matplotlib-users Mailing List for matplotlib

On Mon, Dec 8, 2008 at 1:58 PM, Ryan May <rm...@gm...> wrote:
> John Hunter wrote:
> The scaling changes are in, as well as the warning and the corresponding lines in
> api_changes and CHANGELOG. I also added the converted matlab demo I used to
> figure this stuff out. Now would probably be the time to see if I did something
> wrong (especially the warning).
Question from the peanut gallery: are you guys putting in numerical
tests as part of your unit test suite for this stuff? Rather than in
the future saying 'I think in December 2008 it used to match matlab'
it would be nice to have some automated testing of this. Useful if
someone ever wants to optimize it, cythonize it, etc...
Just a thought...
f

John,
I remember that "-" is also rendedered as a dot in the HTML output.
Using "\-" instead of "-" seems to work. But, I'm afraid that the help
(or similar) command, which I frequently use in interactive sessions,
may show "\-" instead of "-".
So, quoting seems to be a best option to me.
Regards,
-JJ
On Mon, Dec 8, 2008 at 5:04 PM, John Hunter <jd...@gm...> wrote:
> In the Curve class of patches.py, where we are doing:
>
> _style_list["-"] = Curve
>
> and interpolating this into a rest table via patches._pprint_styles,
> which looks like this::
>
> ======== ====== =============================================
> Class Name Attrs
> ======== ====== =============================================
> Curve - None
> CurveB -> head_length=0.4,head_width=0.2
> BracketB -[ widthB=1.0,lengthB=0.2,angleB=None
> CurveA <- head_length=0.4,head_width=0.2
> CurveAB <-> head_length=0.4,head_width=0.2
> Fancy fancy head_length=0.4,head_width=0.4,tail_width=0.4
> Simple simple head_length=0.5,head_width=0.5,tail_width=0.2
> Wedge wedge tail_width=0.3,shrink_factor=0.5
> ======== ====== =============================================
>
>
> tex is inserting the following into the tex doc::
>
> Curve
> & \begin{itemize}
> \item {}
> \end{itemize}
> &
> None
> \\
>
> It looks like it thinks the hyphen is starting an itemize list, and
> this is breaking the tex table environment. Any ideas how to fix
> this?
>
> For now, I am going to quote the Name column with single quotes, eg '-' and '->'
>
> Thanks,
> JDH
>
> ------------------------------------------------------------------------------
> SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
> The future of the web can't happen without you. Join us at MIX09 to help
> pave the way to the Next Web now. Learn more and register at
> http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>

I've been looking for an elegant way to change, on a per-plot basis, 
the font attributes of my x and y tick labels. The best I've come up 
with is getting a list/collection of tick labels and looping through 
it to change the font attributes. Is there an easy one or two line 
way to do it?
--Chad Kidder

In the Curve class of patches.py, where we are doing:
_style_list["-"] = Curve
and interpolating this into a rest table via patches._pprint_styles,
which looks like this::
 ======== ====== =============================================
 Class Name Attrs
 ======== ====== =============================================
 Curve - None
 CurveB -> head_length=0.4,head_width=0.2
 BracketB -[ widthB=1.0,lengthB=0.2,angleB=None
 CurveA <- head_length=0.4,head_width=0.2
 CurveAB <-> head_length=0.4,head_width=0.2
 Fancy fancy head_length=0.4,head_width=0.4,tail_width=0.4
 Simple simple head_length=0.5,head_width=0.5,tail_width=0.2
 Wedge wedge tail_width=0.3,shrink_factor=0.5
 ======== ====== =============================================
tex is inserting the following into the tex doc::
 Curve
 & \begin{itemize}
 \item {}
 \end{itemize}
 &
 None
 \\
It looks like it thinks the hyphen is starting an itemize list, and
this is breaking the tex table environment. Any ideas how to fix
this?
For now, I am going to quote the Name column with single quotes, eg '-' and '->'
Thanks,
JDH

