matplotlib-users — Discussion related to using matplotlib

[Matplotlib-users] Does matplotlib support "Chinese Characters"?

[zhihua o. <zx...@ya...>]

Hi everyone,
  I just found matplotlib and like high quality charts
very much. I am wondering if matplotlib support
Chinese Characters? When the pie Chart function will
be added into matplotlib? 

Thanks

Ouyang


		
__________________________________ 
Do you Yahoo!? 
Send holiday email and support a worthy cause. Do good. 
http://celebrity.mail.yahoo.com

RE: [Matplotlib-users] how to make matplotlib faster?

[Perry G. <pe...@st...>]

John Hunter wrote:

> Plots of this size should be extremely fast - you should be able to
> plot arrays 10 times this big with good performance.  From your
> description "It does first draw a default plot ..and then overplot on
> it for each subplot." it sounds like you may have interactive mode
> turned on.  This would kill your performance in a case like this,
> because the entire figure would be redrawn with the update of every
> single plotting command.  See
> http://matplotlib.sourceforge.net/interactive.html and
> http://matplotlib.sourceforge.net/faq.html#SHOW .
> 
> To definitively determine what mode you are in, run your script with 
> 
>  > python simple_plot.py --verbose-helpful
> 
> and verify that 'interactive is False'.  Fernando Perez's ipython has
> support for running scripts from the interactive shell, turning off
> interactive mode for the duration of the run, and then restoring it.
> 
I wondered the same thing and mentioned that to him privately
(in effect you are doing n*(n-1)/2 plots instead of n).

But that made me wonder whether or not there was a need for some
sort of switch that delayed any update for just this case where
one is looping over many plots (say you wrote a ploting function
that did this that you want to run in interactive mode, and you
wanted to use basic plotting functions like plot). Is there a  
simple mechanism to turn off interactive mode temporarily within
the function and restore it at the end? If not, could it be added?
(akin to the hold() function)

Perry

[Matplotlib-users] Re: Contour plots

[Chris B. <Chr...@no...>]

John Hunter wrote:
>>>>>>"Chris" == Chris Barker <Chr...@no...> writes:
>     Chris> completely irregular? or only orthogonal structured
>     Chris> grids. From your description, it sounds like the
>     Chris> later. Could it take an unstructured set of (x,y,z) points
>     Chris> and contour the z values?
> 
> The latter, I believe.

yup. from the below referenced link:

"""
General purpose contour tracer for quadrilateral meshes.
"""

So it won't handle arbitrary unstructured points, but it's nice none the 
less. With an interpolator to a rectangular grid, you could use it for 
any array of points, I think someone posted an example of this on the 
matplotlib list.

-Chris



-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Peter G. <pgr...@ge...>]

John Hunter wrote:

>>>>>> "Peter" == Peter Groszkowski <pgr...@ge...> writes:
>>>>>
>
> Peter> I use Hardy's multiquadric interpolation to to do the math,
> Peter> then use imshow (or pcolor) to make a surface map. I only
> Peter> have data for the 120 points (where the circle are - those
> Peter> are actuators), and interpolate the rest.
>
> Peter> If people are interested, I can clean up the code a little
> Peter> and post it.
>
> This sounds pretty close to matlab's griddata function. 

Yup. That's the functionality I needed. As it says in MATLAB docstrig 
below, GRIDDATA uses Delaunay triangulation however.

> It would be
> very nice to have this in matplotlib.mlab, perhaps as a wrapper to
> some core scipy functionality, which could be conditionally imported.
> What requirements does your code have -- pure python, extension code,
> scipy, numarray?
>
The interpolating is done all in python with the use of Numeric (this is 
what I have been using, and what my matplotlib installation uses - maybe 
will upgrade to numarray one of these days). Performance wise, It's not 
very practical for a very large number of points N as it has to solve a 
NxN system (my N=120 and takes ~2.3seconds on a P4 3.2Ghz with 2GB ram - 
cant remember how long MATLAB's griddata took). Maybe numarray would be 
faster?!

The drawing of the mirror, actuators, etc is done using matplotlibs 
imshow(), plot() and fill() - all very straight forward.

I will post the code in the next few days when I have a minute to clean 
it up a litte.

