matplotlib-users Mailing List for matplotlib

I have been working to display an image of a USGS 7.5 minute quad sheet. 
These are provided at various locations about the Web.  Since the 
range of colors on these maps is limited, the *.tif files appear to 
use an indexed color map wherein each pixel has a value 0 to 255 and 
the color is found from a table with 256 entries having triplets of
the RGB in the range of 0-255.    I have not been able to sort out
how to get the gdal package to give me the color map from within python, so I dumped it
from the image file using gdalinfo and then cut and pasted to get
the following script:
 
---------------------------------------------------------------------------------
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm
from matplotlib.colors import ListedColormap

import osgeo.gdal as gdal
from osgeo.gdalconst import *

gd = gdal.Open('o37122d1.tif')

#Setup to compute the colormap from the RGB triplets from the geotif file
div = np.zeros( (256,3), np.float32)

div = 255.0
ctab = np.array([
            [ 255,255,255],
            [ 0,0,0],
            [ 255,255,255],
            [ 91,159,230],
            [ 230,45,30],
            [ 162,96,71],
            [ 210,255,177],
            [ 197,101,197],
            [ 255,240,0],
            [ 202,225,245],
            
            ...snip.... (deleted many lines:)
            [ 250,202,250],
            [ 230,230,230],
            [ 222,167,146],
            [ 255,255,255],
            [ 255,255,255] ], dtype=np.uint8)

#Compute colors in range 0.0 to 1.0. 
fctab= ctab/div

usgscm = ListedColormap(fctab, name='usgs',N=None)


doq = gd.ReadAsArray()
# doq proves to be a uint8 array: 8802 rows and 7066 columns

# Cut out a subset from the main array--to large to display and
# slow as 'molasses in January' to process:)
suba = doq[0:1101 ,0:1801 ]

fig = plt.figure()
ax = fig.add_subplot(111)
ax.imshow(suba, cmap=usgscm, origin='upper')

plt.show()
---------------------------------------------------------------------------------
This script does give me an image--but in badly wrong colors:(  The script does 
properly display gray-scaled Digital Ortho-quadrangles using cm.gray as the color
map in imshow.  Consequently something is not quite correct with respect to the 
definition or the use of the color map.  It appears that each map, and there 
are about 56,000 of them available on one site, could have its own color map.
Thus my application must be able to compute a color map unique to each of the 
topographic maps. 

Questions:

1.  What am I missing to get imshow to pick out the correct colors from the 
    color map?

2. Should I be using the name, usgs, given in the ListedColormap instance someplace?
   Not clear to me what role that name plays. 

3. Is there a way to use ctab directly in the ListedColormap instance? Class Colormap
has a bytes argument, False by default, but I am not yet sure if it has any bearing
on my problem. 
 
I am using Matplotlib 0.98 with the latest numpy, and gdal.  I am not using 
Basemap, after some study, because it contains far more "horsepower" than my 
simple topographic-map viewer application requires.  Also, this is my first 
foray into "image processing".  I am finding the learning curve a bit steep,
as usual, with multiple names for the same thing popping up in descriptions
from various sources--business as usual in the computer business:) 

Thanks from a happy matplotlib user in California!

