problem with basemap/quiver with latest versions of basemap&mpl

_P_R.M · August 12, 2009, 1:44am

Hi,
I just updated to the latest versions of mpl & basemap.
Im getting strange output when using the quiver function from basemap (see
attached image).

I ran the quiver_demo scripts for both basemap & also mpl, and the output
looked normal.

Im using pygrads, a python interface to GrADS, in order to generate my
X,Y,U,V input arrays for quiver.
I can provide these arrays if needed...

The data imported using pygrads previously worked fine (prior to updating to
the new versions).

Before I start diving into the pygrads code in order to troubleshoot, I
wanted to first check to see if this issue is a possible quiver bug in the
new version(s)...

Or, were there any significant changes to the mpl/basemap code with respect
to how quiver works???

Here's the basic method Im using...
U = <my_pygrads_imported_U>
V = <my_pygrads_imported_V>
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))
#x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(), V.ravel())
cs2 = m.quiver(X,Y,U,V,*kwargs)

again, this *used* to work fine, prior to the updates...

Please help,
Thanks,
P.Romero

_Jeff_Whitaker1 · August 12, 2009, 2:25am

P.R. wrote:

Hi,
I just updated to the latest versions of mpl & basemap.
Im getting strange output when using the quiver function from basemap (see
attached image).

I ran the quiver_demo scripts for both basemap & also mpl, and the output
looked normal.

Im using pygrads, a python interface to GrADS, in order to generate my
X,Y,U,V input arrays for quiver.
I can provide these arrays if needed...

The data imported using pygrads previously worked fine (prior to updating to
the new versions).

Before I start diving into the pygrads code in order to troubleshoot, I
wanted to first check to see if this issue is a possible quiver bug in the
new version(s)...

Or, were there any significant changes to the mpl/basemap code with respect
to how quiver works???

Here's the basic method Im using...
U = <my_pygrads_imported_U>
V = <my_pygrads_imported_V>
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))
#x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(), V.ravel())
cs2 = m.quiver(X,Y,U,V,*kwargs)

again, this *used* to work fine, prior to the updates...

Please help,
Thanks,
P.Romero

P.R.: I can't see anything wrong - you'd better post your data so we can try it. You might also try to upgrade to numpy 1.3.0, if you haven't already, I know there were some masked array changes between 1.2.1 and 1.3.

-Jeff

_P_R.M · August 12, 2009, 3:31am

Ok, I upgraded to numpy 1.30 and the problem still showed up...
Im still getting those strange diagonal line defects.
(looks as if some of the edge arrow polygons are being clipped/drawn/closed
improperly & thus 'dragged' across the plot area???)

I attached my x,y,u,v arrays...

Here's my basemap setup...
m=bm(projection='cyl',llcrnrlat=20,urcrnrlat=30,llcrnrlon=-100,urcrnrlon=-80
,resolution='h')

though it might not help much, here's the exact export from pygrads
(ga.expr() exports from grads into a customized masked array in python)
'dirpw' is the grads variable being imported into mpl.

U = ga.expr('(-1*sin(dirpw*3.1414/180))')
V = ga.expr('(-1*cos(dirpw*3.1414/180))')
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))

I then was also doing this to convert to simple ndarrays
(a while back, I was recommended this method to solve a problem I was having
with quiver performing really slowly when using those grads masked arrays as
input)

x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(), V.ravel())

Let me know if you need anything else...

Thanks,
P.R.

xy_uv.npz (27.6 KB)

···

-----Original Message-----
From: Jeff Whitaker [mailto:jswhit@…146…]
Sent: 2009-08-11 9:26 PM
To: P.R.
Cc: matplotlib-users@lists.sourceforge.net
Subject: Re: [Matplotlib-users] problem with basemap/quiver with latest
versions of basemap&mpl

P.R. wrote:

Hi,
I just updated to the latest versions of mpl & basemap.
Im getting strange output when using the quiver function from basemap (see
attached image).

I ran the quiver_demo scripts for both basemap & also mpl, and the output
looked normal.

Im using pygrads, a python interface to GrADS, in order to generate my
X,Y,U,V input arrays for quiver.
I can provide these arrays if needed...

The data imported using pygrads previously worked fine (prior to updating

to

the new versions).

Before I start diving into the pygrads code in order to troubleshoot, I
wanted to first check to see if this issue is a possible quiver bug in the
new version(s)...

Or, were there any significant changes to the mpl/basemap code with

respect

to how quiver works???

