matplotlib-users Mailing List for matplotlib

On Fri, Aug 2, 2013 at 3:32 PM, Eric Firing <ef...@ha...> wrote:

> On 2013/08/02 8:55 AM, Drain, Theodore R (392P) wrote:
> > Thanks - we'll look into that.  We might also see how hard it would
> > be to implement an update or refresh method on the legend that could
> > be called when the lines change to keep the legend in sync.  Seems
> > like the legend should own that functionality since it set up the
> > mapping between the lines and what it's displaying...
>
> I agree entirely.  It would be logical for the legend to either have a
> manual refresh method, or perhaps to be coupled to its targets the way a
> colorbar is coupled to its mappable, tracking it automatically.  The
> Legend is a very complex beast, however, so I suspect this is a real
> project.
>
>
Mike and I discussed this at the recent conference. What makes
ScalarMappable different from other artists is that it has attribute
caching and it has callback mechanisms for changes to certain attributes.
This is why colorbar can change with its image. These features needs to be
better generalized and cleaned up, and then applied to *all* attributes.

Another issue is that the artist objects contained in the legend are
created from the artist objects that it represents -- at the time of legend
creation. All color, linestyle marker, etc. attributes are copied rather
than referenced in the artists in the legend. Therefore, any changes to
either doesn't impact the other, unfortunately.

So, there are two approaches to solve this. 1) implement a generic,
efficient callback mechanism for attributes (and proper caching to reduce
unneeded dispatches) and have the legend object register callbacks for any
changes in the legend entries. 2) have a mechanism for some sort of shared
attributes.

Just thinking out loud... wishing I had the time to actually implement my
ideas...

Cheers!
Ben Root

On 2013/08/02 8:55 AM, Drain, Theodore R (392P) wrote:
> Thanks - we'll look into that.  We might also see how hard it would
> be to implement an update or refresh method on the legend that could
> be called when the lines change to keep the legend in sync.  Seems
> like the legend should own that functionality since it set up the
> mapping between the lines and what it's displaying...

I agree entirely.  It would be logical for the legend to either have a 
manual refresh method, or perhaps to be coupled to its targets the way a 
colorbar is coupled to its mappable, tracking it automatically.  The 
Legend is a very complex beast, however, so I suspect this is a real 
project.

On Sun, Jul 28, 2013 at 7:14 AM, Jeffrey Spencer <jef...@gm...>wrote:

> Ben,
>
> Thanks that works great and also one more question. If you look at the
> previous example. I have noticed that at the angle the figure is at the
> ticklabels look like they are at the center of the grid boxes. This is not
> the case because if you turn it to the side as the example below image.pdf.
> The 0.00, 0.05 on the I_lw axis are supposed to be at the actual tick marks.
>
>
>
Thus, can the axis be manually moved too? This isn't a huge deal but I was
> just wondering if it was possible to offset the ticklabels. I looked in the
> _axinfo but looks like for ticklabel their is only a space_factor parameter.
>
> Cheers,
> Jeff
>
>
The effect you are seeing is that the tick labels are set to be below and
in front of the tick marks. More specifically, what happens is that the
ticklabels (and the axis label) are offsetted away from the center of the
domain. The space_factor value you found is just some empirical value that
I have found to work fairly well.  Effectively, there is only a
"radial"-like control over the spacing, not a finer-grained control.

That being said, you can modify the "va" value of the _axinfo to control
the vertical allignment of the labels. I would wonder if messing around
with that might have some desired impact.  Another possibility is to use
'\n' characters before or after the main text for the label to make an
apparent shift.  Just some ideas to play around with.

Cheers!
Ben Root

Thanks - we'll look into that.  We might also see how hard it would be to implement an update or refresh method on the legend that could be called when the lines change to keep the legend in sync.  Seems like the legend should own that functionality since it set up the mapping between the lines and what it's displaying...

________________________________________
From: Eric Firing [ef...@ha...]
Sent: Thursday, August 01, 2013 8:37 PM
To: mat...@li...
Subject: Re: [Matplotlib-users] Update legend when lines change color?

