Matplotlib: An apparent error

[Owen Densmore]

I work with a group, many of whom use the numpy, scipy, matplotlib stack, while I much prefer using Sage.  So I modified Sage to let me use the matplotlib GUI which is not include in the standard Sage release.

To do so, I followed the instructions for installing the matplotlib backend

    http://wiki.sagemath.org/sage_matlab

.. which for me required only setting the two env variables and running

    sage -i matplotlib

To test this out, I actually installed the numpy, scipy, matplotlib as well, and in both, ran this test from a Machine Learning book, using "ipython" and "sage" command line .. i.e. no arguments like -pylab etc:

import numpy as np

import matplotlib.pyplot as plt

import os

os.chdir(os.path.expanduser("~/src/ML/2 Linear"))

pima = np.loadtxt('./pima-indians-diabetes.data',delimiter=',')

indices0 = np.where(pima[:,8]==0)

indices1 = np.where(pima[:,8]==1)

plt.plot(pima[indices0,0],pima[indices0,1],'go');

plt.plot(pima[indices1,0],pima[indices1,1],'rx');

plt.show()

The results look alike, but the axes appear to lack the lowest decimal value.  So for 10, 11, 12 ... I see 1, 1, 1!

The dataset is on this public site: http://archive.ics.uci.edu/ml/datasets/Pima+Indians+Diabetes

Here are the plots, the first from the scipy install, the second from sage.  I'd like to know if this is a mistake on my part or a bug that needs to be reported.

My environment is 

sage -v -> Detected SAGE64 flag, Building Sage on OS X in 64-bit mode, Sage Version 4.7, Release Date: 2011-05-23

        -- Owen

[William Stein]

On Sunday, August 21, 2011, Owen Densmore <ow...@backspaces.net> wrote:
> I work with a group, many of whom use the numpy, scipy, matplotlib stack, while I much prefer using Sage.  So I modified Sage to let me use the matplotlib GUI which is not include in the standard Sage release.
> To do so, I followed the instructions for installing the matplotlib backend
>     http://wiki.sagemath.org/sage_matlab
> .. which for me required only setting the two env variables and running
>     sage -i matplotlib
> To test this out, I actually installed the numpy, scipy, matplotlib as well, and in both, ran this test from a Machine Learning book, using "ipython" and "sage" command line .. i.e. no arguments like -pylab etc:
>

Obvious question: are you using exactly the same versions of numpy, etc., in both of your tests?

> import numpy as np
> import matplotlib.pyplot as plt
> import os
> os.chdir(os.path.expanduser("~/src/ML/2 Linear"))
> pima = np.loadtxt('./pima-indians-diabetes.data',delimiter=',')
> indices0 = np.where(pima[:,8]==0)
> indices1 = np.where(pima[:,8]==1)
> plt.plot(pima[indices0,0],pima[indices0,1],'go');
> plt.plot(pima[indices1,0],pima[indices1,1],'rx');
> plt.show()
>
> The results look alike, but the axes appear to lack the lowest decimal value.  So for 10, 11, 12 ... I see 1, 1, 1!
> The dataset is on this public site: http://archive.ics.uci.edu/ml/datasets/Pima+Indians+Diabetes
> Here are the plots, the first from the scipy install, the second from sage.  I'd like to know if this is a mistake on my part or a bug that needs to be reported.
> My environment is 
> sage -v -> Detected SAGE64 flag, Building Sage on OS X in 64-bit mode, Sage Version 4.7, Release Date: 2011-05-23
>         -- Owen
>
>
>
>
>
>
>

> --
> To post to this group, send email to sage-s...@googlegroups.com
> To unsubscribe from this group, send email to sage-support...@googlegroups.com <sage-support%2Bunsu...@googlegroups.com>
> For more options, visit this group at http://groups.google.com/group/sage-support
> URL: http://www.sagemath.org
>

--
William Stein
Professor of Mathematics
University of Washington
http://wstein.org

[Rajeev Singh]

--
To post to this group, send email to sage-s...@googlegroups.com
To unsubscribe from this group, send email to sage-support...@googlegroups.com

For more options, visit this group at http://groups.google.com/group/sage-support
URL: http://www.sagemath.org

also the y-axis has 5, 10, 15, 20 instead of 50, 100, 150, 200. just pointing out since u didn't mention it.