                                               Delbert

On Monday 23 June 2008 13:25:19 John Hunter wrote:
> On Mon, Jun 23, 2008 at 11:58 AM, Erik Tollerud <eri...@gm...> 
wrote:
> > Hmm... ok, so it is possible to pass some of the text in a plot
> > through TeX, but not all of the text?  That's what the text.markup rc
> > parameter seems to be about, but I get an error saying that its an
> > unrecognized key if I use it...
>
> No, it's all or none.  But with usetex=False you can use mathtext to
> format your labels, and then only the stuff in $$ will get formatted
> as math using the matplotlib fonts and math layout engine.
>
> > I could have sworn I saw a post way back where someone managed to get
> > the tick numbers to be bold when usetex was on, but I've forgotten how
> > it was done and can't find the post now.  Anyway, it's possible that
> > I'd turned usetex off as a test and not been paying close attention to
> > turning it back on...
>
> You should be able to do this with a custom formatter::
>
>     import matplotlib.ticker as ticker
>
>     fmtbld = ticker.FormatStrFormatter(r'$\textbf{%1.2f}$')
>     ax.xaxis.set_major_formatter(fmtbld)
>
> You should also be able to use textbf in the title, xlabel and ylabel.
>
> Of course, for this simple solution, you have to hard-code your
> precision.  To take advantage of all the logic in
> ticker.ScalarFormatter, you could inherit from it and override some
> key pieces.  But it might be better if we provide some hooks for you.
>
> Darren, in ScalarFormatter.set_format, we have code like::
>
>         if self._usetex:
>             self.format = '$%s$' % self.format
>         elif self._useMathText:
>             self.format = '$\mathdefault{%s}$' % self.format
>
> We could consider exposing a font setting here.  Something along the lines
> of
>
>     self.fontcommand = None
>     if self._usetex:
>         if self.fontcommand = None:
>             self.format = '$%s$' % self.format
>         else:
>             self.format = '$%s{%s}$' % (self.fontcommand, self.format)
>
> Then the user could do::
>
>     formatter = ticker.ScalarFormatter()
>     formatter.fontcommand = r'\mathbf'
>
> and something similar for mathtext.

I'm not excited about adding user-accessible, usetex-specific properties and 
methods to classes outside of rcParams. It was a fair amount of work 
integrating as much customizability as we have now via rcParams, but we don't 
have the resources to support all the font properties in rcParams in usetex. I 
think it would be best to just use a custom formatter for this.

Thanks for the reply. The trick with clf() is that it redraws the entire
image, and hold(False), while much better, changes the axes values. Ideally
I could keep my axes setup and redraw only the bars, ala the animation
example of the plot() sine wave. Looking at that code, I'm trying something
like:

line, = bar(i + 0.25 , name_value_dict[key], color='red'), and calling
line.set_y(), but with similar issues.

Any other pointers would be greatly appreciated.

-Matt


On Sun, Jun 22, 2008 at 2:54 PM, Ryan May <rm...@gm...> wrote:

> Matt C wrote:
>
>> I'm very new to MPL, and I'm having a blast with it - great work.  I've
>> looked around  the docs, lists and  other  random places for a hint on this,
>> but I'm still stumped.
>>
>> The goal is simple: I'd like to create a very basic animated bar or barh
>> chart. I've been playing via ion() from the IPython shell but I run into the
>> following issue: The bar is painted correctly, but any value lower than the
>> max doesn't show up, as the max value bar doesn't clear. For example, from
>> IPython:
>>
>> bar(10+0.25, 10) shows perfectly, then an update (using interactive mode
>> set to on)
>> bar(10+0.25, 4)
>>
>
> Try adding a clf(), which clears the current figure, in between the calls
> to bar.  Another option is to use hold(False) signals that you want a new
> plotting command to start from a clean slate.
>
> Ryan
>
> --
> Ryan May
> Graduate Research Assistant
> School of Meteorology
> University of Oklahoma
>

On Mon, Jun 23, 2008 at 11:58 AM, Erik Tollerud <eri...@gm...> wrote:
> Hmm... ok, so it is possible to pass some of the text in a plot
> through TeX, but not all of the text?  That's what the text.markup rc
> parameter seems to be about, but I get an error saying that its an
> unrecognized key if I use it...

No, it's all or none.  But with usetex=False you can use mathtext to
format your labels, and then only the stuff in $$ will get formatted
as math using the matplotlib fonts and math layout engine.

> I could have sworn I saw a post way back where someone managed to get
> the tick numbers to be bold when usetex was on, but I've forgotten how
> it was done and can't find the post now.  Anyway, it's possible that
> I'd turned usetex off as a test and not been paying close attention to
> turning it back on...
>

You should be able to do this with a custom formatter::

    import matplotlib.ticker as ticker

    fmtbld = ticker.FormatStrFormatter(r'$\textbf{%1.2f}$')
    ax.xaxis.set_major_formatter(fmtbld)

You should also be able to use textbf in the title, xlabel and ylabel.

Of course, for this simple solution, you have to hard-code your
precision.  To take advantage of all the logic in
ticker.ScalarFormatter, you could inherit from it and override some
key pieces.  But it might be better if we provide some hooks for you.

