruben wrote:
>> Thanks John,
>>
>> That shows how long it is since I used line markers in my plots. Because 
>> I use them so infrequently, I'm probably not the best one to suggest it, 
>> but I think it would be nicer for the default colour to match the line 
>> colour by default, or for an option to be added to allow its simple 
>> selection without users having to search through the mailing list to 
>> find Norbert's solution. If I was publishing a colour plot with line 
>> markers I would definitely want to do this.
>>
>> Gary
>>
>> John Hunter wrote:
>>   
>>> On Mon, Jan 26, 2009 at 6:17 AM, Gary Ruben wrote:
>>>     
>>>> Has the mec always been black? I thought it used to be the same as the
>>>> line colour. I expected it to default to the line colour, as Che expected.
>>>>       
>>> It's been this way since at least 2004:
>>>
>>>   http://matplotlib.svn.sourceforge.net/viewvc/matplotlib/trunk/matplotlib/lib/matplotlib/__init__.py?revision=540&view=markup
>>>
>>> JDH
>>>     
>>
>> ------------------------------------------------------------------------------
>> This SF.net email is sponsored by:
>> SourcForge Community
>> SourceForge wants to tell your story.
>> http://p.sf.net/sfu/sf-spreadtheword
>> _______________________________________________
>> Matplotlib-users mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>
>>   
> 
> 
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Dear All,

I apologise if this is naive, but I am having problems with matplotlib. I downloaded it as a tar ball from SourceForge
and it all compiled and installed without any error. I have the following
setup.

Download source
	SourceForge - matplotlib-0.98.5.2.tar.gz

OS
	OpenSUSE 11.1

gcc
	gcc version 4.3.2 [gcc-4_3-branch revision 141291] (SUSE Linux)

uname -a
	Linux trevor 2.6.27.7-9-pae #1 SMP 2008-12-04 18:10:04 +0100 i686 i686 i386 GNU/Linux

python
	Python 2.6

matplotlib
	>>> import matplotlib; print matplotlib.__version__
	0.98.5.2

I have taken the default 'matplotlibrc' from the web site and it is stored in ~/.matplotlib

The problem I have is then when I issue the following

>>> from pylab import *
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python2.6/site-packages/pylab.py", line 1, in <module>
  File "/usr/local/lib/python2.6/site-packages/matplotlib/pylab.py", line 206, in <module>
  File "/usr/local/lib/python2.6/site-packages/matplotlib/mpl.py", line 1, in <module>
  File "/usr/local/lib/python2.6/site-packages/matplotlib/artist.py", line 5, in <module>
  File "/usr/local/lib/python2.6/site-packages/matplotlib/transforms.py", line 34, in <module>
ImportError: No module named _path

Any suggestions will be greatly appreciated, thanks

Peter

Hi,

I am creating a scatter plot of some data that spans many orders of
magnitude, and so I am changing the axes using
gca().set_xscale('log')
gca().set_yscale('log')

Clearly, this only shows non-zero data.  Is there an easy way to also
include data has zero values for coordinates (say one tick mark to the
left/below)?

Thanks in advance!

Uri


-- 
Uri Laserson
PhD Candidate, Biomedical Engineering
Harvard Medical School (Genetics)
Massachusetts Institute of Technology (Mathematics)
phone +1 917 742 8019
las...@mi...

massimo sandal wrote:
> Dmitrey ha scritto:
>> hi matplotlib developers and users,
>> after installation, it yields
>>
>> ImportError: No module named _tkinter, please install the python-tk package

> all complex packages have dependencies. 
> It just seems Kubuntu missed one.

Maybe, Maybe not -- MPL is perfectly usable without tk -- you just can't 
use the tk back-end. If someone is going to use it with GTK, or wx, or 
to generate images for a web app, they don't need TK, for instance, and 
may not want it installed.

However, it looks like MPL is setting TK as the default back-end, so 
maybe TK should be a dependency if they're doing that.

-Chris




-- 
Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Derek Hohls wrote:
> I hope I speak for others when I say: for those of us who cannot
> be at PyCon, please consider making tutorials etc available for 
> download afterwards.  We'd love to be able to spread the good
> word about these excellent libraries.