On 2013/08/01 4:23 PM, Drain, Theodore R (392P) wrote:
> I have an application where the user can edit line colors and other attributes after the plot is drawn.  The artists update just fine but the legend doesn't change.
>
>>From what I can see in the legend code, it doesn't seem like there is any mechanism in place for doing this.  Does anyone have any ideas on the best way to implement something like this?
>
> Here is a simplified script to show the issue:
>
> import pylab as p
> p.ion()
> l = p.plot( [1,2,3], 'b', label="foo" )
> p.legend()
>
> raw_input( "press return..." )
> l[0].set_color( "green" )
> p.draw()
>
> Thanks,
> Ted


If you keep a reference to the Legend object, then you can call its
get_lines() method to get a list of Line2D objects corresponding to the
objects returned by plot().  You can use the set_color() method on them.
Maybe this is enough if your application is simple enough.

Eric

------------------------------------------------------------------------------
Get your SQL database under version control now!
Version control is standard for application code, but databases havent
caught up. So what steps can you take to put your SQL databases under
version control? Why should you start doing it? Read more to find out.
http://pubads.g.doubleclick.net/gampad/clk?id=49501711&iu=/4140/ostg.clktrk
_______________________________________________
Matplotlib-users mailing list
Mat...@li...
https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Thanks!  Either of those looks like it will work.  I'll play w/ both of them to see which fits better w/ my existing code.



Ted



________________________________
From: Goodman, Alexander (398J-Affiliate) [go...@jp...]
Sent: Friday, August 02, 2013 11:16 AM
To: Benjamin Root
Cc: Drain, Theodore R (392P); mat...@li...
Subject: Re: [Matplotlib-users] Splitting arrays into chunks that satisfy a condition?

Hi Ted,

As far as actually splitting up a numpy array into contiguous chunks fulfilling a condition, there is a very good solution posted on stackoverflow:
http://stackoverflow.com/questions/4494404/find-large-number-of-consecutive-values-fulfilling-condition-in-a-numpy-array

If you use the contiguous_regions function from the first answer, this code should give you what you want:
xneg = [x[slice(*reg)] for reg in contiguous_regions(z < 0)]
xpos = [x[slice(*reg)] for reg in contiguous_regions(z >= 0)]

Thanks,
Alex


On Fri, Aug 2, 2013 at 10:45 AM, Benjamin Root <ben...@ou...<mailto:ben...@ou...>> wrote:

On Fri, Aug 2, 2013 at 1:36 PM, Drain, Theodore R (392P) <the...@jp...<mailto:the...@jp...>> wrote:
I have three arrays (x,y,z).  I want plot x vs y and draw the line segments differently depending on whether or not z is positive or negative.  So I'm trying to split the x,y arrays into chunks depending on the value of z.  Using numpy.where, I can find the indeces in z that satisfy a condition but I can't figure out an efficient way (other than brute force) to split the array up into continuous chunks.  Does anyone know of a numpy trick that would help with this?

Here's a simple example:

# index:    0  1  2  3   4  5  6  7 8 9
z=numpy.array([-1,-1,-1, 1, -1,-1,-1, 1,1,1] )
x=numpy.array([-2,-3,-4, 2, -5,-6,-7, 3,4,5] )

# Want: xneg = [ x[0:3], x[4:7] ], xpos = [ x[3:4], x[7:10] ]
xneg = [ [-2,-3,-4], [-5,-6,-7] ]
xpos = [ [ 2 ], [ 3, 4, 5 ] ]

idxneg = numpy.where( z < 0 )[0]
# == [ 0,1,2, 4,5,6 ]
idxpos = numpy.where( z >= 0 )[0]
# == [ 3, 7,8,9 ]

Thanks,
Ted

One way I would go about it is to do this:

z1 = numpy.where(z < 0, z, numpy.nan)
z2 = numpy.where(z >= 0, z, numpy.nan)

And then plot those against x.  matplotlib ignores nans and would break up the line where-ever a nan shows up (assuming that is the effect you want).

Cheers!
Ben Root



--
Alex Goodman

Yes, I see that now. I hadn't noticed that the face color for the bottom of
the 3d plot is off as well. Thanks for the update and keep me posted.