John Hunter wrote:
> On Mon, Dec 8, 2008 at 10:56 AM, Ryan May <rm...@gm...> wrote:
> 
>> My only other concern is whether this belongs in 0.98.x. This is a behavior
>> change from 0.98.3, not necessarily a bug fix. I'll defer to John, et al.
>> on whether this should go in 0.98.x or go in a later release.
> 
> It's a judgement call, but we have always had new features and minor
> API changes in the point releases. Except on the 0.91 maintenance
> branch, which is only bugfix, we have continuously added new stuff
> on every release. When the breakage is likely to be difficult, or the
> new feature really significant, we will push the major version. I
> don't think these changes are so significant that they require waiting
> until a new major version number, but you may want to consider issuing
> a warning in addition to the requisite explanation in the docstring
> and CHANGELOG.
The scaling changes are in, as well as the warning and the corresponding lines in 
 api_changes and CHANGELOG. I also added the converted matlab demo I used to 
figure this stuff out. Now would probably be the time to see if I did something 
wrong (especially the warning).
Ryan
-- 
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

On Dec 8, 2008, at 2:00 PM, Ryan May wrote:
> Chad Kidder wrote:
>> On Dec 8, 2008, at 12:55 PM, Ryan May wrote:
>>> Chad Kidder wrote:
>>>> I've got many series of data that I want to plot, and each has 
>>>> an additional scalar that is valid for the whole series. What 
>>>> I want to do is plot all these series on top of each other 
>>>> (plot can do this just fine), but with the additional scalar 
>>>> changing the color, efectively using color as the z-axis. I'm 
>>>> not seeing how to do that. If there was a function where I 
>>>> could give a color map a value and it would spit out the color, 
>>>> that would work, but I haven't seen it. Thanks for your help.
>>> Try using scatter instead of plot. Specifically, the 'c' keyword 
>>> argument:
>>>
>>> *c*:
>>> a color. *c* can be a single color format string, or a
>>> sequence of color specifications of length *N*, or a
>>> sequence of *N* numbers to be mapped to colors using the
>>> *cmap* and *norm* specified via kwargs (see below). Note
>>> that *c* should not be a single numeric RGB or RGBA
>>> sequence because that is indistinguishable from an array
>>> of values to be colormapped. *c* can be a 2-D array in
>>> which the rows are RGB or RGBA, however.
>
> Here's what you're looking for:
>
> import numpy as n
> import matplotlib.pyplot as p
> import matplotlib.colors as mcolors
> import matplotlib.cm as cm
>
> cmap = cm.get_cmap('winter')
> norm = mcolors.Normalize(0, 1) #Range of z
>
> nlines = 100
> z = n.random.rand(nlines)
> x = n.arange(nlines)
> t1, t2 = n.meshgrid(x,z)
> y = t1+t2-0.5
>
> #Uses normalize to map z values to range of 0 to 1.
> #Cmap maps these normalized values to colors
> colors = cmap(norm(z))
> for ii in range(nlines):
> p.plot(x, y[ii], color=colors[ii])
> p.show()
>
> Ryan
Thanks all. I had tried something close, but it didn't work. This 
works great.

Chad Kidder wrote:
> On Dec 8, 2008, at 12:55 PM, Ryan May wrote:
> 
>> Chad Kidder wrote:
>>> I've got many series of data that I want to plot, and each has an 
>>> additional scalar that is valid for the whole series. What I want 
>>> to do is plot all these series on top of each other (plot can do 
>>> this just fine), but with the additional scalar changing the 
>>> color, efectively using color as the z-axis. I'm not seeing how 
>>> to do that. If there was a function where I could give a color 
>>> map a value and it would spit out the color, that would work, but 
>>> I haven't seen it. Thanks for your help.
>> Try using scatter instead of plot. Specifically, the 'c' keyword 
>> argument:
>>
>> *c*:
>> a color. *c* can be a single color format string, or a
>> sequence of color specifications of length *N*, or a
>> sequence of *N* numbers to be mapped to colors using the
>> *cmap* and *norm* specified via kwargs (see below). Note
>> that *c* should not be a single numeric RGB or RGBA
>> sequence because that is indistinguishable from an array
>> of values to be colormapped. *c* can be a 2-D array in
>> which the rows are RGB or RGBA, however.
Here's what you're looking for:
import numpy as n
import matplotlib.pyplot as p
import matplotlib.colors as mcolors
import matplotlib.cm as cm
cmap = cm.get_cmap('winter')
norm = mcolors.Normalize(0, 1) #Range of z
nlines = 100
z = n.random.rand(nlines)
x = n.arange(nlines)
t1, t2 = n.meshgrid(x,z)
y = t1+t2-0.5
#Uses normalize to map z values to range of 0 to 1.
#Cmap maps these normalized values to colors
colors = cmap(norm(z))
for ii in range(nlines):
 p.plot(x, y[ii], color=colors[ii])
p.show()
Ryan
-- 
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

