Thread: [Matplotlib-users] how to display data with a "fixed" color scale ?

I want to generate a 2-d figure with a (fixed) color scale that  does 
not vary with the range of the data being plotted.

How do I do this?  Attempts to specify vimin and vmax appear to be ignored.

The following example:

#<code>
import numpy
data = numpy.zeros(shape=(240,240),dtype=int)
data[  0: 80] = -1
data[ 80:160] =  0
data[160:]    =  1

import matplotlib.pyplot as plot
figure = plot.figure()
ax     = figure.add_subplot(111)
cax    = ax.imshow(data, interpolation='bilinear')
ax.set_title('test data with fixed colorbar')
colorbar = figure.colorbar(cax, ticks=[-1, 0, 1])
colorbar.ax.set_yticklabels(['-1', '0', '1'])
plot.show()
#</code>

produces a figure with 3 color bands (blue,green,red) and matching color 
bar with labels (-1,0,1) as expected.

if the data[160:]=1 specification is deleted, in the above code, the 
resulting figure has 2 color bands (blue,red) and the associated color 
bar is identical to the original, but the labels are (-1,0).

What I want, in this second case, is a blue-green figure and a color bar 
with labels identical to the original example.

-- jv

On Thu, Jun 3, 2010 at 4:00 PM, Jim Vickroy <Jim...@no...> wrote:
> How do I do this?  Attempts to specify vimin and vmax appear to be ignored.
>

Hmm,
vmin and vmax should work.

cax    = ax.imshow(data, interpolation='bilinear', vmin=-1, vmax=1)

If these are still ignored, what the following line prints?

print cax.get_clim()

Regards,

-JJ

Jae-Joon Lee wrote:
> On Thu, Jun 3, 2010 at 4:00 PM, Jim Vickroy <Jim...@no...> wrote:
>   
>> How do I do this?  Attempts to specify vimin and vmax appear to be ignored.
>>
>>     
>
> Hmm,
> vmin and vmax should work.
>
> cax    = ax.imshow(data, interpolation='bilinear', vmin=-1, vmax=1)
>
> If these are still ignored, what the following line prints?
>
> print cax.get_clim()
>
> Regards,
>
> -JJ
>   
You are absolutely correct!  My apologies for not testing this prior to 
posting.

I'm really specifying vmin and vmax as parameters for pcolormesh and the 
labels, for the associated color bar, vary (despite being explicitly 
specified in figure.colorbar(...,ticks=...)) with the range of the data 
being plotted.

I over-simplified, so I will try to produce a better version of the 
actual code demonstrating the problem.

Thank-you for your quick reply.

-- jv

On 06/03/2010 10:00 AM, Jim Vickroy wrote:
> I want to generate a 2-d figure with a (fixed) color scale that  does
> not vary with the range of the data being plotted.
>
> How do I do this? Attempts to specify vimin and vmax appear to be ignored.
>
> The following example:
>
> #<code>
> import numpy
> data = numpy.zeros(shape=(240,240),dtype=int)
> data[ 0: 80] = -1
> data[ 80:160] = 0
> data[160:] = 1
>
> import matplotlib.pyplot as plot
> figure = plot.figure()
> ax = figure.add_subplot(111)
> cax = ax.imshow(data, interpolation='bilinear')
> ax.set_title('test data with fixed colorbar')

Adding to what JJ said, note that setting the ticks on the colorbar has 
no effect on the norm used in color mapping.  The vmin and vmax kwargs 
to imshow get passed to the norm, so they do set the mapping range.

> colorbar = figure.colorbar(cax, ticks=[-1, 0, 1])
> colorbar.ax.set_yticklabels(['-1', '0', '1'])

Please avoid setting the ticklabels directly--it is almost always 
unnecessary, and it is too easy to shoot yourself in the foot.  If the 
default tick label formatting is inadequate, you can use the format 
kwarg in colorbar.

 From the docstring:

         *ticks*       [ None | list of ticks | Locator object ]
                       If None, ticks are determined automatically from the
                       input.
         *format*      [ None | format string | Formatter object ]
                       If None, the
                       :class:`~matplotlib.ticker.ScalarFormatter` is used.
                       If a format string is given, e.g. '%.3f', that is
                       used. An alternative
                       :class:`~matplotlib.ticker.Formatter` object may be
                       given instead.