Cheers,

--
Peter Groszkowski         Gemini Observatory
Tel: +1 808 9742509       670 N. A'ohoku Place
Fax: +1 808 9359235       Hilo, Hawai'i 96720, USA





> Here is the matlab docstring, FYI
>
> GRIDDATA Data gridding and surface fitting.
> ZI = GRIDDATA(X,Y,Z,XI,YI) fits a surface of the form Z = F(X,Y)
> to the data in the (usually) nonuniformly-spaced vectors (X,Y,Z)
> GRIDDATA interpolates this surface at the points specified by
> (XI,YI) to produce ZI. The surface always goes through the data
> points. XI and YI are usually a uniform grid (as produced by
> MESHGRID) and is where GRIDDATA gets its name.
>
> XI can be a row vector, in which case it specifies a matrix with
> constant columns. Similarly, YI can be a column vector and it
> specifies a matrix with constant rows.
>
> [XI,YI,ZI] = GRIDDATA(X,Y,Z,XI,YI) also returns the XI and YI
> formed this way (the results of [XI,YI] = MESHGRID(XI,YI)).
>
> [...] = GRIDDATA(...,'method') where 'method' is one of
> 'linear' - Triangle-based linear interpolation (default).
> 'cubic' - Triangle-based cubic interpolation.
> 'nearest' - Nearest neighbor interpolation.
> 'v4' - MATLAB 4 griddata method.
> defines the type of surface fit to the data. The 'cubic' and 'v4'
> methods produce smooth surfaces while 'linear' and 'nearest' have
> discontinuities in the first and zero-th derivative respectively. All
> the methods except 'v4' are based on a Delaunay triangulation of the
> data.
>
> See also GRIDDATA3, GRIDDATAN, DELAUNAY, INTERP2, MESHGRID.
>
>
>
>

Re: [Matplotlib-users] how to make matplotlib faster?

[John H. <jdh...@ac...>]

>>>>> "Matt" == Matt Newville <new...@ca...> writes:

    Matt> For me, this block (run twice for a plot()) typically takes
    Matt> at least 50% of the plot time.  Commenting out the
    Matt> tick.draw(renderer) and the following two 'extent' lines
    Matt> roughly doubles the drawing rate (though no grid or ticks
    Matt> are shown). I was surprised by this, but have not tracked it
    Matt> down much beyond this.  I'm not using mathtext in the labels
    Matt> and had only standard numerical Tick labels in this example.

    Matt> I don't know if this is applicable to the slowness of the
    Matt> contour plots or error bars or if collections would help
    Matt> here.  But it doesn't seem like tick drawing should be the
    Matt> bottleneck. Anyway, this seems like a simple place to test
    Matt> in other situations, and may be a good place to look for
    Matt> possible optimizations.

This is a known bottleneck.  Text layout is non-trivial in matplotlib.
Put it this way: you don't get multiline text with arbitrary rotation,
font properties, horizontal, vertical, and multiline alignment for
free.  Take a look at matplotlib.text.Text._get_layout.

I do cache the layout information because I've seen this performance
hit on animated demos before.  But if your text properties are
changing, caching doesn't help.  The cache key is returned by
Text.get_prop_tup.

It is probably worthwhile to run your animation through the profiler
so we can get a better idea of where exactly the problems are.  I
think the matrix multiplication that _get_layout uses for rotations is
slow.  It would be possible to special case the most common case
(rotation angle = 0) for some speedups, but the code is already fairly
hairy so I've been resisting special casing it.

FYI, the wxagg backend uses string methods to transfer the agg image
to the canvas.  tk and gtk use extension code.  fltk uses a python
buffer object.  I investigated the latter for wxagg but couldn't make
it work.  You may want to look into FigureCanvasWxAgg.draw to see if
you can do this image transfer faster, possibly adding some extension
code.  If you do go the extension code route, I suggest you try/except
the import on your extension code and fall back on the string method
already in place.

Oh, I added "newville" to the list of CVS developers.  Everyone,
welcome Matt, the new wx maintainer!

JDH

RE: [Matplotlib-users] Re: Contour plots

[Perry G. <pe...@st...>]

John Hunter Wrote:
> >>>>> "Chris" == Chris Barker <Chr...@no...> writes:
> 
>     Chris> Perry Greenfield wrote:
> 
>     >>> Actually, I believe that the low level contour engine we are
>     >>> using supports this. It takes 2-d arrays for both x and y that
>     >>> represent the x and y coordinates of the array being contoured
>     >>> and generates plotting points based on those x and y
>     >>> arrays. These arrays allow for irregular grids.
> 
>     Chris> completely irregular? or only orthogonal structured
>     Chris> grids. From your description, it sounds like the
>     Chris> later. Could it take an unstructured set of (x,y,z) points
>     Chris> and contour the z values?
> 
> The latter, I believe.  The contouring routine was implemented by
> Nadia Dencheva (CCd on this mail) and is based on a gist routine.  You

Yes, the latter. For an unstructured set some other approach would be
needed. Sorry if I misunderstood. I thought what was being discussed
was irregular spacings rather than irregular organization.