Chad Kidder wrote:
> On Dec 8, 2008, at 12:55 PM, Ryan May wrote:
> 
>> Chad Kidder wrote:
>>> I've got many series of data that I want to plot, and each has an 
>>> additional scalar that is valid for the whole series. What I want 
>>> to do is plot all these series on top of each other (plot can do 
>>> this just fine), but with the additional scalar changing the 
>>> color, efectively using color as the z-axis. I'm not seeing how 
>>> to do that. If there was a function where I could give a color 
>>> map a value and it would spit out the color, that would work, but 
>>> I haven't seen it. Thanks for your help.
>> Try using scatter instead of plot. Specifically, the 'c' keyword 
>> argument:
>>
>> *c*:
>> a color. *c* can be a single color format string, or a
>> sequence of color specifications of length *N*, or a
>> sequence of *N* numbers to be mapped to colors using the
>> *cmap* and *norm* specified via kwargs (see below). Note
>> that *c* should not be a single numeric RGB or RGBA
>> sequence because that is indistinguishable from an array
>> of values to be colormapped. *c* can be a 2-D array in
>> which the rows are RGB or RGBA, however.
>>
>> For example:
>>
>> import numpy as np
>> import matplotlib.pyplot as plt
>>
>> x = np.random.randn(100)
>> y = np.random.randn(100)
>> data = x**2 + y**2
>> plt.scatter(x, y, c=data)
>> plt.show()
>>
>> Ryan
>>
> 
> Close, but not quite what I want. Maybe this will tell what I want to 
> do better:
> 
> ----------------------
> import numpy as n
> import matplotlib.pyplot as p
> 
> nlines = 100
> z = n.random.rand(nlines)
> x = n.array(range(nlines))
> t1, t2 = n.meshgrid(x,z)
> y = t1+t2-0.5
> for ii in range(nlines):
> p.plot(x,y[ii,:],color = str(z[ii]))
> p.show()
> -------------------
> Instead of getting a grayscale plot out, I'd like to use a colormap 
> like jet() or winter(). Any ideas there?
In [6]:cmap = get_cmap('jet')
In [7]:cmap(0.2)
Out[7]:(0.0, 0.29999999999999999, 1.0, 1.0)
In [8]:cmap(0.8)
Out[8]:(1.0, 0.40740740740740755, 0.0, 1.0)
The pyplot.get_cmap() function gets a colormap by name. Calling that 
colormap with a floating-point argument in the 0-1 range returns the 
mapped color as an rgba tuple, which will be accepted by the color kwarg 
of plot. You can use pyplot.normalize to map your z range to the 0-1 range:
In [2]:norm = normalize(vmin=2, vmax=4)
In [3]:norm(3)
Out[3]:0.5
Alternatively, you can use a LineCollection. See the 
examples/pylab_examples/line_collection2.py script.
Eric
> 
> 
> 
> 
> 
> --Chad Kidder
> 
> ------------------------------------------------------------------------------
> SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
> The future of the web can't happen without you. Join us at MIX09 to help
> pave the way to the Next Web now. Learn more and register at
> http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

On Mon, Dec 8, 2008 at 1:25 PM, Chad Kidder <cck...@gm...> wrote:
-------------------
> Instead of getting a grayscale plot out, I'd like to use a colormap
> like jet() or winter(). Any ideas there?
How about?
 import matplotlib.cm as cm
 for ii in range(nlines):
 color = cm.jet(z[ii])
 p.plot(x,y[ii,:],color=color)
All of the mpl colormaps are callable, so if you pass in a [0..1]
normalized value they will return an RGB tuple
JDH