Rajeev

[Jason Grout]

On 8/21/11 6:54 PM, William Stein wrote:
>
> Obvious question: are you using exactly the same versions of numpy,
> etc., in both of your tests?

Exactly. We are currently behind the latest release of matplotlib. Can
you do this in both systems?

import matplotlib
print matplotlib.__version__

and report your results? Please do the same for numpy as well.

Thanks,

Jason

[Owen Densmore]

@william: Obvious question: are you using exactly the same versions of numpy, etc., in both of your tests?

No they are very unlikely to be the same versions.  I used the last pre-Lion superpack_10.6_2011.07.10.sh

Chris Fonnesbeck: http://stronginference.com/scipy-superpack/

@jason: Exactly.  We are currently behind the latest release of matplotlib.  Can you do this in both systems?

Yes: Here are the versions:

sage: import matplotlib

sage: print matplotlib.__version__

1.0.1

sage: import numpy

sage: print numpy.__version__

1.5.1

In [1]: import matplotlib

In [2]: print matplotlib.__version__

1.1.0

In [3]: import numpy

In [4]: print numpy.__version__

2.0.0.dev-b5cdaee

All: This morning I entered a bug report (the google docs form) with more information.  I didn't see it in Trac but I don't know how these two are coordinated.  I created a small directory to repeat the bugs: http://backspaces.net/temp/plotbug.

I don't think this is a version problem due to a prior similar report: 

https://groups.google.com/d/topic/sage-support/t5CnEiHTaNI/discussion

.. and the successful plot in the sage notebook.

Pix:

http://backspaces.net/temp/plotbug/SagePlt.png

        Snapshot of truncated axes

http://backspaces.net/temp/plotbug/SagenbPlt.png

        Snapshot of correct sagenb example

http://backspaces.net/temp/plotbug/SciPyPlt.png

        Snapshot of non-sage, scipy stack example

Here's the report, in case the google docs had a problem.  Let me know if you would like me to download a more recent version of Sage and try again.

////

'Sage Version 4.7, Release Date: 2011-05-23' 

....

Sage Matplotlib Plots Truncate x,y axes values

....

Running the following, starting sage in the directory where the pima-indians-diabetes.data resides, creates a plot with the x & y axes numerical labels truncated.

sage: import numpy as np

sage: import matplotlib.pyplot as plt

sage: pima = np.loadtxt('pima-indians-diabetes.data',delimiter=',')

sage: indices0 = np.where(pima[:,8]==0)

sage: indices1 = np.where(pima[:,8]==1)

sage: plt.plot(pima[indices0,0],pima[indices0,1],'go');

sage: plt.plot(pima[indices1,0],pima[indices1,1],'rx');

sage: plt.show()

See: http://backspaces.net/temp/plotbug/SagePlt.png 

for the truncated labels

Note that this works correctly in sage notebooks (using savefig)

    import numpy as np

    import matplotlib.pyplot as plt

    pima = np.loadtxt(DATA+'pima-indians-diabetes.data',delimiter=',')

    indices0 = np.where(pima[:,8]==0)

    indices1 = np.where(pima[:,8]==1)

    plt.plot(pima[indices0,0],pima[indices0,1],'go');

    plt.plot(pima[indices1,0],pima[indices1,1],'rx');

    plt.savefig('pima.png')

See: http://backspaces.net/temp/plotbug/SagenbPlt.png

And also works correctly in a non-sage scipy,numpy,matplotlib installation:

    http://backspaces.net/temp/plotbug/SciPyPlt.png

http://backspaces.net/temp/plotbug/ contains:

SagePlt.png

        Snapshot of truncated axes

SagenbPlt.png

        Snapshot of correct sagenb example

SciPyPlt.png

        Snapshot of non-sage, scipy stack example

pima-indians-diabetes.data

        Dataset

Note: these two prior posts in the support group discuss the problem:

https://groups.google.com/d/topic/sage-support/t5CnEiHTaNI/discussion

https://groups.google.com/d/topic/sage-support/yMnxwVh1w-s/discussion

....

The intention is to create scipy, numpy, matplotlib, ipython subset within sage so that I can work with fellow workers, having compatibility between sage & the scipy stack.

Then I'll lure them to sage!  :)

....

No backtrace, only truncated labels.

Possibly this is a back-end problem of the ipython version of sage and there is a work around?

////