Perry

Re: [Matplotlib-users] how to make matplotlib faster?

[John H. <jdh...@ac...>]

>>>>> "Eric" == Eric Emsellem <ems...@ob...> writes:

    Eric> Hi, 1. slow plot 2. cursor issue 3. key press event !
    Eric> ------- 1/ Here is the piece of code which is quite slow I
    Eric> think. Compared to pgplot this is a factor of more than
    Eric> 10. It does first draw a default plot (0,1 ?)  and then
    Eric> overplot on it for each subplot.

    Eric> for this particular case I have 10 subplots. The slices are
    Eric> made of about 10-20 points each only (stored in a 3D array
    Eric> which is 48x5x20 points).  I hope this answers the
    Eric> question. Sorry for the ''specifics''.

Plots of this size should be extremely fast - you should be able to
plot arrays 10 times this big with good performance.  From your
description "It does first draw a default plot ..and then overplot on
it for each subplot." it sounds like you may have interactive mode
turned on.  This would kill your performance in a case like this,
because the entire figure would be redrawn with the update of every
single plotting command.  See
http://matplotlib.sourceforge.net/interactive.html and
http://matplotlib.sourceforge.net/faq.html#SHOW .

To definitively determine what mode you are in, run your script with 

 > python simple_plot.py --verbose-helpful

and verify that 'interactive is False'.  Fernando Perez's ipython has
support for running scripts from the interactive shell, turning off
interactive mode for the duration of the run, and then restoring it.

If this doesn't solve your problem please

  - post your entire script

  - report the output of verbose-helpful (requires matplotlib 0.64)

  - what is your platform, machine specs, etc?

  - how are you running the script (IDE, from the command prompt, etc)

JDH

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[John H. <jdh...@ac...>]

>>>>> "Peter" == Peter Groszkowski <pgr...@ge...> writes:

    Peter> I use Hardy's multiquadric interpolation to to do the math,
    Peter> then use imshow (or pcolor) to make a surface map.  I only
    Peter> have data for the 120 points (where the circle are - those
    Peter> are actuators), and interpolate the rest.

    Peter> If people are interested, I can clean up the code a little
    Peter> and post it.

This sounds pretty close to matlab's griddata function.  It would be
very nice to have this in matplotlib.mlab, perhaps as a wrapper to
some core scipy functionality, which could be conditionally imported.
What requirements does your code have -- pure python, extension code,
scipy, numarray?

Here is the matlab docstring, FYI

 GRIDDATA Data gridding and surface fitting.
    ZI = GRIDDATA(X,Y,Z,XI,YI) fits a surface of the form Z = F(X,Y)
    to the data in the (usually) nonuniformly-spaced vectors (X,Y,Z)
    GRIDDATA interpolates this surface at the points specified by
    (XI,YI) to produce ZI.  The surface always goes through the data
    points.  XI and YI are usually a uniform grid (as produced by
    MESHGRID) and is where GRIDDATA gets its name.
  
    XI can be a row vector, in which case it specifies a matrix with
    constant columns. Similarly, YI can be a column vector and it
    specifies a matrix with constant rows.
  
    [XI,YI,ZI] = GRIDDATA(X,Y,Z,XI,YI) also returns the XI and YI
    formed this way (the results of [XI,YI] = MESHGRID(XI,YI)).
  
    [...] = GRIDDATA(...,'method') where 'method' is one of
        'linear'    - Triangle-based linear interpolation (default).
        'cubic'     - Triangle-based cubic interpolation.
        'nearest'   - Nearest neighbor interpolation.
        'v4'        - MATLAB 4 griddata method.
    defines the type of surface fit to the data. The 'cubic' and 'v4'
    methods produce smooth surfaces while 'linear' and 'nearest' have
    discontinuities in the first and zero-th derivative respectively.  All
    the methods except 'v4' are based on a Delaunay triangulation of the
    data.
  
    See also GRIDDATA3, GRIDDATAN, DELAUNAY, INTERP2, MESHGRID.