Cheers,
Jeff


On Sat, Aug 3, 2013 at 3:41 AM, Benjamin Root <ben...@ou...> wrote:

>
>
>
> On Fri, Jul 26, 2013 at 4:41 PM, Jeffrey Spencer <jef...@gm...>wrote:
>
>> I have version 1.2.x of matplotlib. The minimal example shows the case
>> below. The back wall will lose its lines. Is there a reason for this? Is
>> there a transparent layer there so eps has to put it as a solid wall? If
>> so, is there a way to remove that transparent layer?
>>
>> Pdf and the other backends have no problem exporting correctly.
>>
>> import matplotlib.pyplot as plt
>> from mpl_toolkits.mplot3d import Axes3D
>> fig = plt.figure()
>> ax = fig.add_subplot(111, projection='3d')
>> plt.savefig('testing.eps')
>>
>>
> Finally had some time to investigate this a bit...
>
> Diving into the eps output, I can tell that the grid lines are being
> output to the file, but it seems like the color of the gridline matches
> that particular wall's color, which is why we can't see it.  The odd thing
> is that the line colors are correct, but for some reason, the face colors
> are wrong for eps.  Looking at the output for pdf and png, the walls are
> not shaded nearly as much as it is in eps.  There is probably something
> messed up in our PS backend that is misinterpreting the grayscale color
> information it is getting.
>
> Will have to dive in some more...
>
> Cheers!
> Ben Root
>

On Fri, Aug 2, 2013 at 1:36 PM, Drain, Theodore R (392P) <
the...@jp...> wrote:

> I have three arrays (x,y,z).  I want plot x vs y and draw the line
> segments differently depending on whether or not z is positive or negative.
>  So I'm trying to split the x,y arrays into chunks depending on the value
> of z.  Using numpy.where, I can find the indeces in z that satisfy a
> condition but I can't figure out an efficient way (other than brute force)
> to split the array up into continuous chunks.  Does anyone know of a numpy
> trick that would help with this?
>
> Here's a simple example:
>
> # index:    0  1  2  3   4  5  6  7 8 9
> z=numpy.array([-1,-1,-1, 1, -1,-1,-1, 1,1,1] )
> x=numpy.array([-2,-3,-4, 2, -5,-6,-7, 3,4,5] )
>
> # Want: xneg = [ x[0:3], x[4:7] ], xpos = [ x[3:4], x[7:10] ]
> xneg = [ [-2,-3,-4], [-5,-6,-7] ]
> xpos = [ [ 2 ], [ 3, 4, 5 ] ]
>
> idxneg = numpy.where( z < 0 )[0]
> # == [ 0,1,2, 4,5,6 ]
> idxpos = numpy.where( z >= 0 )[0]
> # == [ 3, 7,8,9 ]
>
> Thanks,
> Ted
>

One way I would go about it is to do this:

z1 = numpy.where(z < 0, z, numpy.nan)
z2 = numpy.where(z >= 0, z, numpy.nan)

And then plot those against x.  matplotlib ignores nans and would break up
the line where-ever a nan shows up (assuming that is the effect you want).

Cheers!
Ben Root

On Fri, Jul 26, 2013 at 4:41 PM, Jeffrey Spencer <jef...@gm...>wrote:

> I have version 1.2.x of matplotlib. The minimal example shows the case
> below. The back wall will lose its lines. Is there a reason for this? Is
> there a transparent layer there so eps has to put it as a solid wall? If
> so, is there a way to remove that transparent layer?
>
> Pdf and the other backends have no problem exporting correctly.
>
> import matplotlib.pyplot as plt
> from mpl_toolkits.mplot3d import Axes3D
> fig = plt.figure()
> ax = fig.add_subplot(111, projection='3d')
> plt.savefig('testing.eps')
>
>
Finally had some time to investigate this a bit...

Diving into the eps output, I can tell that the grid lines are being output
to the file, but it seems like the color of the gridline matches that
particular wall's color, which is why we can't see it.  The odd thing is
that the line colors are correct, but for some reason, the face colors are
wrong for eps.  Looking at the output for pdf and png, the walls are not
shaded nearly as much as it is in eps.  There is probably something messed
up in our PS backend that is misinterpreting the grayscale color
information it is getting.

Will have to dive in some more...

Cheers!
Ben Root

I have three arrays (x,y,z).  I want plot x vs y and draw the line segments differently depending on whether or not z is positive or negative.  So I'm trying to split the x,y arrays into chunks depending on the value of z.  Using numpy.where, I can find the indeces in z that satisfy a condition but I can't figure out an efficient way (other than brute force) to split the array up into continuous chunks.  Does anyone know of a numpy trick that would help with this?

