rowan

1.3.2 verified Fri May 01 auth: no python

A Python library for performing quaternion operations using NumPy arrays. Provides functions for quaternion creation, normalization, rotation, interpolation, and conversion between quaternions and other rotation representations (e.g., rotation matrices, axis-angle). Current version: 1.3.2, released 2023-04-03. Release cadence is irregular.

pip install rowan

Common errors

error AttributeError: module 'rowan' has no attribute 'from_axis_angle' ↓

cause Older versions of rowan (<1.0?) use different function names, e.g., `rowan.from_axisangle` (no underscore). Check installed version.

fix

Upgrade to latest rowan: pip install --upgrade rowan or use the correct old name: rowan.from_axisangle.

error TypeError: Object arrays are not supported ↓

cause Input numpy array has dtype=object, often because it contains mixed types or is created from a ragged list.

fix

Ensure array has homogeneous numeric dtype, e.g., np.array(..., dtype=np.float64).

Warnings

gotcha rowan expects inputs as numpy arrays, not lists. Passing plain Python lists may cause unexpected errors or silently produce wrong results. ↓

fix Always convert lists to np.array before passing to rowan functions.

gotcha Quaternion conventions: rowan uses the scalar-first convention (q = [w, x, y, z]). Some other libraries (e.g., scipy) use scalar-last. Mixing conventions will cause incorrect rotations. ↓

fix When interfacing with other libraries, convert conventions explicitly: for rowan use [w, x, y, z].

gotcha The function `rowan.from_axis_angle` expects the axis to be normalized. If the axis is not unit length, the quaternion will represent a rotation with incorrect angle. ↓

fix Normalize the axis vector using `axis / np.linalg.norm(axis)` before passing.

Imports

rowan
wrong
```
from rowan import *
```
correct
```
import rowan
```
Importing * pollutes the namespace; prefer explicit functions.

Quickstart

Create a quaternion from axis-angle, rotate a vector, and convert to rotation matrix.

import rowan
import numpy as np

# Create a quaternion from axis-angle (radians)
axis = np.array([0, 0, 1])
angle = np.pi / 2
q = rowan.from_axis_angle(axis, angle)

# Rotate a vector
v = np.array([1, 0, 0])
v_rot = rowan.rotate(q, v)
print(v_rot)  # [0, 1, 0]

# Convert to rotation matrix
R = rowan.to_matrix(q)
print(R)