[Matplotlib-users] Re: Contour plots

[John H. <jdh...@ac...>]

>>>>> "Chris" == Chris Barker <Chr...@no...> writes:

    Chris> Perry Greenfield wrote:

    >>> Actually, I believe that the low level contour engine we are
    >>> using supports this. It takes 2-d arrays for both x and y that
    >>> represent the x and y coordinates of the array being contoured
    >>> and generates plotting points based on those x and y
    >>> arrays. These arrays allow for irregular grids.

    Chris> completely irregular? or only orthogonal structured
    Chris> grids. From your description, it sounds like the
    Chris> later. Could it take an unstructured set of (x,y,z) points
    Chris> and contour the z values?

The latter, I believe.  The contouring routine was implemented by
Nadia Dencheva (CCd on this mail) and is based on a gist routine.  You
can read more about the core routine at
http://scipy.net/cgi-bin/viewcvsx.cgi/*checkout*/scipy1/xplt/src/gist/gcntr.c?rev=HEAD&content-type=text/plain.

The contour routines have been checked into CVS.  A simple example can
be found in examples/contour_demo.py in CVS.  We have some rudimentary
support for labeling (auto-legend and/or brute-force use of the text
command).  It would be nice to develop a point and click labeling
widget and/or an auto-labeling routine.  Contributors of course always
welcome!

JDH

Re: [Matplotlib-users] plot scatter priority

[John H. <jdh...@ac...>]

>>>>> "Humufr" == Humufr  <hu...@ya...> writes:

    Humufr>        Hello, I have a problem to plot some data.

    Humufr> I use "plot" to plot some data and "scatter" for other. I
    Humufr> obtain a plot whith the point trace with "scatter" are
    Humufr> behind the points from "plot". I tried to change the order
    Humufr> in the script but that change nothing.  Do you know how to
    Humufr> do this? (I want use scatter because I want have a
    Humufr> specific size for this points)

I just added the long awaited zorder attribute to the Artist base
class in CVS, originally discussed here
http://sourceforge.net/mailarchive/message.php?msg_id=9363527.

There is a new example examples/zorder_demo.py that shows you how to
set the zorder.  I'll include it below.

It will be in the next release, probably early next week

    #!/usr/bin/env python
    """
    The default drawing order for axes is patches, lines, text.  This
    order is determined by the zorder attribute.  The following default
    are set

    Artist                      Z-order
    Patch / PatchCollection      1
    Line2D / LineCollection      2
    Text                         3

    You can change the order for individual artists by setting the zorder.

    In the fist subplot below, the lines are drawn above the patch
    collection from the scatter, which is the default.

    In the subplot below, the order is reversed
    """

    from pylab import *
    x = rand(20); y = rand(20)

    subplot(211)
    plot(x, y, 'r', lw=3)
    scatter(x,y,s=120)

    subplot(212)
    plot(x, y, 'r', zorder=1, lw=3)
    scatter(x,y,s=120, zorder=2)

    show()

Re: [Matplotlib-users] colorbar() not working with scatter

[John H. <jdh...@ac...>]

>>>>> "Arnold" == Arnold Moene <arn...@wu...> writes:

    Arnold> Dear all, At the moment I'm heavily using the scatter plot
    Arnold> (great!). But if I want to add a color bar (with the
    Arnold> command colorbar(), directly following the call to
    Arnold> scatter) to explain the meaning of the colors of the
    Arnold> patches, matplotlib (0.64) refuses to make a colorbar with
    Arnold> the following message:

I Arnold, thanks for alerting me to this problem.  This was a bug, and
is now fixed in CVS.  I should be releasing a new version of
matplotlib next week which has these changes plus more.

JDH

[Matplotlib-users] using show() multiple times: no longer

[Alan G I. <ai...@am...>]

I have a few scripts that call show() multiple times.
(Why?  Well I had pedagogical reasons: I like one
graph to come up at a time in class, in the order
I wish, with no other "distractions".)

These scripts work as I wish if run with execfile() from the
(default Python) interpreter prompt.  (I.e., the first
figure is displayed, and when I close it the next is
displayed, etc.) If executed from the cmd line however they
fail:
    Fatal Python error: PyEval_RestoreThread: NULL tstate
    abnormal program termination
Can I add something to the scripts so that they behave
as I wish if executed from a cmd line?  (Am I wrong to
believe this used to work until recently; say last
version number or so?)

Aside from my wishes, should the script fail in this
fashion (rather than being more gracefully rejected)?
I realize we have been warned against using show()
multiple times ...