Here's a simple example:

# index:    0  1  2  3   4  5  6  7 8 9
z=numpy.array([-1,-1,-1, 1, -1,-1,-1, 1,1,1] )
x=numpy.array([-2,-3,-4, 2, -5,-6,-7, 3,4,5] )

# Want: xneg = [ x[0:3], x[4:7] ], xpos = [ x[3:4], x[7:10] ]
xneg = [ [-2,-3,-4], [-5,-6,-7] ]
xpos = [ [ 2 ], [ 3, 4, 5 ] ]

idxneg = numpy.where( z < 0 )[0]
# == [ 0,1,2, 4,5,6 ]
idxpos = numpy.where( z >= 0 )[0]
# == [ 3, 7,8,9 ]

Thanks,
Ted

Can you provide the output of the build?

On 08/02/2013 06:53 AM, Andrew Jaffe wrote:
> Hi,
>
>
> On 01/08/2013 19:06, Michael Droettboom wrote:
>> On behalf of a veritable army of super coders, I'm pleased to announce
>> the release of matplotlib 1.3.0.
> Two issues on OSX 10.8.4. I had been previously using the dmg installer.
> Lacking that, I tried easy-install and pip install, both of which gave
> me the following problems:
>
>    - I needed to set CC=clang
>    - When attempting to load matplotlib, I got the following error:
>
> /Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/font_manager.py
> in <module>()
>        51 import matplotlib
>        52 from matplotlib import afm
> ---> 53 from matplotlib import ft2font
>        54 from matplotlib import rcParams, get_cachedir
>        55 from matplotlib.cbook import is_string_like
>
> ImportError:
> dlopen(/Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/ft2font.so,
> 2): Symbol not found: _FT_Attach_File
>     Referenced from:
> /Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/ft2font.so
>     Expected in: flat namespace
>    in
> /Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/ft2font.so
>
>
> This is a freetype problem, probably an incompatible version somewhere.
> Ideas?
>
> Andrew
>
>
>
> ------------------------------------------------------------------------------
> Get your SQL database under version control now!
> Version control is standard for application code, but databases havent
> caught up. So what steps can you take to put your SQL databases under
> version control? Why should you start doing it? Read more to find out.
> http://pubads.g.doubleclick.net/gampad/clk?id=49501711&iu=/4140/ostg.clktrk
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Hi,


On 01/08/2013 19:06, Michael Droettboom wrote:
> On behalf of a veritable army of super coders, I'm pleased to announce
> the release of matplotlib 1.3.0.

Two issues on OSX 10.8.4. I had been previously using the dmg installer. 
Lacking that, I tried easy-install and pip install, both of which gave 
me the following problems:

  - I needed to set CC=clang
  - When attempting to load matplotlib, I got the following error:

/Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/font_manager.py 
in <module>()
      51 import matplotlib
      52 from matplotlib import afm
---> 53 from matplotlib import ft2font
      54 from matplotlib import rcParams, get_cachedir
      55 from matplotlib.cbook import is_string_like

ImportError: 
dlopen(/Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/ft2font.so, 
2): Symbol not found: _FT_Attach_File
   Referenced from: 
/Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/ft2font.so
   Expected in: flat namespace
  in 
/Volumes/Data/Users/jaffe/Library/Python/2.7/lib/python/site-packages/matplotlib/ft2font.so


This is a freetype problem, probably an incompatible version somewhere. 
Ideas?

Andrew

On 2013/08/01 4:23 PM, Drain, Theodore R (392P) wrote:
> I have an application where the user can edit line colors and other attributes after the plot is drawn.  The artists update just fine but the legend doesn't change.
>
>>From what I can see in the legend code, it doesn't seem like there is any mechanism in place for doing this.  Does anyone have any ideas on the best way to implement something like this?
>
> Here is a simplified script to show the issue:
>
> import pylab as p
> p.ion()
> l = p.plot( [1,2,3], 'b', label="foo" )
> p.legend()
>
> raw_input( "press return..." )
> l[0].set_color( "green" )
> p.draw()
>
> Thanks,
> Ted


If you keep a reference to the Legend object, then you can call its 
get_lines() method to get a list of Line2D objects corresponding to the 
objects returned by plot().  You can use the set_color() method on them.
Maybe this is enough if your application is simple enough.

Eric

I have an application where the user can edit line colors and other attributes after the plot is drawn.  The artists update just fine but the legend doesn't change.

>From what I can see in the legend code, it doesn't seem like there is any mechanism in place for doing this.  Does anyone have any ideas on the best way to implement something like this?

Here is a simplified script to show the issue:

import pylab as p
p.ion()
l = p.plot( [1,2,3], 'b', label="foo" )
p.legend()

raw_input( "press return..." )
l[0].set_color( "green" )
p.draw()

Thanks,
Ted