Darren, in ScalarFormatter.set_format, we have code like::

        if self._usetex:
            self.format = '$%s$' % self.format
        elif self._useMathText:
            self.format = '$\mathdefault{%s}$' % self.format

We could consider exposing a font setting here.  Something along the lines of

    self.fontcommand = None
    if self._usetex:
        if self.fontcommand = None:
            self.format = '$%s$' % self.format
        else:
            self.format = '$%s{%s}$' % (self.fontcommand, self.format)

Then the user could do::

    formatter = ticker.ScalarFormatter()
    formatter.fontcommand = r'\mathbf'

and something similar for mathtext.

JDH

Hmm... ok, so it is possible to pass some of the text in a plot
through TeX, but not all of the text?  That's what the text.markup rc
parameter seems to be about, but I get an error saying that its an
unrecognized key if I use it...

I could have sworn I saw a post way back where someone managed to get
the tick numbers to be bold when usetex was on, but I've forgotten how
it was done and can't find the post now.  Anyway, it's possible that
I'd turned usetex off as a test and not been paying close attention to
turning it back on...

Thanks for the quick response.


On Mon, Jun 23, 2008 at 5:21 AM, Darren Dale <dsd...@gm...> wrote:
> On Sunday 22 June 2008 21:49:03 Erik Tollerud wrote:
>> I'm trying to adjust the font weight on some of my plots - I'd like to
>> have the numbers along the axis ticks be bold instead of regular font
>> like the default setting.  The problem is, nothing I do seems to
>> change the font weight.  I've changed everything I can font in
>> matplotlibrc to bold, and when I go in and explicitly look at the
>> XTick or YTick objects and their text properties, it even claims the
>> text is bold... yet the text is normal weight font when displayed on
>> the plot.
>>
>> A would also like to have the plot labels be bolder, as well, and I
>> can't seem to change that either.
>>
>> Note that I have usetex as True, and I realize that this means the
>> labels, at least, are rendered through TeX... but I'm not clear on how
>> I can change the font properties to render bold versions of all the
>> TeX fonts I'm using (assuming this is possible).
>
> There is currently no support for bold ticklabels with usetex.
>
>> And I've been under
>> the impression that the default number formattes, at least, don't
>> render through TeX.
>
> I don't understand, what default number formats? When you enable usetex, tex
> is used to render all text.
>
>> So does anyone know how I can make all/any of my
>> text elemens bold?
>
> The tex option was not designed to provide all the flexibility of matplotlib's
> text handling. It was pretty hairy trying to support as many font properties
> as we currently support.
>
> Darren
>
>

On Thu, Jun 19, 2008 at 10:35 PM, John Hunter <jd...@gm...> wrote:

> On Thu, Jun 19, 2008 at 9:19 PM, Charles Moad <cw...@gm...> wrote:
>
> > Seems like that one little fix did the trick.  I ran several examples and
> > haven't had any problems.  Committed now.
>
> Thanks Charlie!  Can you provide some basic instructions for those of
> us on osx how we can enable and test this backend?
>

I guess the two option are to set "CocoaAgg" as the default backend in your
rc config or run a script with the "-dCocoaAgg" flag.

- Charlie

Hi Michael,

Thanks!

This really does the trick. The circles look beautiful and smooth now.  ;)

You rock!

Kelvin Quee


Michael Droettboom wrote:
> You can pass in a "marker" argument with a tuple of the form (numsides, 
> style, angle) to scatter.
>
>               *numsides*:
>                 the number of sides
>
>               *style*:
>                 the style of the regular symbol:
>
>                 =====   ==================
>                 Value   Description
>                 =====   ==================
>                 0       a regular polygon
>                 1       a star-like symbol
>                 2       an asterisk
>                 =====   ==================
>
>               *angle*:
>                 the angle of rotation of the symbol
>
> For example:
>
>    scatter(marker=(100, 0, 0))
>
> You can make "numsides" as high as you need until it's smooth enough at 
> the sizes you need.
>
> On a related note, now that matplotlib can use splines to draw circles 
> (in the 0.98.x series), matplotlib should probably do that for scatter 
> circles.  This will require changing some code on our end.  This will 
> avoid requiring the kind of workaround I suggest above.
>
> Cheers,
> Mike
>
> Kelvin Quee wrote:
>   
>> Hi folks,
>>
>> I'm currently trying to plot some Bubble Charts using the scatter plot
>> function.
>>
>> It works beautifully, but the circles drawn are a little too jagged for
>> my liking.
>>
>> I'm following the example shown here -
>>
>> http://matplotlib.sourceforge.net/screenshots.html
>> (look for scatter plot)
>>
>> Even in the example picture, you can see slight jagged edges in bigger
>> circles.
>>
>> http://matplotlib.sourceforge.net/screenshots/scatter_demo2_large.png
>>
>> Does anyone have a clue on how I can improve the way the bubbles are
>> rendered?
>>
>>   
>>     
>
>   