On Dec 8, 2008, at 12:55 PM, Ryan May wrote:
> Chad Kidder wrote:
>> I've got many series of data that I want to plot, and each has an 
>> additional scalar that is valid for the whole series. What I want 
>> to do is plot all these series on top of each other (plot can do 
>> this just fine), but with the additional scalar changing the 
>> color, efectively using color as the z-axis. I'm not seeing how 
>> to do that. If there was a function where I could give a color 
>> map a value and it would spit out the color, that would work, but 
>> I haven't seen it. Thanks for your help.
>
> Try using scatter instead of plot. Specifically, the 'c' keyword 
> argument:
>
> *c*:
> a color. *c* can be a single color format string, or a
> sequence of color specifications of length *N*, or a
> sequence of *N* numbers to be mapped to colors using the
> *cmap* and *norm* specified via kwargs (see below). Note
> that *c* should not be a single numeric RGB or RGBA
> sequence because that is indistinguishable from an array
> of values to be colormapped. *c* can be a 2-D array in
> which the rows are RGB or RGBA, however.
>
> For example:
>
> import numpy as np
> import matplotlib.pyplot as plt
>
> x = np.random.randn(100)
> y = np.random.randn(100)
> data = x**2 + y**2
> plt.scatter(x, y, c=data)
> plt.show()
>
> Ryan
>
Close, but not quite what I want. Maybe this will tell what I want to 
do better:
----------------------
import numpy as n
import matplotlib.pyplot as p
nlines = 100
z = n.random.rand(nlines)
x = n.array(range(nlines))
t1, t2 = n.meshgrid(x,z)
y = t1+t2-0.5
for ii in range(nlines):
 p.plot(x,y[ii,:],color = str(z[ii]))
p.show()
-------------------
Instead of getting a grayscale plot out, I'd like to use a colormap 
like jet() or winter(). Any ideas there?
--Chad Kidder

John Hunter wrote:
> On Mon, Dec 8, 2008 at 10:56 AM, Ryan May <rm...@gm...> wrote:
> 
>> My only other concern is whether this belongs in 0.98.x. This is a behavior
>> change from 0.98.3, not necessarily a bug fix. I'll defer to John, et al.
>> on whether this should go in 0.98.x or go in a later release.
> 
> It's a judgement call, but we have always had new features and minor
> API changes in the point releases. Except on the 0.91 maintenance
> branch, which is only bugfix, we have continuously added new stuff
> on every release. When the breakage is likely to be difficult, or the
> new feature really significant, we will push the major version. I
> don't think these changes are so significant that they require waiting
> until a new major version number, but you may want to consider issuing
> a warning in addition to the requisite explanation in the docstring
> and CHANGELOG.
How about making the default None, and if the default is used set to True and 
issue the warning for this release. In the future the warning is remove and the 
default goes to True.
As long as we're going for Matlab compatibility, what about changing Fs from 
defaulting to 2 to 2*pi? That seems to be the default.
Ryan
-- 
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

Chad Kidder wrote:
> I've got many series of data that I want to plot, and each has an 
> additional scalar that is valid for the whole series. What I want to 
> do is plot all these series on top of each other (plot can do this 
> just fine), but with the additional scalar changing the color, 
> efectively using color as the z-axis. I'm not seeing how to do that. 
> If there was a function where I could give a color map a value and it 
> would spit out the color, that would work, but I haven't seen it. 
> Thanks for your help.
Try using scatter instead of plot. Specifically, the 'c' keyword argument:
 *c*:
 a color. *c* can be a single color format string, or a
 sequence of color specifications of length *N*, or a
 sequence of *N* numbers to be mapped to colors using the
 *cmap* and *norm* specified via kwargs (see below). Note
 that *c* should not be a single numeric RGB or RGBA
 sequence because that is indistinguishable from an array
 of values to be colormapped. *c* can be a 2-D array in
 which the rows are RGB or RGBA, however.
For example:
import numpy as np
import matplotlib.pyplot as plt
x = np.random.randn(100)
y = np.random.randn(100)
data = x**2 + y**2
plt.scatter(x, y, c=data)
plt.show()
Ryan
-- 
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