Somewhat related:
can I control the order in which figures are displayed
when the show() command is given, or will the highest
numbered figure always display on top?

Thank you,
Alan Isaac
Win 2000, Python 2.3.3, MatPlotLib 0.63

Re: [Matplotlib-users] Date demos from examples fail to run

[John H. <jdh...@ac...>]

>>>>> "Delbert" == Delbert D Franz <iq...@so...> writes:

    Delbert> After downloading and installing these two packages all
    Delbert> but date_demo_rrule.py completed properly.  The error in
    Delbert> this case was an unknown name "rand".  A check of the
    Delbert> Python Library reference stated it was obsolete.  I
    Delbert> replaced it with random.randrange but got another error,
    Delbert> an assertion error apparently on the y value.  Being
    Delbert> somewhat new to Python and even newer to matplotlib I
    Delbert> gave up on that demo.

    Delbert> Perhaps someone else can test date_demo_rrule.py and see
    Delbert> what happens.  It is always a good thing when demos in
    Delbert> fact run!

True!  But all of these demos do run for me.  I suggest you flush your
existing matplotlib by removing site-packages/matplotlib and your
"build" directory and reinstall from the official source at
http://sourceforge.net/project/showfiles.php?group_id=80706&package_id=82474&release_id=281218.
Please follow the instructions at
http://matplotlib.sourceforge.net/installing.html, eg make sure you
have numeric or numarray installed when you compile matplotlib.

Let us know if you have more troubles, and please include a full
traceback from one of the date demos and run it with

> python date_demo1.py --verbose-helpful

and report the output.


    Delbert> I am also testing under MS Windows and the dateutils and
    Delbert> pytz files came with that install but none of the example
    Delbert> files came.  Not sure why they are not included in the
    Delbert> *.exe installer.

It's a distutils thing.  Suggestions here welcome.

JDH

Re: [Matplotlib-users] how to make matplotlib faster?

[John H. <jdh...@ac...>]

>>>>> "Chris" == Chris Barker <Chr...@no...> writes:

    Chris> One thing that could help here is if all the drawing
    Chris> commands were "vectorized". This would mean that rather
    Chris> than asking the back-end to draw one primitive at a time, a
    Chris> whole set could be passed in. This would allow the back end
    Chris> to possibly optimize the drawing of the set. An example is
    Chris> wxPython's DC.DrawXXXList() methods. These methods take a
    Chris> python sequence of drawing primitives, and loop through
    Chris> that sequence in C++ code. It makes a huge difference when
    Chris> drawing simple objects, like points or line segments.

    Chris> I haven't looked closely at the matplotlib code to see if
    Chris> this can be done, but it could make a difference.


This is basically what collections already do --
http://matplotlib.sourceforge.net/matplotlib.collections.html.
Typically when we find an area of code where a bunch of primitives are
being drawn independently, we refactor them as a collection.  There is
a default draw implementation in python that the backends can override
in extension code (as agg does).  Even if you just use the default
python drawing implementation, eg
backend_bases.RendererBase.draw_line_collection, the result is much
faster that instantiating a large number of independent objects.

Every property in a collection is a sequence (might be just length
one) and the drawing code iterates over the collection and gets the
value of a property for the i-th element of the collection as

  thisprop = prop[i % len(props)]

So if you have a len(1) list, every element in the collection shares
the prop, if you have a len(props) list, every element has a unique
property and < len(props) the props cycle.

Actually, a dictionary mapping element index to property value, which
has a default value, would be more flexible, and the performance hit
might not be bad.  Might be worth refactoring.

Actually, the code to plot line markers could be sped up by using
collections to draw line markers.  Currently we're using plain old
python loops for this.  

JDH

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Chris B. <Chr...@no...>]

Perry Greenfield wrote:

>> Actually, I believe that the low level contour engine we are using
>> supports this. It takes 2-d arrays for both x and y that represent
>> the x and y coordinates of the array being contoured and generates
>> plotting points based on those x and y arrays. These arrays allow
>> for irregular grids.

completely irregular? or only orthogonal structured grids. From your 
description, it sounds like the later. Could it take an unstructured set 
of (x,y,z) points and contour the z values?

-Chris



-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Chris B. <Chr...@no...>]

Peter Groszkowski wrote:
> I use Hardy's multiquadric interpolation to to do the math,  then use 
> imshow (or pcolor) to make a surface map.  I only have data for the 120 
> points (where the circle are - those are actuators), and interpolate the 
> rest.
> 
> If people are interested, I can clean up the code a little and post it.