Eric


> plot.show()
> #</code>
>
> produces a figure with 3 color bands (blue,green,red) and matching color
> bar with labels (-1,0,1) as expected.
>
> if the data[160:]=1 specification is deleted, in the above code, the
> resulting figure has 2 color bands (blue,red) and the associated color
> bar is identical to the original, but the labels are (-1,0).
>
> What I want, in this second case, is a blue-green figure and a color bar
> with labels identical to the original example.
>
> -- jv
>
>
>
> ------------------------------------------------------------------------------
> ThinkGeek and WIRED's GeekDad team up for the Ultimate
> GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
> lucky parental unit.  See the prize list and enter to win:
> http://p.sf.net/sfu/thinkgeek-promo
>
>
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Eric Firing wrote:
> On 06/03/2010 10:00 AM, Jim Vickroy wrote:
>   
>> I want to generate a 2-d figure with a (fixed) color scale that  does
>> not vary with the range of the data being plotted.
>>
>> How do I do this? Attempts to specify vimin and vmax appear to be ignored.
>>
>> The following example:
>>
>> #<code>
>> import numpy
>> data = numpy.zeros(shape=(240,240),dtype=int)
>> data[ 0: 80] = -1
>> data[ 80:160] = 0
>> data[160:] = 1
>>
>> import matplotlib.pyplot as plot
>> figure = plot.figure()
>> ax = figure.add_subplot(111)
>> cax = ax.imshow(data, interpolation='bilinear')
>> ax.set_title('test data with fixed colorbar')
>>     
>
> Adding to what JJ said, note that setting the ticks on the colorbar has 
> no effect on the norm used in color mapping.  The vmin and vmax kwargs 
> to imshow get passed to the norm, so they do set the mapping range.
>
>   
>> colorbar = figure.colorbar(cax, ticks=[-1, 0, 1])
>> colorbar.ax.set_yticklabels(['-1', '0', '1'])
>>     
>
> Please avoid setting the ticklabels directly--it is almost always 
> unnecessary, and it is too easy to shoot yourself in the foot.  If the 
> default tick label formatting is inadequate, you can use the format 
> kwarg in colorbar.
>
>  From the docstring:
>
>          *ticks*       [ None | list of ticks | Locator object ]
>                        If None, ticks are determined automatically from the
>                        input.
>          *format*      [ None | format string | Formatter object ]
>                        If None, the
>                        :class:`~matplotlib.ticker.ScalarFormatter` is used.
>                        If a format string is given, e.g. '%.3f', that is
>                        used. An alternative
>                        :class:`~matplotlib.ticker.Formatter` object may be
>                        given instead.
>
>
> Eric
>   

Thanks for this advice.

In my case, the data being plotted is in the range 0-255, but the 
color-bar labels are to be in the range 1-4095.  So I have the following 
code snippet:

   colorbar = figure.colorbar(image, cax, orientation='vertical', 
ticks=(0, 64, 128, 192, 254))
   colorbar.ax.set_yticklabels(('1','8','64','512','4095')) # colorbar 
labels (which are to be in units of DN/sec on a log10 scale)

Is there a better way to do this?

-- jv


>
>   
>> plot.show()
>> #</code>
>>
>> produces a figure with 3 color bands (blue,green,red) and matching color
>> bar with labels (-1,0,1) as expected.
>>
>> if the data[160:]=1 specification is deleted, in the above code, the
>> resulting figure has 2 color bands (blue,red) and the associated color
>> bar is identical to the original, but the labels are (-1,0).
>>
>> What I want, in this second case, is a blue-green figure and a color bar
>> with labels identical to the original example.
>>
>> -- jv
>>
>>
>>
>> ------------------------------------------------------------------------------
>> ThinkGeek and WIRED's GeekDad team up for the Ultimate
>> GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
>> lucky parental unit.  See the prize list and enter to win:
>> http://p.sf.net/sfu/thinkgeek-promo
>>
>>
>>
>> _______________________________________________
>> Matplotlib-users mailing list
>> Mat...@li...
>> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>>     
>
>
> ------------------------------------------------------------------------------
> ThinkGeek and WIRED's GeekDad team up for the Ultimate 
> GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the 
> lucky parental unit.  See the prize list and enter to win: 
> http://p.sf.net/sfu/thinkgeek-promo
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>   