On Mon, Dec 8, 2008 at 12:45 PM, Chad Kidder <cck...@gm...> wrote:
> I've got many series of data that I want to plot, and each has an
> additional scalar that is valid for the whole series. What I want to
> do is plot all these series on top of each other (plot can do this
> just fine), but with the additional scalar changing the color,
> efectively using color as the z-axis. I'm not seeing how to do that.
> If there was a function where I could give a color map a value and it
> would spit out the color, that would work, but I haven't seen it.
> Thanks for your help.
Check out the scatter_demo -- scatter takes an optional argument 'c'
for the color and an optional colormap
http://matplotlib.sourceforge.net/users/screenshots.html#scatter-demo
>
> --Chad Kidder
>
>
>
> ------------------------------------------------------------------------------
> SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
> The future of the web can't happen without you. Join us at MIX09 to help
> pave the way to the Next Web now. Learn more and register at
> http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>

I've got many series of data that I want to plot, and each has an 
additional scalar that is valid for the whole series. What I want to 
do is plot all these series on top of each other (plot can do this 
just fine), but with the additional scalar changing the color, 
efectively using color as the z-axis. I'm not seeing how to do that. 
If there was a function where I could give a color map a value and it 
would spit out the color, that would work, but I haven't seen it. 
Thanks for your help.
--Chad Kidder

Dear Jeff,
2008年12月8日 Jeff Whitaker <js...@fa...>:
> Mauro: That's because the default is not really a map projection at all -
> it just displays the data in lat/lon coordinates. The map scale really
> has no meaning in that case.
Sure, however some biological journals insist that authors provide
these scales (and in at least one case I known of, also a wind
rose!!!!) on all maps, even if when they have no meaning (as is the
case with the Equirectangular projection). Regrettably, as I mentioned
earlier, there have been no discussion of map projections -- their
meaning and significance -- among biogeographers.
With best regards,
-- 
Dr. Mauro J. Cavalcanti
Ecoinformatics Studio
P.O. Box 46521, CEP 20551-970
Rio de Janeiro, RJ, BRASIL
E-mail: mau...@gm...
Web: http://studio.infobio.net
Linux Registered User #473524 * Ubuntu User #22717
"Life is complex. It consists of real and imaginary parts."

Mauro Cavalcanti wrote:
> Dear Jeff,
>
> Thanks.... Unfortunately, the map scale cannot be drawn for
> cylindrical projections like the Equirectangular default... :-(
> 
Mauro: That's because the default is not really a map projection at all 
- it just displays the data in lat/lon coordinates. The map scale 
really has no meaning in that case.
-Jeff
> Best wishes,
>
> 2008年12月8日 Jeff Whitaker <js...@fa...>:
> 
>> Mauro Cavalcanti wrote:
>> 
>>> Dear Jeff & ALL,
>>>
>>> Are there any examples of the use of the drawmapscale() Basemap
>>> method? I can use this myself and will also add any examples to the
>>> forthcoming "Basemap Cookbook" website (for this, I am compiling all
>>> available examples).
>>>
>>> Thanks in advance!
>>>
>>> Best wishes,
>>>
>>>
>>> 
>> Mauro:
>>
>> see the ortho_demo.py example.
>>
>> -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
>>
>>
>> 
>
>
>
> 
-- 
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

Dear Jeff,
Thanks.... Unfortunately, the map scale cannot be drawn for
cylindrical projections like the Equirectangular default... :-(
Best wishes,
2008年12月8日 Jeff Whitaker <js...@fa...>:
> Mauro Cavalcanti wrote:
>>
>> Dear Jeff & ALL,
>>
>> Are there any examples of the use of the drawmapscale() Basemap
>> method? I can use this myself and will also add any examples to the
>> forthcoming "Basemap Cookbook" website (for this, I am compiling all
>> available examples).
>>
>> Thanks in advance!
>>
>> Best wishes,
>>
>>
>
> Mauro:
>
> see the ortho_demo.py example.
>
> -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
>
>
-- 
Dr. Mauro J. Cavalcanti
Ecoinformatics Studio
P.O. Box 46521, CEP 20551-970
Rio de Janeiro, RJ, BRASIL
E-mail: mau...@gm...
Web: http://studio.infobio.net
Linux Registered User #473524 * Ubuntu User #22717
"Life is complex. It consists of real and imaginary parts."