Please do.



-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Peter G. <pgr...@ge...>]

Perry Greenfield wrote:

>
> On Dec 9, 2004, at 2:12 PM, Chris Barker wrote:
>
>> LUK ShunTim wrote:
>>
>>> As it's being implemented, here is a little wish. I'd like to see 
>>> the capability of contouring on an arbitrary grid. That is, 
>>> matplotlab would be able to plot the contours of a function f(x_i, 
>>> y_i) given on an arbitrary set of points (x_i, y_i), not necessarily 
>>> set out on a regular grid.
>>
>>
>> This would be nice, but it's a bit of a project. One way to do it 
>> would be to Delaunay triangulate the points, then you can compute the 
>> contours from the triangular grid. Delaunay triangulation is not 
>> trivial, and you really want to use an efficient scheme to do it. One 
>> possibility is:
>>
>> http://www-2.cs.cmu.edu/~quake/triangle.html
>>
>> It is very robust and fast, and can be compiled as a library. I've 
>> been planning for ages to write a Python wrapper for it, but haven't 
>> gotten to it yet.
>>
>> If someone works on this, I'd like to help.
>>
>> -Chris
>>
>
> Actually, I believe that the low level contour engine we are using
> supports this. It takes 2-d arrays for both x and y that represent
> the x and y coordinates of the array being contoured and generates
> plotting points based on those x and y arrays. These arrays allow
> for irregular grids. At the moment, the routine generates uniform
> x and y grids as arguments to pass along, but it could be generalized
> to take these as extra arguments without much trouble.


I use Hardy's multiquadric interpolation to to do the math,  then use 
imshow (or pcolor) to make a surface map.  I only have data for the 120 
points (where the circle are - those are actuators), and interpolate the 
rest.

If people are interested, I can clean up the code a little and post it.

Re: [Matplotlib-users] how to make matplotlib faster?

[Matt N. <new...@ca...>]

Hi,

I'm doing relatively simple line plots the WXAgg backend, but I
also find matplotlib to be somewhat slower than I'd hope for.

On a WindowsXP box (P4 1.7GHz, 512Mb RAM), in a wx event loop
issuing a plot() as fast as I can go, I get about 1 plot every
0.25 to 0.30 sec.  This is just barely fast enough for my needs.  
If I could reliably go at 10 plots/sec, that would be great.

It turns out that the dynamic_demo_wx.py example does go much
faster, but it does not actually re-do a plot(). Instead it just
changes the subplots line data.  That's interesting, but I need
the view to be adjusted as well, as the scale will change with
time for my data.  So far, I'm just re-issuing plot(), but I'd be
willing to do something slightly fancier.

Anyway, that led me to try to track down where the slowness in
plot() was coming from. Using nothing more sophisticated than
print statements, I believe the performance bottleneck is in
axis.py in Axis.draw(), in this block:

        for tick, loc, label in zip(majorTicks, majorLocs, majorLabels):
            if not interval.contains(loc): continue
            seen[loc] = 1
            tick.update_position(loc)
            tick.set_label1(label)
            tick.set_label2(label)            
            tick.draw(renderer)
            extent = tick.label1.get_window_extent(renderer) 
            ticklabelBoxes.append(extent)

For me, this block (run twice for a plot()) typically takes at
least 50% of the plot time.  Commenting out the
tick.draw(renderer) and the following two 'extent' lines roughly
doubles the drawing rate (though no grid or ticks are shown). I
was surprised by this, but have not tracked it down much beyond
this.  I'm not using mathtext in the labels and had only standard
numerical Tick labels in this example.

I don't know if this is applicable to the slowness of the contour
plots or error bars or if collections would help here.  But it
doesn't seem like tick drawing should be the bottleneck. Anyway,
this seems like a simple place to test in other situations, and
may be a good place to look for possible optimizations.

Thanks, 

--Matt

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Perry G. <pe...@st...>]

On Dec 9, 2004, at 2:33 PM, Perry Greenfield wrote:
>
> Actually, I believe that the low level contour engine we are using
> supports this. It takes 2-d arrays for both x and y that represent
> the x and y coordinates of the array being contoured and generates
> plotting points based on those x and y arrays. These arrays allow
> for irregular grids. At the moment, the routine generates uniform
> x and y grids as arguments to pass along, but it could be generalized
> to take these as extra arguments without much trouble.
>
> Let me know if I misunderstand what you are trying to do.
>
> Perry
>
Correction. It already supports that feature now (but it isn't
checked in yet).

