plot_surface and transposed arrays

[micfe03]

plot_surface(X, Y, Z, ..)
if you have non-square matrices and Z is transposed (Z.T), you
would expect an error message. However, X and Y become reshaped, such that

1. the plot is messed up
2. as a side effect X and Y are scrambled after the call

[WeatherGod]

On Wed, Mar 20, 2013 at 7:19 AM, micfe03 notifications@github.com wrote:

plot_surface(X, Y, Z, ..)
if you have non-square matrices and Z is transposed (Z.T), you
would expect an error message. However, X and Y become reshaped, such that

1. the plot is messed up
2. as a side effect X and Y are scrambled after the call

Which version of matplotlib are you using? I believe I fixed this at some
point.

[micfe03]

I installed the version that was most recent in autumn 2012. It says:

In [149]: ? plt
Type: module
String Form:<module 'matplotlib.pyplot' from
'/Library/Python/2.7/site-packages/matplotlib-1.2.x-py2.7-macosx-10.7-intel
.egg/matplotlib/pyplot.pyc'>
File:
/Library/Python/2.7/site-packages/matplotlib-1.2.x-py2.7-macosx-10.7-intel.
egg/matplotlib/pyplot.py

I could not find any old issue on the problem, so I assumed that it was
still present.

On 3/20/13 2:52 PM, "Benjamin Root" notifications@github.com wrote:

On Wed, Mar 20, 2013 at 7:19 AM, micfe03 notifications@github.com wrote:

plot_surface(X, Y, Z, ..)
if you have non-square matrices and Z is transposed (Z.T), you
would expect an error message. However, X and Y become reshaped, such
that

1. the plot is messed up
2. as a side effect X and Y are scrambled after the call

Which version of matplotlib are you using? I believe I fixed this at some
point.
‹
Reply to this email directly or
view it on GitHub
<#1838 (comment)
4>.

[WeatherGod]

On Wed, Mar 20, 2013 at 9:56 AM, micfe03 notifications@github.com wrote:

I installed the version that was most recent in autumn 2012. It says:

In [149]: ? plt
Type: module
String Form:<module 'matplotlib.pyplot' from
'/Library/Python/2.7/site-packages/matplotlib-1.2.x-py2.7-macosx-10.7-intel

.egg/matplotlib/pyplot.pyc'>
File:
/Library/Python/2.7/site-packages/matplotlib-1.2.x-py2.7-macosx-10.7-intel.

egg/matplotlib/pyplot.py

I could not find any old issue on the problem, so I assumed that it was
still present.

/Michael

Oooh, just found the issue. It is an edge-case where one dimension is a
multiple of the other (e.g., 5x10). This, by itself, is bad enough. But
apparently, the underlying call to np.broadcast_arrays() appears to be
modifying the original object's strides, which I think is a numpy bug.

[WeatherGod]

Actually, just realized I screwed up my own example... I can not reproduce your issue. Could you provide a short example to demonstrate it.

[micfe03]

Here you go.

If you want to see the scrambled plot, you have to replace Z with
something deterministic ;-)

On 3/20/13 4:44 PM, "Benjamin Root" notifications@github.com wrote:

Actually, just realized I screwed up my own example... I can not
reproduce your issue. Could you provide a short example to demonstrate it.
‹
Reply to this email directly or
view it on GitHub
<#1838 (comment)
0>.

[WeatherGod]

nothing attached. I think you need to paste the code in the text because github strips out attachments, I think.

[micfe03]

Hi,

I sent this yesterday from my mobile, but I think it could not handle the
reply address construction in your email:

Sorry, did not know that. Here is the code:

import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
​
if name == 'main':
Z = np.random.uniform(0,1,(6,7))
x = np.arange(Z.shape[0])
y = np.arange(Z.shape[1])
X, Y = np.meshgrid(x, y)
print X
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.plot_surface(X, Y, Z)
plt.show()
print X

On 3/20/13 5:13 PM, "Benjamin Root" notifications@github.com wrote:

nothing attached. I think you need to paste the code in the text because
github strips out attachments, I think.
—
Reply to this email directly or
view it on GitHub
<#1838 (comment)
8>.

[WeatherGod]

I was only able to reproduce your problem with matplotlib v1.1.x. With v1.2.x and above, the patch that I mentioned earlier catches these sorts of issues and raises an exception saying that the arrays have incompatible dimensions.

Somehow, you are running a different version of matplotlib than you think. Try putting in a "import matplotlib; print matplotlib.version" line in your code to make certain.

[micfe03]

In [18]: import matplotlib; print matplotlib.version
1.2.x

But it is on Mac OS, maybe that makes a difference?

On 3/21/13 2:19 PM, "Benjamin Root" notifications@github.com wrote:

I was only able to reproduce your problem with matplotlib v1.1.x. With
v1.2.x and above, the patch that I mentioned earlier catches these sorts
of issues and raises an exception saying that the arrays have
incompatible dimensions.
Somehow, you are running a different version of matplotlib than you
think. Try putting in a "import matplotlib; print matplotlib.version"
line in your code to make certain.
‹
Reply to this email directly or
view it on GitHub
<#1838 (comment)
5>.

[WeatherGod]

On Thu, Mar 21, 2013 at 10:48 AM, micfe03 notifications@github.com wrote:

In [18]: import matplotlib; print matplotlib.version
1.2.x

But it is on Mac OS, maybe that makes a difference?

Where did you get your install binary? This isn't an official release
binary because an official release wouldn't have an 'x' in the version
string. I suspect that your binary was built from the development branch
prior to my patch, which is in the 1.2.0 release.

[micfe03]

I have installed via Scipy superpack for Mac os v 10.7.4:

https://github.com/fonnesbeck/ScipySuperpack/tags

On 3/21/13 3:54 PM, "Benjamin Root" notifications@github.com wrote:

On Thu, Mar 21, 2013 at 10:48 AM, micfe03 notifications@github.com
wrote:

In [18]: import matplotlib; print matplotlib.version
1.2.x

But it is on Mac OS, maybe that makes a difference?

Where did you get your install binary? This isn't an official release
binary because an official release wouldn't have an 'x' in the version
string. I suspect that your binary was built from the development branch
prior to my patch, which is in the 1.2.0 release.
‹
Reply to this email directly or
view it on GitHub
<#1838 (comment)
3>.

[WeatherGod]

From the Superpack README:

All builds are based on recent development code from each package, which means though some bugs may be
fixed and features added, they also may be more unstable than the official releases.

Now that we have that mystery solved, I will be closing this bug.

[Tillsten]

This issues is still somewhate there:
matplot is showing garbage instead a error msg if x, y, z are all 1-dim arrays with the same shape.

[WeatherGod]

Gah! didn't think of that edge case... handling this is gonna get
complicated. Could you open up a new issue for this edge case? Complete
with a test-able example?

On Wed, Jun 4, 2014 at 4:36 PM, Till Stensitzki notifications@github.com
wrote:

This issues is still somewhate there:
matplot is showing garbage instead a error msg if x, y, z are all 1-dim
arrays with the same shape.

—
Reply to this email directly or view it on GitHub
#1838 (comment)
.

[Tillsten]

See #3116