Mauro Cavalcanti wrote:
> Dear Jeff & ALL,
>
> Are there any examples of the use of the drawmapscale() Basemap
> method? I can use this myself and will also add any examples to the
> forthcoming "Basemap Cookbook" website (for this, I am compiling all
> available examples).
>
> Thanks in advance!
>
> Best wishes,
>
> 
Mauro:
see the ortho_demo.py example.
-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

Dear Jeff & ALL,
Are there any examples of the use of the drawmapscale() Basemap
method? I can use this myself and will also add any examples to the
forthcoming "Basemap Cookbook" website (for this, I am compiling all
available examples).
Thanks in advance!
Best wishes,
-- 
Dr. Mauro J. Cavalcanti
Ecoinformatics Studio
P.O. Box 46521, CEP 20551-970
Rio de Janeiro, RJ, BRASIL
E-mail: mau...@gm...
Web: http://studio.infobio.net
Linux Registered User #473524 * Ubuntu User #22717
"Life is complex. It consists of real and imaginary parts."

Mauro Cavalcanti wrote:
> Dear Jeff,
>
> 2008年12月8日 Jeff Whitaker <js...@fa...>:
> 
>> Mauro: I just added a 'scale' keyword to bluemarble that downsamples the
>> image to speed things up. scale=1.0 (the default) gives the same answer as
>> before (the full resolution image). scale=0.5 downsamples the image to half
>> the original size, speeding things up considerably (and using a lot less
>> memory, which I think is the real problem).
>> 
>
> Thank you very much for that! Yes, the real problem with the slow
> plotting of Blue Marble very probably concerns the amount of available
> memory and not the processor speed. I will try it ASAP. I suppose this
> modification is also available only via SVN? Is there a time schedule
> for releasing a new distribution version incorporating this recent
> changes? I need a setup package that can be downloaded and installed
> by computer-unsophisticated Windows users :-(
>
> More soon...
>
> With best wishes,
>
> 
Mauro:
I can make a new release shortly, after the new changes have been tested 
a bit.
-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

Dear Jeff,
2008年12月8日 Jeff Whitaker <js...@fa...>:
> Mauro: I just added a 'scale' keyword to bluemarble that downsamples the
> image to speed things up. scale=1.0 (the default) gives the same answer as
> before (the full resolution image). scale=0.5 downsamples the image to half
> the original size, speeding things up considerably (and using a lot less
> memory, which I think is the real problem).
Thank you very much for that! Yes, the real problem with the slow
plotting of Blue Marble very probably concerns the amount of available
memory and not the processor speed. I will try it ASAP. I suppose this
modification is also available only via SVN? Is there a time schedule
for releasing a new distribution version incorporating this recent
changes? I need a setup package that can be downloaded and installed
by computer-unsophisticated Windows users :-(
More soon...
With best wishes,
-- 
Dr. Mauro J. Cavalcanti
Ecoinformatics Studio
P.O. Box 46521, CEP 20551-970
Rio de Janeiro, RJ, BRASIL
E-mail: mau...@gm...
Web: http://studio.infobio.net
Linux Registered User #473524 * Ubuntu User #22717
"Life is complex. It consists of real and imaginary parts."

On Mon, Dec 8, 2008 at 10:56 AM, Ryan May <rm...@gm...> wrote:
> My only other concern is whether this belongs in 0.98.x. This is a behavior
> change from 0.98.3, not necessarily a bug fix. I'll defer to John, et al.
> on whether this should go in 0.98.x or go in a later release.
It's a judgement call, but we have always had new features and minor
API changes in the point releases. Except on the 0.91 maintenance
branch, which is only bugfix, we have continuously added new stuff
on every release. When the breakage is likely to be difficult, or the
new feature really significant, we will push the major version. I
don't think these changes are so significant that they require waiting
until a new major version number, but you may want to consider issuing
a warning in addition to the requisite explanation in the docstring
and CHANGELOG.
JDH