Here's the basic method Im using...
U = <my_pygrads_imported_U>
V = <my_pygrads_imported_V>
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))
#x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(),

V.ravel())

cs2 = m.quiver(X,Y,U,V,*kwargs)

again, this *used* to work fine, prior to the updates...

Please help,
Thanks,
P.Romero

P.R.: I can't see anything wrong - you'd better post your data so we
can try it. You might also try to upgrade to numpy 1.3.0, if you
haven't already, I know there were some masked array changes between
1.2.1 and 1.3.

-Jeff

_Jeff_Whitaker1 · August 12, 2009, 4:05am

P.R. wrote:

Ok, I upgraded to numpy 1.30 and the problem still showed up...
Im still getting those strange diagonal line defects.
(looks as if some of the edge arrow polygons are being clipped/drawn/closed
improperly & thus 'dragged' across the plot area???)

I attached my x,y,u,v arrays...

P.R. The file appears to be corrupted:

In [4]: from numpy import load

In [5]: load('xy_uv.npz').files
Out[5]: ['y', 'x', 'u', 'v']

In [6]: load('xy_uv.npz')['u']

···

---------------------------------------------------------------------------
BadZipfile Traceback (most recent call last)

/Users/jsw/python/basemap/<ipython console> in <module>()

/opt/local/lib/python2.5/site-packages/numpy/lib/io.pyc in __getitem__(self, key)
    110 key += '.npy'
    111 if member:
--> 112 bytes = self.zip.read(key)
    113 if bytes.startswith(format.MAGIC_PREFIX):
    114 value = cStringIO.StringIO(bytes)

/opt/local/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/zipfile.pyc in read(self, name)
    477 fheader = self.fp.read(30)
    478 if fheader[0:4] != stringFileHeader:
--> 479 raise BadZipfile, "Bad magic number for file header"
    480
    481 fheader = struct.unpack(structFileHeader, fheader)

BadZipfile: Bad magic number for file header

-Jeff

Here's my basemap setup...
m=bm(projection='cyl',llcrnrlat=20,urcrnrlat=30,llcrnrlon=-100,urcrnrlon=-80
,resolution='h')

though it might not help much, here's the exact export from pygrads
(ga.expr() exports from grads into a customized masked array in python)
'dirpw' is the grads variable being imported into mpl.

U = ga.expr('(-1*sin(dirpw*3.1414/180))')
V = ga.expr('(-1*cos(dirpw*3.1414/180))')
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))

I then was also doing this to convert to simple ndarrays
(a while back, I was recommended this method to solve a problem I was having
with quiver performing really slowly when using those grads masked arrays as
input)

x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(), V.ravel())

Let me know if you need anything else...

Thanks,
P.R.

-----Original Message-----
From: Jeff Whitaker [mailto:jswhit@…146…] Sent: 2009-08-11 9:26 PM
To: P.R.
Cc: matplotlib-users@lists.sourceforge.net
Subject: Re: [Matplotlib-users] problem with basemap/quiver with latest
versions of basemap&mpl

P.R. wrote:


Hi,
I just updated to the latest versions of mpl & basemap.
Im getting strange output when using the quiver function from basemap (see
attached image).

I ran the quiver_demo scripts for both basemap & also mpl, and the output
looked normal.

Im using pygrads, a python interface to GrADS, in order to generate my
X,Y,U,V input arrays for quiver.
I can provide these arrays if needed...

The data imported using pygrads previously worked fine (prior to updating


to


the new versions).

Before I start diving into the pygrads code in order to troubleshoot, I
wanted to first check to see if this issue is a possible quiver bug in the
new version(s)...

Or, were there any significant changes to the mpl/basemap code with


respect


to how quiver works???

Here's the basic method Im using...
U = <my_pygrads_imported_U>
V = <my_pygrads_imported_V>
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))
#x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(),


V.ravel())


cs2 = m.quiver(X,Y,U,V,*kwargs)

again, this *used* to work fine, prior to the updates...

Please help,
Thanks,
P.Romero

P.R.: I can't see anything wrong - you'd better post your data so we can try it. You might also try to upgrade to numpy 1.3.0, if you haven't already, I know there were some masked array changes between 1.2.1 and 1.3.

-Jeff

_P_R.M · August 12, 2009, 4:37am

Jeff,
Nevermind...

