Re: [Matplotlib-users] map projections

[Jeff W. <js...@fa...>]

On Tue, 13 Jul 2004, John Hunter wrote:

John:

>
> OK, made some headway here.
>
> When profiling for performace, I often turn off the GUI so my numbers
> aren't influenced by the user interface
>
> > /usr/local/lib/python/profile.py plotmap.py -dAgg > prof.out
>
> Here are the profile results sorted by cumulative time for your script
> - only the most expensive functions are shown
>
>   7.590  proj.py:43(__call__)
>   7.080  os.py:611(popen2)
>   7.050  popen2.py:31(__init__)
>   7.050  popen2.py:143(popen2)
>   2.610  matlab.py:1305(savefig)
>
> On my system, over half the time is spent running your popen process.
> Can't help you with this one :-).  To get better numbers and focus on
> matplotlib, I cached the xy points that the popen process is
> generating to a pickled list.

OK - looks like I have to work on the proj module. I probably should make 
it a C extension that calls the proj library directly instead of 
communicating with the proj application via pipes.

>
> You're using pcolor to generate the colormap.  pcolor uses
> collections, which is already pretty fast (agg implements collection
> drawing in extension code, no python loops needed to make a pcolor).
> However, you can get better performace still from imshow with the data
> limits set.  This will approximately double the performance of the
> image part of the map.  I had to make some changes to matplotlib (see
> below) because image origin wasn't playing nicely with image extent -
> this also fixes Andrew's bug.
>
> For the line part of the map, I extended the
> matplotlib.collections.LineCollection class to handle a sequence of
> lines, where each line is defined by a list of tuples (x0,y0), (x1,
> y1), ...  Thus all of your lines are handled by a single object,
> rather than having 1800+ separate line objects created in plot.
> Again, no python loops required.
>
> In the current form, the code takes about 1.15s to run on my machine
> and is about 30x faster than the original code you posted which
> includes the data loading part.  Nonetheless, the matplotlib part is
> much faster too, as you'll see when you interact with the data.
>
> I'm including a link to a matplotlib snapshot below which includes the
> required changes.  As a bonus, you can try out the new navigation
> toolbar (in progress, only works w/ gtk and gtkagg).  It includes a
> view limits stack, hand pan, and zoom to rectangle.  Much nicer for
> map navigation.  And with the changes, you can actually interact with
> your data with reasonable performance.  I need to add some more
> features to the toolbar, but give it a test drive and let me know if
> you have suggestions.  Set 'toolbar: toolbar2' in matplotlibrc

Excellent!  I'll do some more work on it and repost a (hopefully) faster 
version.

Thanks for the quick response.

-Jeff


-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/CDC  R/CDC1        Email  : Jef...@no...
325 Broadway                Web    : www.cdc.noaa.gov/~jsw
Boulder, CO, USA 80303-3328 Office : Skaggs Research Cntr 1D-124