This is the first bugfix release of the 0.98 series, with lots of
small bugs fixed and a few features added.  Thanks to Charlie Moad for
the builds: src and binaries available from
http://sourceforge.net/project/showfiles.php?group_id=80706.  For
those of you on the 0.91 maintenance branch, we have also put out a
0.91.4 bugfix release with downloads at the link above.

0.98.1 CHANGELOG:

2008-06-22 Released 0.98.1 at revision 5637

2008-06-22 Removed axes3d support and replaced it with a
          NotImplementedError for one release cycle

2008-06-21 fix marker placement bug in backend_ps - DSD

2008-06-20 [ 1978629 ] scale documentation missing/incorrect for log - MGD

2008-06-20 Added closed kwarg to PolyCollection.  Fixes bug [ 1994535
          ] still missing lines on graph with svn (r 5548). - MGD

2008-06-20 Added set/get_closed method to Polygon; fixes error
          in hist - MM

2008-06-19 Use relative font sizes (e.g. 'medium' and 'large') in
          rcsetup.py and matplotlibrc.template so that text will
          be scaled by default when changing rcParams['font.size'] -
          EF

2008-06-17 Add a generic PatchCollection class that can contain any
          kind of patch. - MGD

2008-06-13 Change pie chart label alignment to avoid having labels
          overwrite the pie - MGD

2008-06-12 Added some helper functions to the mathtext parser to
          return bitmap arrays or write pngs to make it easier to use
          mathtext outside the context of an mpl figure.  modified
          the mathpng sphinxext to use the mathtext png save
          functionality - see examples/api/mathtext_asarray.py - JDH

2008-06-11 Use matplotlib.mathtext to render math expressions in
          online docs - MGD

2008-06-11 Move PNG loading/saving to its own extension module, and
          remove duplicate code in _backend_agg.cpp and _image.cpp
          that does the same thing - MGD

2008-06-11 Numerous mathtext bugfixes, primarily related to
          dpi-independence - MGD

2008-06-10 Bar now applies the label only to the first patch only, and
          sets '_nolegend_' for the other patch labels.  This lets
          autolegend work as expected for hist and bar - see
          https://sourceforge.net/tracker/index.php?func=detail&aid=1986597&group_id=80706&atid=560720
          JDH

2008-06-10 Fix text baseline alignment bug.  [ 1985420 ] Repair of
          baseline alignment in Text._get_layout.  Thanks Stan West -
          MGD

2008-06-09 Committed Gregor's image resample patch to downsampling
          images with new rcparam image.resample - JDH

2008-06-09 Don't install Enthought.Traits along with matplotlib. For
          matplotlib developers convenience, it can still be
          installed by setting an option in setup.cfg while we figure
          decide if there is a future for the traited config - DSD

2008-06-09 Added range keyword arg to hist() - MM

2008-06-07 Moved list of backends to rcsetup.py; made use of lower
          case for backend names consistent; use validate_backend
          when importing backends subpackage - EF

2008-06-06 hist() revision, applied ideas proposed by Erik Tollerud and
          Olle Engdegard: make histtype='step' unfilled by default
          and introduce histtype='stepfilled'; use default color
          cycle; introduce reverse cumulative histogram; new align
          keyword - MM

2008-06-06 Fix closed polygon patch and also provide the option to
          not close the polygon - MGD

2008-06-05 Fix some dpi-changing-related problems with PolyCollection,
          as called by Axes.scatter() - MGD

2008-06-05 Fix image drawing so there is no extra space to the right
          or bottom - MGD