Mauro Cavalcanti wrote:
> Dear ALL,
>
> I am using Basemap version 0.90 with MPL version 0.98.3 under Linux
> Ubuntu, both installed from Andrew Straw repository (I just prefer to
> install from repositories via apt than directly from the sources,
> although in this case I do not have always the latest versions).
>
> I stumbled upon a few issues that I could not solve by looking at the
> available documentation. They are not really critical for me, but I
> would be interested in finding some more about them anyway.
>
> Here they are:
>
> i) When trying some of the projections listed in the Basemap
> documentation, I got a message saying that this or that projection is
> not available, what leads me to think that these projections have only
> been included in later versions of Basemap. However, I could not find
> this in the documentation. This is not critical for my current
> project, as I have no need to offer so many projections, but I would
> like to know.
>
> ii) The plotting of rivers using the drawrivers() method is painfully
> slow, even with resolution set to "crude". The FAQ provided with the
> Basemap source package offers an interesting suggestion for
> potentially solving this, but an exemple would be in order (the FAQ
> mentions an "ireland.py" example which I have however not found in the
> examples directory).
>
> iii) In the same context as the above, I would be interested in
> knowing if there is some way to speed up the plotting of a map with
> the Blue Marble image superimposed on it.\
> 
Mauro: I just added a 'scale' keyword to bluemarble that downsamples 
the image to speed things up. scale=1.0 (the default) gives the same 
answer as before (the full resolution image). scale=0.5 downsamples the 
image to half the original size, speeding things up considerably (and 
using a lot less memory, which I think is the real problem).
-Jeff
> In general, I would like to suggest that the Basemap FAQ could be
> expanded and made available in the website, along with, perhaps, some
> the examples provided in the Basemap distribution package (in the same
> way the examples are provided in the MPL page).
>
> Well, hope this helps!
>
> Have a nice weekend.
>
> With best regards,
>
> 
-- 
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

John Hunter wrote:
> On Mon, Dec 8, 2008 at 9:10 AM, Fago, Matt - AES <Mat...@it...> wrote:
>> So what is to be done here? It seems to me that at least the factor of two should be
>> fixed for one-sided PSDs, and the 1/fs normalization difference with Matlab
>> documented. Ideally, I'd think this normalization would be on by default, with the
>> option to turn it off.
>>
>> Are you planning to submit a patch, or shall I look into it? If I submit a patch, does matplotlib
>> require a copyright assignment?
> 
> If you submit a patch, we assume you are agreeing to the mpl licensing
> terms, so you do not need to explicitly do anything. As for the
> decision about how the defaults and kwargs should behave, I'll defer
> to you and Ryan.
I'm fine with coding up a patch to handle the scaling. Since it seems Matlab 
compatibility is the key desire here, making the Matlab scaling the default is 
fine by me. I, myself, am more than capable of turning on a keyword argument. I 
see two independent changes here:
1) Scaling by factor of two for one-sided psd -- Doesn't need a keyword argument, 
just makes sense.
2) Scaling by 1/Fs. I'm still not sure why Matlab is doing this, other than 
perhaps to make the PSD dB/Hz instead of dB/(rad/s). Needs a new keyword 
argument. Suggestions on what exactly this should be called? scale_by_freq?
My only other concern is whether this belongs in 0.98.x. This is a behavior 
change from 0.98.3, not necessarily a bug fix. I'll defer to John, et al. on 
whether this should go in 0.98.x or go in a later release.
Ryan
-- 
Ryan May
Graduate Research Assistant
School of Meteorology
University of Oklahoma

On Mon, Dec 8, 2008 at 9:10 AM, Fago, Matt - AES <Mat...@it...> wrote:
>
> So what is to be done here? It seems to me that at least the factor of two should be
> fixed for one-sided PSDs, and the 1/fs normalization difference with Matlab
> documented. Ideally, I'd think this normalization would be on by default, with the
> option to turn it off.
>
> Are you planning to submit a patch, or shall I look into it? If I submit a patch, does matplotlib
> require a copyright assignment?
If you submit a patch, we assume you are agreeing to the mpl licensing
terms, so you do not need to explicitly do anything. As for the
decision about how the defaults and kwargs should behave, I'll defer
to you and Ryan.
JDH