Using tf.transformations quaternion_from_euler with numpy array arguments can silently change the input arguments!

[peci1]

This code in quaternion_from_euler is actually dangerous when called with 1-D 1-element numpy arrays on input:

geometry/tf/src/tf/transformations.py

Lines 1100 to 1128 in 63c3c7b

	def quaternion_from_euler(ai, aj, ak, axes='sxyz'):
	"""Return quaternion from Euler angles and axis sequence.
	ai, aj, ak : Euler's roll, pitch and yaw angles
	axes : One of 24 axis sequences as string or encoded tuple
	>>> q = quaternion_from_euler(1, 2, 3, 'ryxz')
	>>> numpy.allclose(q, [0.310622, -0.718287, 0.444435, 0.435953])
	True
	"""
	try:
	firstaxis, parity, repetition, frame = _AXES2TUPLE[axes.lower()]
	except (AttributeError, KeyError):
	_ = _TUPLE2AXES[axes]
	firstaxis, parity, repetition, frame = axes
	i = firstaxis
	j = _NEXT_AXIS[i+parity]
	k = _NEXT_AXIS[i-parity+1]
	if frame:
	ai, ak = ak, ai
	if parity:
	aj = -aj
	ai /= 2.0
	aj /= 2.0
	ak /= 2.0

E.g.

import numpy as np
from tf.transformations import quaternion_from_euler
angles = np.array([[0], [0], [1.0]])
print(angles[2], type(angles[2]))  # prints [ 1. ], <type 'numpy.ndarray'>
q = quaternion_from_euler(0, 0, angles[2])
print(angles[2])  # prints [ 0.5 ], <type 'numpy.ndarray'>

The problem is that 1-element arrays behave like scalars, so the function works, however, different from scalars, they are passed to functions as reference as opposed to by value.

[salihmarangoz]

Created a PR with passing checks to solve this: #241