matplotlib-users Mailing List for matplotlib

Hmmm, it isn't the same, but I wonder if it is related to the oddities
around https://github.com/matplotlib/matplotlib/pull/2925 . What I mean is
that in that case, none of our unit tests caught the problem, even though
it should have. In your case, your unit tests is catching a similar
problem, but there is no obvious reason why there should be a problem.
Now, technically speaking, in your case, there is a bug in the unittest
package (values() is an iterator in py3k rather than a list), but I
wouldn't know why that dictionary would change in the first place.
On Mon, Apr 14, 2014 at 2:54 PM, John Evans <joh...@gm...>wrote:
> Helllo, I'm seeing a strange issue when running unittests on python3.3 and
> python3.4 that somehow involves matplotlib. My code has a somewhat
> complicated setup, but I think I've boiled the issue down to the following
> reproduction steps
>
>
> import unittest
> import warnings
>
> import matplotlib.pyplot
>
> class TestMe(unittest.TestCase):
> def test_warn(self):
> with self.assertWarns(UserWarning):
> warnings.warn("a warning", UserWarning)
>
> if __name__ == "__main__":
> unittest.main()
>
>
> It looks like it should pass, but it errors as follows
>
> E
> ======================================================================
> ERROR: test_warn (__main__.TestMe)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
> File "/homes/5/jevans/Downloads/testit.py", line 8, in test_warn
> with self.assertWarns(UserWarning):
> File
> "/space/getafix/1/users/jevans/anaconda/envs/py3k/lib/python3.3/unittest/case.py",
> line 177, in __enter__
> for v in sys.modules.values():
> RuntimeError: dictionary changed size during iteration
>
> ----------------------------------------------------------------------
> Ran 1 test in 0.002
>
>
> If the matplotlib import is changed to just
>
> import matlotlib
>
> or if the matplotlib import is commented out altogether, it then passes.
> I'm seeing the behavior on Anaconda with python 3.3 and matplotlib 1.3.1
> on both mac and linux, but also with MacPorts with pythons 3.3 and 3.4,
> matplotlib 1.3.1. All seems fine with a Fedora 20 laptop with python 3.3
> and also matplotlib 1.3.1.
>
>
> --
> John Evans
>
>
> ------------------------------------------------------------------------------
> Learn Graph Databases - Download FREE O'Reilly Book
> "Graph Databases" is the definitive new guide to graph databases and their
> applications. Written by three acclaimed leaders in the field,
> this first edition is now available. Download your free book today!
> http://p.sf.net/sfu/NeoTech
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>
>

Dear Phil,
Thank you. This solves my problem. So the title of my mail is wrong, the
behaviour is reasonable but I am using wrong coordinates.
And also thanks to Jeff.
Cheers,
Chao
On Wed, Apr 16, 2014 at 10:14 AM, Phil Elson <pel...@gm...> wrote:
> Hi Chao,
>
> The warning you are getting:
>
> WARNING: x coordinate not monotonically increasing - contour plot
> may not be what you expect. If it looks odd, your can either
> adjust the map projection region to be consistent with your data, or
> (if your data is on a global lat/lon grid) use the shiftgrid
> function to adjust the data to be consistent with the map projection
> region (see examples/contour_demo.py).
>
> Is important here. It looks like the x coordinate is not in appropriate
> longitudes. Printing the first 5 and last 5 longitudes gives us our first
> clue:
>
> First 5: [-180. -178.99720764 -177.99443054 -176.99163818 -175.98886108]
> Last 5 : [ 175.98886108 176.9916687 177.9944458 178.9972229 180.00003052]
>
> Notice that the last longitude wraps around beyond 180. So if we were to
> clip these numbers to -180 and +180 we will see that the warning disappears
> and the contour is correct. This can be achieved with:
>
> lon = np.clip(lon, -180, 180)
>
> Alternatively, we can just construct the latitudes and longitudes
> ourselves directly with:
>
> lon, lat = np.meshgrid(np.linspace(-180, 180, 360), np.linspace(-90, 90,
> 180))
>
> Incidentally, I tried these numbers with cartopy which has been designed
> to handle dateline wrapping automatically, and the contour worked with the
> unmodified longitudes (http://nbviewer.ipython.org/gist/pelson/10830039).
>
> ---------------------------
>
> @JeffWhitaker - This looks like a bug with float tolerances in the
> makegrid function. It currently does:
>
> def makegrid(self,nx,ny,returnxy=False):
> dx = (self.urcrnrx-self.llcrnrx)/(nx-1)
> dy = (self.urcrnry-self.llcrnry)/(ny-1)
>
> But might be better if it did:
>
> def makegrid(self,nx,ny,returnxy=False):
> x = np.linspace(self.llcrnrx, self.urcrnrx, nx)
> y = np.linspace(self.llcrnry, self.urcrnry, ny)
>
> To avoid the multiplicative floating point drift that is currently being
> seen.
>
> HTH,
>
> Phil
>
-- 
please visit:
http://www.globalcarbonatlas.org/
***********************************************************************************
Chao YUE
Laboratoire des Sciences du Climat et de l'Environnement (LSCE-IPSL)
UMR 1572 CEA-CNRS-UVSQ
Batiment 712 - Pe 119
91191 GIF Sur YVETTE Cedex
Tel: (33) 01 69 08 29 02; Fax:01.69.08.77.16
************************************************************************************