Keep an eye out on the PyConUS website after the conference, I believe 
they'll make the tutorials available there...

cheers,

Chris

-- 
Simplistix - Content Management, Zope & Python Consulting
            - http://www.simplistix.co.uk

Le mardi 27 janvier 2009 à 13:17 +0100, Marcin Krol a écrit :
> How do I set the attributes of Line2D instance? I have:
> 
> 	origline = plt.plot(origlist)
> 	valine = plt.plot(valist)
> 	plt.xlim(0,256)
> 	valine.set_color('r')
> 	valine.set_linewidth(2)
> 
> Traceback (most recent call last):
> AttributeError: 'list' object has no attribute 'set_color'

As mentioned in the traceback, valine is a list and not a Line2D
instance. In fact, the plt.plot function returns the list of the Line2D
instances added to the axes by the instruction. Even if you only plot
one curve.
As a solution, you may correct to 
        valine, = plt.plot(valist)
which unpack the list and extract the first item (the Line2D instance
you want) to valine.

-- 
Fabrice Silva <si...@lm...>
LMA UPR CNRS 7051 - équipe S2M

mfabulous wrote:
> Hi all,
> 
> I apologize if this was asked before. I just started adding legends to my
> plots 
> and found the legend stile for a point-plot, i.e. a plot using points
> rather
> than 
> lines to be a bit confusing. 
> 
> Example:
> pylab.plot(arange(5), arange(5), 'o', label="foo")
> pylab.legend()
> 
> Will produce a legend which looks like "o  o foo".
> Why are there two points? Can I force it to give me only one? 

That may depend on how old your version of matplotlib is; for quite a 
while there has been a "numpoints" kwarg to the legend command which 
lets you set that number, e.g., "pylab.legend(numpoints=1)".

Eric

Thats it...

thank you!

Cheers,

Maximilian
-- 
View this message in context: http://www.nabble.com/legend-for-plots-with-points-tp21680968p21689742.html
Sent from the matplotlib - users mailing list archive at Nabble.com.

Hi,

I would like to change the hue range of the hsv color map to use only hue
values between 0 and 0.6667. I checked the documentation of matplotlib and
searched in the internet but couldn't find the answer.

Best regards,
KB
-- 
View this message in context: http://www.nabble.com/Change-Hue-Range-in-HSV-color-map-tp21689235p21689235.html
Sent from the matplotlib - users mailing list archive at Nabble.com.

mfabulous wrote:
> Hi all,
> 
> I apologize if this was asked before. I just started adding legends to my
> plots 
> and found the legend stile for a point-plot, i.e. a plot using points rather
> than 
> lines to be a bit confusing. 
> 
> Example:
> pylab.plot(arange(5), arange(5), 'o', label="foo")
> pylab.legend()
> 
> Will produce a legend which looks like "o  o foo".
> Why are there two points? Can I force it to give me only one? 

That may depend on how old your version of matplotlib is; for quite a 
while there has been a "numpoints" kwarg to the legend command which 
lets you set that number, e.g., "pylab.legend(numpoints=1)".

Eric

Hi Norbert,

Both of your proposals (b) and (c) sound better to me than the current 
behaviour, although they don't sound as obvious to me as simply 
defaulting to always setting the mec to the line colour unless 
overridden using mec="k" - you could label this proposal (d).

Since others seem to have been happy with the black edges and I don't 
know how much extra logic (and consequently extra overhead) is required 
to change to (d), I'd be easily persuaded that (b) or (c) would be OK 
since, as you say, the black edge of filled markers is a matter of style 
preference and if black edges are not wanted on a particular plot that 
uses filled markers, the edge width can simply be set to zero. The 
decision might be guided by whichever results in the simplest logic or 
least overhead.