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Perry G. <pe...@st...>]

On Dec 9, 2004, at 2:12 PM, Chris Barker wrote:

> LUK ShunTim wrote:
>> As it's being implemented, here is a little wish. I'd like to see the 
>> capability of contouring on an arbitrary grid. That is, matplotlab 
>> would be able to plot the contours of a function f(x_i, y_i) given on 
>> an arbitrary set of points (x_i, y_i), not necessarily set out on a 
>> regular grid.
>
> This would be nice, but it's a bit of a project. One way to do it 
> would be to Delaunay triangulate the points, then you can compute the 
> contours from the triangular grid. Delaunay triangulation is not 
> trivial, and you really want to use an efficient scheme to do it. One 
> possibility is:
>
> http://www-2.cs.cmu.edu/~quake/triangle.html
>
> It is very robust and fast, and can be compiled as a library. I've 
> been planning for ages to write a Python wrapper for it, but haven't 
> gotten to it yet.
>
> If someone works on this, I'd like to help.
>
> -Chris
>

Actually, I believe that the low level contour engine we are using
supports this. It takes 2-d arrays for both x and y that represent
the x and y coordinates of the array being contoured and generates
plotting points based on those x and y arrays. These arrays allow
for irregular grids. At the moment, the routine generates uniform
x and y grids as arguments to pass along, but it could be generalized
to take these as extra arguments without much trouble.

Let me know if I misunderstand what you are trying to do.

Perry

Re: [Matplotlib-users] how to make matplotlib faster?

[Chris B. <Chr...@no...>]

Arnd Baecker wrote:
> P.S.: I agree on the speed issues. Unfortunately
>       most of the newer python graphics packages tend to be
>       slower than older packages.

I think this has two reasons:

1) They are written more in Python, rather than wrapping an existing 
library written in C or whatever.

2) They often are back-end independent. This introduces an extra layer 
at every drawing command, and makes it difficult to take advantage of 
possible optimizations available for a given back end, like Arnd has 
done for his stuff.

One thing that could help here is if all the drawing commands were 
"vectorized". This would mean that rather than asking the back-end to 
draw one primitive at a time, a whole set could be passed in. This would 
allow the back end to possibly optimize the drawing of the set. An 
example is wxPython's DC.DrawXXXList() methods. These methods take a 
python sequence of drawing primitives, and loop through that sequence in 
C++ code. It makes a huge difference when drawing simple objects, like 
points or line segments.

I haven't looked closely at the matplotlib code to see if this can be 
done, but it could make a difference.

-Chris


-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Re: Contour plots [Was: Re: [Matplotlib-users] how to make matplotlib faster?]

[Chris B. <Chr...@no...>]

LUK ShunTim wrote:
> As it's being implemented, here is a little wish. I'd like to see the 
> capability of contouring on an arbitrary grid. That is, matplotlab would 
> be able to plot the contours of a function f(x_i, y_i) given on an 
> arbitrary set of points (x_i, y_i), not necessarily set out on a regular 
> grid.

This would be nice, but it's a bit of a project. One way to do it would 
be to Delaunay triangulate the points, then you can compute the contours 
from the triangular grid. Delaunay triangulation is not trivial, and you 
really want to use an efficient scheme to do it. One possibility is:

http://www-2.cs.cmu.edu/~quake/triangle.html

It is very robust and fast, and can be compiled as a library. I've been 
planning for ages to write a Python wrapper for it, but haven't gotten 
to it yet.

If someone works on this, I'd like to help.

-Chris

-- 
Christopher Barker, Ph.D.
Oceanographer
                                     		
NOAA/OR&R/HAZMAT         (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chr...@no...

Re: [Matplotlib-users] how to make matplotlib faster?

[Norbert N. <Nor...@gm...>]

Am Donnerstag, 9. Dezember 2004 17:38 schrieb John Hunter:
> Note it
> would be possible to define setitem, getitem, and possibly setslice,
> getslice and iter for collections to make them behave more like lists
> of objects, which would be nice if we (you) want to make this change.

Would that be possible at all? Do individual items in a collection have an 
identity at all that could be exposed in Python? Do they have individual 
properties?

Note, that I probably won't have the time to look into this matter myself. 
Maybe one day, but certainly not in the near future.

