quiver problems when angles is an array of values rather than 'uv' or 'xy'

[mlandriau]

I use the version of matplotlib that is included in the following distribution:
Python 3.5.2 |Anaconda 4.1.1 (x86_64)| (default, Jul 2 2016, 17:52:12)
IPython 4.2.0 -- An enhanced Interactive Python.

quiver produces the following warnings when called with angles=array :

/Users/mlandriau/anaconda/lib/python3.5/site-packages/matplotlib/quiver.py:609: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
if self.angles == 'xy' and self.scale_units == 'xy':
/Users/mlandriau/anaconda/lib/python3.5/site-packages/matplotlib/quiver.py:614: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
elif self.angles == 'xy' or self.scale_units == 'xy':
/Users/mlandriau/anaconda/lib/python3.5/site-packages/matplotlib/quiver.py:647: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
if self.angles == 'xy':
/Users/mlandriau/anaconda/lib/python3.5/site-packages/matplotlib/quiver.py:649: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
elif self.angles == 'uv':

It seems to produce the right answer, except quiverkey no longer works properly (the arrow is a filled half-circle instead of a line).

[tacaswell]

What does matplotlib.__version__ return?

[tacaswell]

and can you provide a minimal working example?

[mlandriau]

The version of matplotlib is 1.5.1 and here is the code I used. The commented out line works fine, except it does not have the desired angles.

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import rc
import matplotlib.ticker as ticker

def fmt(x, pos):
    a, b = '{:.2e}'.format(x).split('e')
    b = int(b)
    return r'${} \times 10^{{{}}}$'.format(a, b)

rc('font',**{'family':'sans-serif','sans-serif':['Helvetica']})
rc('text', usetex=True)

norm = 1.0
nmap = 1
n = int(ngrid)
for i in range(nmap):
    infile = 'map' + str(i+1) + '.dat'
    outfile = 'map' + str(i+1) + '.eps'
    data = np.genfromtxt(infile, names=['t', 'q', 'u'])
    t = np.reshape( data['t']  * norm, (n,n) )
    q = np.reshape( data['q'] * norm, (n,n) )
    u = np.reshape( data['u'] * norm, (n,n) )
    psi = (0.5 * np.arctan2(u,q)) * 180.0 / np.pi
    p = np.sqrt(q*q + u*u)
    pkey = 5.0 * np.mean
    pkeystr = fmt(pkey,0)
    fig = plt.figure()
    plt.axis('off')
    fig = plt.imshow(t, cmap='jet')
    plt.colorbar(format=ticker.FuncFormatter(fmt))
    x, y = np.meshgrid( range(n), range(n) )
    skip = ( slice(None, None, 4), slice(None, None, 4) )
    qv = plt.quiver(x[skip], y[skip], q[skip], u[skip], angles=psi[skip], headwidth=1, headlength=0.1, pivot='middle')
    #qv = plt.quiver(x[skip], y[skip], q[skip], u[skip], headwidth=1, headlength=0.1, pivot='middle')
    plt.quiverkey(qv, 0.4, 0.95, pkey, pkeystr, coordinates='figure', labelpos='E')
    plt.savefig(outfile, format='eps')
    plt.show()

[tacaswell]

This is still not minimal example, I meant something more like

import matplotlib.pyplot as plt
import numpy as np

fig, ax = plt.subplots()

X, Y = np.meshgrid(np.arange(15), np.arange(10))
U = V = np.ones_like(X)
phi = (np.random.rand(15, 10) - .5) * 150
ax.quiver(X, Y, U, V, angles=phi)

which is sufficient to show the problem, is understandable, and (most importantly) is copy-paste-runnable.

[tacaswell]

and I have a fix for this, just adding a test.

[tacaswell]

@mlandriau Can you open a new issue for the quiverkey? It is not drawing a half-circle, it is drawing an arrow for each angle

import matplotlib.pyplot as plt
import numpy as np

fig, ax = plt.subplots()

X, Y = np.meshgrid(np.arange(5), np.arange(7))
U = V = np.ones_like(X)

phi = (np.random.rand(5, 7) - .5) * 150
q = ax.quiver(X, Y, U, V*.5, angles=phi, color='r', alpha=1)

qk = ax.quiverkey(q, 0.5, 0.92, 2, r'$2 \frac{m}{s}$',
                      labelpos='W',
                      fontproperties={'weight': 'bold'})

[mlandriau]

Thanks, I've opened a new issue using the test code in your comment with one line added.

[jenshnielsen]

For the record the new issue is #6972 this is closed by #6971