I figured out the problem...
It appears that the grads mpl interface (pygrads) wasn't creating an
appropriate mask for the data.
All of the '-9.99e+8' values that correspond to 'undefined' points weren't
being masked out, and were thus causing quiver to freak out & try to draw
some extreme vector arrows...

I fixed the mask, and it works fine now...
Sorry for the bother, but thanks anyway for the help...

Regards,
P.R.

···

-----Original Message-----
From: Jeff Whitaker [mailto:jswhit@…146…]
Sent: 2009-08-11 11:05 PM
To: P.R.
Cc: matplotlib-users@lists.sourceforge.net
Subject: Re: [Matplotlib-users] problem with basemap/quiver with latest
versions of basemap&mpl

P.R. wrote:

Ok, I upgraded to numpy 1.30 and the problem still showed up...
Im still getting those strange diagonal line defects.
(looks as if some of the edge arrow polygons are being

clipped/drawn/closed

improperly & thus 'dragged' across the plot area???)

I attached my x,y,u,v arrays...

P.R. The file appears to be corrupted:

In [4]: from numpy import load

In [5]: load('xy_uv.npz').files
Out[5]: ['y', 'x', 'u', 'v']

In [6]: load('xy_uv.npz')['u']
---------------------------------------------------------------------------
BadZipfile Traceback (most recent call last)

/Users/jsw/python/basemap/<ipython console> in <module>()

/opt/local/lib/python2.5/site-packages/numpy/lib/io.pyc in
__getitem__(self, key)
    110 key += '.npy'
    111 if member:
--> 112 bytes = self.zip.read(key)
    113 if bytes.startswith(format.MAGIC_PREFIX):
    114 value = cStringIO.StringIO(bytes)

/opt/local/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/zi
pfile.pyc
in read(self, name)
    477 fheader = self.fp.read(30)
    478 if fheader[0:4] != stringFileHeader:
--> 479 raise BadZipfile, "Bad magic number for file header"
    480
    481 fheader = struct.unpack(structFileHeader, fheader)

BadZipfile: Bad magic number for file header

-Jeff

Here's my basemap setup...

m=bm(projection='cyl',llcrnrlat=20,urcrnrlat=30,llcrnrlon=-100,urcrnrlon=-80

,resolution='h')

though it might not help much, here's the exact export from pygrads
(ga.expr() exports from grads into a customized masked array in python)
'dirpw' is the grads variable being imported into mpl.

U = ga.expr('(-1*sin(dirpw*3.1414/180))')
V = ga.expr('(-1*cos(dirpw*3.1414/180))')
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))

I then was also doing this to convert to simple ndarrays
(a while back, I was recommended this method to solve a problem I was

having

with quiver performing really slowly when using those grads masked arrays

as

input)

x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(), V.ravel())

Let me know if you need anything else...

Thanks,
P.R.

-----Original Message-----
From: Jeff Whitaker [mailto:jswhit@…146…]
Sent: 2009-08-11 9:26 PM
To: P.R.
Cc: matplotlib-users@lists.sourceforge.net
Subject: Re: [Matplotlib-users] problem with basemap/quiver with latest
versions of basemap&mpl

P.R. wrote:

Hi,
I just updated to the latest versions of mpl & basemap.
Im getting strange output when using the quiver function from basemap

(see

attached image).

I ran the quiver_demo scripts for both basemap & also mpl, and the output
looked normal.

Im using pygrads, a python interface to GrADS, in order to generate my
X,Y,U,V input arrays for quiver.
I can provide these arrays if needed...

The data imported using pygrads previously worked fine (prior to updating

to

the new versions).

Before I start diving into the pygrads code in order to troubleshoot, I
wanted to first check to see if this issue is a possible quiver bug in

the

new version(s)...

Or, were there any significant changes to the mpl/basemap code with


respect


to how quiver works???

Here's the basic method Im using...
U = <my_pygrads_imported_U>
V = <my_pygrads_imported_V>
X,Y=m(*np.meshgrid(U.grid.lon,U.grid.lat))
#x,y,u,v = delete_masked_points(X.ravel(), Y.ravel(), U.ravel(),


V.ravel())


cs2 = m.quiver(X,Y,U,V,*kwargs)

again, this *used* to work fine, prior to the updates...

Please help,
Thanks,
P.Romero

P.R.: I can't see anything wrong - you'd better post your data so we
can try it. You might also try to upgrade to numpy 1.3.0, if you
haven't already, I know there were some masked array changes between
1.2.1 and 1.3.

-Jeff