On 06/03/2010 10:43 AM, Jim Vickroy wrote:
> Eric Firing wrote:
>> On 06/03/2010 10:00 AM, Jim Vickroy wrote:
>>
>>> I want to generate a 2-d figure with a (fixed) color scale that  does
>>> not vary with the range of the data being plotted.
>>>
>>> How do I do this? Attempts to specify vimin and vmax appear to be ignored.
>>>
>>> The following example:
>>>
>>> #<code>
>>> import numpy
>>> data = numpy.zeros(shape=(240,240),dtype=int)
>>> data[ 0: 80] = -1
>>> data[ 80:160] = 0
>>> data[160:] = 1
>>>
>>> import matplotlib.pyplot as plot
>>> figure = plot.figure()
>>> ax = figure.add_subplot(111)
>>> cax = ax.imshow(data, interpolation='bilinear')
>>> ax.set_title('test data with fixed colorbar')
>>>
>>
>> Adding to what JJ said, note that setting the ticks on the colorbar has
>> no effect on the norm used in color mapping.  The vmin and vmax kwargs
>> to imshow get passed to the norm, so they do set the mapping range.
>>
>>
>>> colorbar = figure.colorbar(cax, ticks=[-1, 0, 1])
>>> colorbar.ax.set_yticklabels(['-1', '0', '1'])
>>>
>>
>> Please avoid setting the ticklabels directly--it is almost always
>> unnecessary, and it is too easy to shoot yourself in the foot.  If the
>> default tick label formatting is inadequate, you can use the format
>> kwarg in colorbar.
>>
>>   From the docstring:
>>
>>           *ticks*       [ None | list of ticks | Locator object ]
>>                         If None, ticks are determined automatically from the
>>                         input.
>>           *format*      [ None | format string | Formatter object ]
>>                         If None, the
>>                         :class:`~matplotlib.ticker.ScalarFormatter` is used.
>>                         If a format string is given, e.g. '%.3f', that is
>>                         used. An alternative
>>                         :class:`~matplotlib.ticker.Formatter` object may be
>>                         given instead.
>>
>>
>> Eric
>>
>
> Thanks for this advice.
>
> In my case, the data being plotted is in the range 0-255, but the
> color-bar labels are to be in the range 1-4095. So I have the following
> code snippet:
>
> colorbar = figure.colorbar(image, cax, orientation='vertical', ticks=(0,
> 64, 128, 192, 254))
> colorbar.ax.set_yticklabels(('1','8','64','512','4095')) # colorbar
> labels (which are to be in units of DN/sec on a log10 scale)
>
> Is there a better way to do this?

The advantage of using a custom formatter is that it formats actual tick 
values, so if you decide to use a different set of tick locations, you 
don't have to remember to change the labels.  A formatter for a 
complicated case such as the above could use a dictionary, which would 
at least generate a KeyError if you changed a tick without adding the 
new location to the dictionary, or, better, it could calculate the label 
numbers.  Suppose you have a function to do the translation:

def to_DNpersec(x):
     dn = ... whatever function of x
     return dn

import matplotlib as mpl

class DNpersecFormatter(mpl.ticker.Formatter):
     def __call__(self, val, pos=None):
         dn = to_DNpersec(val)
         return "%d" % round(dn)

...
colorbar = figure.colorbar(image, cax, orientation='vertical',
                              ticks=(0, 64, 128, 192, 254),
                              format=DNpersecFormatter())

Eric

>
> -- jv
>
>
>>
>>
>>> plot.show()
>>> #</code>
>>>
>>> produces a figure with 3 color bands (blue,green,red) and matching color
>>> bar with labels (-1,0,1) as expected.
>>>
>>> if the data[160:]=1 specification is deleted, in the above code, the
>>> resulting figure has 2 color bands (blue,red) and the associated color
>>> bar is identical to the original, but the labels are (-1,0).
>>>
>>> What I want, in this second case, is a blue-green figure and a color bar
>>> with labels identical to the original example.
>>>
>>> -- jv

OK, upon  a more careful review of the code, I made a simple(-minded) 
error.  Specifying vmin and vmax do work (as everyone already knew).

Thanks to Jae-Joon and Eric for their quick replies. and valuable 
suggestions.

-- jv

Jim Vickroy wrote:
> I want to generate a 2-d figure with a (fixed) color scale that  does 
> not vary with the range of the data being plotted.
>
> How do I do this?  Attempts to specify vimin and vmax appear to be 
> ignored.
>
> The following example:
>
> #<code>
> import numpy
> data = numpy.zeros(shape=(240,240),dtype=int)
> data[  0: 80] = -1
> data[ 80:160] =  0
> data[160:]    =  1
>
> import matplotlib.pyplot as plot
> figure = plot.figure()
> ax     = figure.add_subplot(111)
> cax    = ax.imshow(data, interpolation='bilinear')
> ax.set_title('test data with fixed colorbar')
> colorbar = figure.colorbar(cax, ticks=[-1, 0, 1])
> colorbar.ax.set_yticklabels(['-1', '0', '1'])
> plot.show()
> #</code>
>
> produces a figure with 3 color bands (blue,green,red) and matching 
> color bar with labels (-1,0,1) as expected.
>
> if the data[160:]=1 specification is deleted, in the above code, the 
> resulting figure has 2 color bands (blue,red) and the associated color 
> bar is identical to the original, but the labels are (-1,0).
>
> What I want, in this second case, is a blue-green figure and a color 
> bar with labels identical to the original example.
>
> -- jv
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> ThinkGeek and WIRED's GeekDad team up for the Ultimate 
> GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the 
> lucky parental unit.  See the prize list and enter to win: 
> http://p.sf.net/sfu/thinkgeek-promo
> ------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>