Hi Chao,
The warning you are getting:
WARNING: x coordinate not monotonically increasing - contour plot
may not be what you expect. If it looks odd, your can either
adjust the map projection region to be consistent with your data, or
(if your data is on a global lat/lon grid) use the shiftgrid
function to adjust the data to be consistent with the map projection
region (see examples/contour_demo.py).
Is important here. It looks like the x coordinate is not in appropriate
longitudes. Printing the first 5 and last 5 longitudes gives us our first
clue:
First 5: [-180. -178.99720764 -177.99443054 -176.99163818 -175.98886108]
Last 5 : [ 175.98886108 176.9916687 177.9944458 178.9972229 180.00003052]
Notice that the last longitude wraps around beyond 180. So if we were to
clip these numbers to -180 and +180 we will see that the warning disappears
and the contour is correct. This can be achieved with:
lon = np.clip(lon, -180, 180)
Alternatively, we can just construct the latitudes and longitudes ourselves
directly with:
lon, lat = np.meshgrid(np.linspace(-180, 180, 360), np.linspace(-90, 90,
180))
Incidentally, I tried these numbers with cartopy which has been designed to
handle dateline wrapping automatically, and the contour worked with the
unmodified longitudes (http://nbviewer.ipython.org/gist/pelson/10830039).
---------------------------
@JeffWhitaker - This looks like a bug with float tolerances in the makegrid
function. It currently does:
 def makegrid(self,nx,ny,returnxy=False):
 dx = (self.urcrnrx-self.llcrnrx)/(nx-1)
 dy = (self.urcrnry-self.llcrnry)/(ny-1)
But might be better if it did:
 def makegrid(self,nx,ny,returnxy=False):
 x = np.linspace(self.llcrnrx, self.urcrnrx, nx)
 y = np.linspace(self.llcrnry, self.urcrnry, ny)
To avoid the multiplicative floating point drift that is currently being
seen.
HTH,
Phil

Cool notebook. I took the liberty of giving it a go with cartopy, and you
can see the results here http://nbviewer.ipython.org/gist/pelson/10822698
I'd agree that the issue you linked to does look very similar to the issue
you are seeing, so I think this is very likely a bug.
Cheers,
On 16 April 2014 00:54, Scott Henderson <st...@co...> wrote:
> Hello,
>
> I’m trying to make a plot data on a map with the ‘cyl’ projection with a
> shifted centerline (lon_0=180), but I receive an error when shiftdata() is
> called. Since the plot works when lon_0=0, this seems to be a bug.
>
> I’ve posted the code, error, and plots here:
>
> http://nbviewer.ipython.org/gist/anonymous/cbfe6d0f66ff3a8186c8/shiftdata_issue.ipynb
>
> It might be related to this issue:
> https://github.com/matplotlib/basemap/issues/126
>
> Any insight would be appreciated!
> Scott
>
> ------------------------------------------------------------------------------
> Learn Graph Databases - Download FREE O'Reilly Book
> "Graph Databases" is the definitive new guide to graph databases and their
> applications. Written by three acclaimed leaders in the field,
> this first edition is now available. Download your free book today!
> http://p.sf.net/sfu/NeoTech
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>