Ciao,
Nobbi

-- 
_________________________________________Norbert Nemec
         Bernhardstr. 2 ... D-93053 Regensburg
     Tel: 0941 - 2009638 ... Mobil: 0179 - 7475199
           eMail: <No...@Ne...>

Re: [Matplotlib-users] how to make matplotlib faster?

[Eric E. <ems...@ob...>]

Hi,

1. slow plot
2. cursor issue
3. key press event !
-------
1/ 

Here is the piece of code which is quite slow I think. Compared to 
pgplot this
is a factor of more than 10. It does first draw a default plot (0,1 ?) 
and then
overplot on it for each subplot.

for this particular case I have 10 subplots. The slices are made of 
about 10-20
points each only (stored in a 3D array which is  48x5x20 points).
I hope this answers the question. Sorry for the ''specifics''.

        for i in arange(ncoef) :
           subplot(nrow, 2, i+1)
           if i > 1 :
              temparray = self.Vcoef[indgal][i][:minind] / 
self.Vcoef[indgal][1][:minind]
              plot(self.Vrad[indgal][:minind], temparray, 'b-o', ms=4)
           else :
              plot(self.Vrad[indgal][:minind], 
self.Vcoef[indgal][i][:minind], 'b-o', ms=4)
           ylabel('$C_{%d'%i+'}$')

        for i in arange(ncoef) :
           j = i + ncoef
           subplot(nrow, 2, j+1)
           plot(self.Vrad[indgal][:minind], 
self.Vphi[indgal][i][:minind], 'b-o', ms=4)
           ylabel('$\phi_{%d'%i+'}$')

2/ the cursor issue and how to interact with it:
yes indeed the solution I took is close to what is shown in some examples.
I used a new class which I then use later on in interactive mode or not.
Below is a simple/shortened example of the structure I create by just 
transferring
the data (x,y,key, etc) to the cursor structure. I can then use : toto = 
cursor()
to have it working. (I in fact define several different cursor_* classes 
for different purposes).
So sorry if my mail sounded like ''I have found a new way...''.

class cursor :
   def __init__(self):
      print 'Class initialized'
      self.figure = Figure()
      self.canvas = get_current_fig_manager().canvas
      self.canvas.mpl_connect('button_press_event', self.on_click)
      self.x = 0
      self.y = 0
      self.button = 1

   def on_click(self, event):
      self.x = event.x
      self.y = event.y
      self.xd = event.xdata
      self.yd = event.ydata
      self.button = event.button
      ...

3/ key press event

>Currently key_press_event is not implemented (though mouse move and
>motion to capture and report key presses as event.key).  We added this
>because this is how we do the event handling across backends for the
>toolbar.  There are also some fixes in CVS to make disconnects work
>properly, eg in the tk backend.
>
>It would be fairly straightforward to add a key_press_event under the
>same framework.
>
>  
>
That WOULD be great since this is exactly what I needed. For example being
able to type ''h'' to make an horizontal cut of my image at the location 
corresponding
to the mouse position.... I do that easily with ppgplot for example.

-- 
===============================================================
Observatoire de Lyon                 ems...@ob...
9 av. Charles-Andre                      tel: +33 4 78 86 83 84
69561 Saint-Genis Laval Cedex            fax: +33 4 78 86 83 86
France               http://www-obs.univ-lyon1.fr/eric.emsellem
===============================================================

Re: [Matplotlib-users] how to make matplotlib faster?

[John H. <jdh...@ac...>]

>>>>> "Eric" == Eric Emsellem <ems...@ob...> writes:

    Eric> P.S.: by the way I solved the cursor problem I posted (and
    Eric> got no answer) by defining a new cursor class (something
    Eric> already hinted by many on the web), if anyone is
    Eric> interested..

Been out of town at meetings for the last week...

Have you seen
http://matplotlib.sourceforge.net/examples/coords_demo.py, which shows
how to connect to mouse motion and click?  The interface is being
streamlined in 0.65.  Ie in CVS, one simply needs to do

  def on_click(event): pass
  connect('button_press_event', on_click)

Currently key_press_event is not implemented (though mouse move and
motion to capture and report key presses as event.key).  We added this
because this is how we do the event handling across backends for the
toolbar.  There are also some fixes in CVS to make disconnects work
properly, eg in the tk backend.

It would be fairly straightforward to add a key_press_event under the
same framework.

For "cursoring", see the *cursor*.py examples in the examples subdir
of the matplotlib src distro.

But please also post your solution which may be useful....

JDH