2006-06-04 Added a figure title command subtitle as a Figure method
          and pyplot command -- see examples/figure_title.py - JDH

2008-06-02 Added support for log to hist with histtype='step' and fixed
          a bug for log-scale stacked histograms - MM


0.91.4 CHANGELOG
===================


2008-06-22 Released 0.91.4 at revision 5638

2008-06-09 Don't install Enthought.Traits along with matplotlib. For
          matplotlib developers convenience, it can still be
          installed by setting an option in setup.cfg while we figure
          decide if there is a future for the traited config - DSD

2008-06-02 Fix crash in Agg when drawing collections with no fill and/or
          edge color - MGD

1. "requires python2.4" seems to imply that python 2.5 won't work.

2. "produces publication quality figures" should be "produces publication-quality figures".

Keith

You can pass in a "marker" argument with a tuple of the form (numsides, 
style, angle) to scatter.

              *numsides*:
                the number of sides

              *style*:
                the style of the regular symbol:

                =====   ==================
                Value   Description
                =====   ==================
                0       a regular polygon
                1       a star-like symbol
                2       an asterisk
                =====   ==================

              *angle*:
                the angle of rotation of the symbol

For example:

   scatter(marker=(100, 0, 0))

You can make "numsides" as high as you need until it's smooth enough at 
the sizes you need.

On a related note, now that matplotlib can use splines to draw circles 
(in the 0.98.x series), matplotlib should probably do that for scatter 
circles.  This will require changing some code on our end.  This will 
avoid requiring the kind of workaround I suggest above.

Cheers,
Mike

Kelvin Quee wrote:
> Hi folks,
>
> I'm currently trying to plot some Bubble Charts using the scatter plot
> function.
>
> It works beautifully, but the circles drawn are a little too jagged for
> my liking.
>
> I'm following the example shown here -
>
> http://matplotlib.sourceforge.net/screenshots.html
> (look for scatter plot)
>
> Even in the example picture, you can see slight jagged edges in bigger
> circles.
>
> http://matplotlib.sourceforge.net/screenshots/scatter_demo2_large.png
>
> Does anyone have a clue on how I can improve the way the bubbles are
> rendered?
>
>   

-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA

On Sunday 22 June 2008 21:49:03 Erik Tollerud wrote:
> I'm trying to adjust the font weight on some of my plots - I'd like to
> have the numbers along the axis ticks be bold instead of regular font
> like the default setting.  The problem is, nothing I do seems to
> change the font weight.  I've changed everything I can font in
> matplotlibrc to bold, and when I go in and explicitly look at the
> XTick or YTick objects and their text properties, it even claims the
> text is bold... yet the text is normal weight font when displayed on
> the plot.
>
> A would also like to have the plot labels be bolder, as well, and I
> can't seem to change that either.
>
> Note that I have usetex as True, and I realize that this means the
> labels, at least, are rendered through TeX... but I'm not clear on how
> I can change the font properties to render bold versions of all the
> TeX fonts I'm using (assuming this is possible).  

There is currently no support for bold ticklabels with usetex.

> And I've been under
> the impression that the default number formattes, at least, don't
> render through TeX.   

I don't understand, what default number formats? When you enable usetex, tex 
is used to render all text.

> So does anyone know how I can make all/any of my
> text elemens bold?

The tex option was not designed to provide all the flexibility of matplotlib's 
text handling. It was pretty hairy trying to support as many font properties 
as we currently support.

Darren

I'm trying to adjust the font weight on some of my plots - I'd like to
have the numbers along the axis ticks be bold instead of regular font
like the default setting.  The problem is, nothing I do seems to
change the font weight.  I've changed everything I can font in
matplotlibrc to bold, and when I go in and explicitly look at the
XTick or YTick objects and their text properties, it even claims the
text is bold... yet the text is normal weight font when displayed on
the plot.

A would also like to have the plot labels be bolder, as well, and I
can't seem to change that either.

Note that I have usetex as True, and I realize that this means the
labels, at least, are rendered through TeX... but I'm not clear on how
I can change the font properties to render bold versions of all the
TeX fonts I'm using (assuming this is possible).  And I've been under
the impression that the default number formattes, at least, don't
render through TeX.   So does anyone know how I can make all/any of my
text elemens bold?