regards,
Gary

Norbert Nemec wrote:
> Before my work in 2004, the colors were not following the line color at 
> all, which was clearly bad behavior.
> 
> Now, there are two categories: filled markers (with edge color black and 
> filling following the line color) and non-filled markers (with edge 
> color following line color).
> 
> The black edge of filled markers is a matter of style which I personally 
> like and would not want to change.
> 
> The thing that was up for dispute was only about what the edge color of 
> filled markers should do when the filling is switched off. I see three 
> ways to solve this:
> 
> a) Leave it black. (current behavior)
> b) Switch mec to line color if mfc is either "none" or "white".
> c) Switch mec to line color if mfc is not "auto"
> 
> b) or c) might be what people would expect and prefer, but I feared that 
> it would be one step too many in built-in intelligence. But then - maybe 
> c) would be ok? After all, switching from c) to a) by an explicit 
> mec="k" is simple and obvious, the other way around takes a bit more.
> 
> Greetings,
> Norbert

I figured out why your show() works...you had WxPython installed and
I'm assuming your backend was defaulting to this since Tkinter wasn't
found.  Once I noticed that I installed WxPython and then re-buily
matplotlib with WxPython.  If I used wxpython as my backend, then my
GUIs worked.  I can import pylab just fine and I don't appear to have
any errors.  While I can't figure out what your problem is, you are
free to try the binary I built and see if it works:
http://code.patricktmarsh.com/matplotlib-0.98.6svn.win32-py2.6.exe

If you do decide to try this, let me know if it works.

-Patrick





On Tue, Jan 27, 2009 at 9:22 AM, Marcin Krol <mr...@gm...> wrote:
> John Hunter wrote:
>
>> First of all, on the clean build, when you do
>>
>>   import matplotlib._path
>>
>> do you still get the same error.
>
> No, it's working fine on the new build (the import above and the whole
> installation0 -- once I switched to VC++ from MinGW.
>
> <snip info, ack>
>
> Regards,
> mk
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>



-- 
Patrick Marsh
Graduate Research Assistant
School of Meteorology
University of Oklahoma
http://www.patricktmarsh.com

On Tue, Jan 27, 2009 at 9:27 AM, Nils Wagner
<nw...@ia...> wrote:

>> If you want to only return a single point, you can write
>>a custom
>> picker function to return only the point closest to the
>>click point.
>>
>
> Unfortunately, I am not sure how to implement a custom
> picker function.
>
> Any pointer would be appreciated.

The example I pointed you to in my first post has an example of a
custom picker function.  It is in the section labelled "picking with a
custom hit test function"

  http://matplotlib.sourceforge.net/examples/event_handling/pick_event_demo.html

We've taken a lot of time to write the documentation and examples.
When you ask us questions on the mailing list and we point you to
documentation and examples, please take the time to read them.  One
reason we wrote them is so that we would not have to continuously
solve the same problems over and over again on the mailing list.  I'm
happy to help if you find the documentation or examples confusing, but
it will serve us both if you spend some time trying to digest them.

JDH

On Tue, 27 Jan 2009 08:57:17 -0600
  John Hunter <jd...@gm...> wrote:
> On Tue, Jan 27, 2009 at 8:18 AM, Nils Wagner
> <nw...@ia...> wrote:
>>> Thank you very much for your response.
>>> The horizontal distance between consecutive points is
>>> 5. in my application.
>>>
>>> Is the value of picker linked to the distance ?
>>> I mean
>>> How should I use picker to circumvent multiple results ?
> 
> The scalar value of the picker is a epsilon tolerance in 
>points (1
> point = 1/72 inches) so you will need to set this 
>distance to be small
> enough to only overlap one point when you click.
> 
> If you want to only return a single point, you can write 
>a custom
> picker function to return only the point closest to the 
>click point.
> 

Unfortunately, I am not sure how to implement a custom 
picker function.

Any pointer would be appreciated.

Thanks in advance.

                        Nils
  

John Hunter wrote:

> First of all, on the clean build, when you do
> 
>   import matplotlib._path
> 
> do you still get the same error.

No, it's working fine on the new build (the import above and the whole 
installation0 -- once I switched to VC++ from MinGW.

<snip info, ack>

Regards,
mk

On Tue, Jan 27, 2009 at 7:09 AM, Marcin Krol <mr...@gm...> wrote:
> John Hunter wrote:
>> It might help if you recursively remove the buld directory, do a clean
>> rebuld, capture the build output in a file, and post that here so we
>> can see if we can get an idea what is going on
>
> Since it's over 250kB, I pasted it here:
>
> http://pastebin.com/f466bd0db

First of all, on the clean build, when you do

  import matplotlib._path

do you still get the same error.

I don't know a ton about windows and even less about visual studio,
but according to the build output you posted a _path.pyd file is built
and apparently installed.   It may be that it is trying to linking
against some dll that is not on your system.  There is a windows
freeware program I used to use to debug this stuff, in which you could
load the pyd file and see what it links against, which reported
failures.  Perhaps a windows guru can advise here.

In the past, there was a bug in distutils that was forcing the
compiler to link against msvcp71.dll which could be fixed by patching
distutils.  Don't know if this is related

On Tue, Jan 27, 2009 at 8:18 AM, Nils Wagner
<nw...@ia...> wrote:
>> Thank you very much for your response.
>> The horizontal distance between consecutive points is
>> 5. in my application.
>>
>> Is the value of picker linked to the distance ?
>> I mean
>> How should I use picker to circumvent multiple results ?

The scalar value of the picker is a epsilon tolerance in points (1
point = 1/72 inches) so you will need to set this distance to be small
enough to only overlap one point when you click.

If you want to only return a single point, you can write a custom
picker function to return only the point closest to the click point.

JDH

On Tue, Jan 27, 2009 at 2:15 AM, Nils Wagner
<nw...@ia...> wrote:
> Hi all,
>
> Is it possible to restrict the mouse movements to the points of a graph
> (curve tracking) ?


No, but you can obtain the vertices of the line that are within a
tolerance of the clicked point using the picking functions.  See the
tutorial at

  http://matplotlib.sourceforge.net/users/event_handling.html

and the "pick*" demos at

  http://matplotlib.sourceforge.net/examples/event_handling/index.html

JDH

On Tue, Jan 27, 2009 at 4:31 AM, Marcin Krol <mr...@gm...> wrote:
> Hello Patrick,
>
> Thanks for reply!
>
>> I didn't catch on the first email you were trying to build for
>> python2.6.  I haven't been able to get matplotlib to build correctly
>> with python2.6.  I'm actually curious as to how you were able to build
>> matplotlib using the win32_static folder since it doesn't contain the
>> libs and header files for tcl/tk 8.5 which python2.6 is build against.
>
> I didn't do anything really, except I have had installed Python 2.5 on
> the same system before. Maybe this left DLLs lying around and the
> compiled matplotlib uses those old DLLs?
>
>>  This was the original error for me and I had to modify the
>> setupext.py to look for the tcl/tk 8.5 includes (which I had to
>> manually add to win32_static).  What happens if you try:
>>
>> import matplotlib.pyplot as plt
>> x = range(0,100)
>> y = range(0,100)
>> plt.plot(x,y)
>> plt.show()
>
> I have used win32_static_vs (I have VC++ 9.0 installed) and it just
> worked! It displayed the graph and everything.
>
>> As for the error you specifically mentioned, have you checked to make
>> sure that the correct dll exists?  This error typically happens to me
>> when the dll didn't build properly (if at all).
>
> How do I do that? After this successful build, I can't find any DLL in
> the 'build' folder. I have unpacked matplotlib-0.98.5.2.win32-py2.6.exe
> installer and it doesn't contain any dll either.

It might help if you recursively remove the buld directory, do a clean
rebuld, capture the build output in a file, and post that here so we
can see if we can get an idea what is going on

Hello everyone,

Newbie alert.

How do I set the attributes of Line2D instance? I have:

	origline = plt.plot(origlist)
	valine = plt.plot(valist)
	plt.xlim(0,256)
	valine.set_color('r')
	valine.set_linewidth(2)


I get:

Traceback (most recent call last):
   File "C:/Python26/loadavg2.py", line 121, in <module>
     plotit(valist, smalist)
   File "C:/Python26/loadavg2.py", line 106, in plotit
     valine.set_color('r')
AttributeError: 'list' object has no attribute 'set_color'

WTF?? After all, Line2D api docs, e.g.

http://matplotlib.sourceforge.net/api/artist_api.html#matplotlib.lines.Line2D.set_color

clearly state there is set_color method to the Line2D instance!

I get the same error with get_* methods.

Regards,
mk

Hello everyone,

Got it now, plot() apparently returns a list and I have to unpack it, I 
feel silly now...

Regards,
mk