nphelper 0.0.5

nphelper - convenient numpy helper functions

This package contains some convenient helper function for numpy. Nothing fancy, but quite useful. It works with python 2.7, 3.4, and 3.5.

Install

Simply install via pip:

pip install nphelper

nphelper only depends on numpy.

Usage / Features

Create block arrays (in matlabs’s [A A; B B] spirit):

>>> from nphelper import block
>>> A = np.array([[1, 2, 3]])
>>> B = np.array([[2, 3, 4]])
>>> C = np.ones((2, 6))
>>> block([[A, B], [C]])
array([[1, 2, 3, 2, 3, 4],
      [[1, 1, 1, 1, 1, 1],
      [[1, 1, 1, 1, 1, 1]])

Compute the cartesian product (similar to itertools.product):

>>> from nphelper import cartesian_product
>>> cartesian_product([[1, 2], [3, 4]])
array([[1, 3],
       [1, 4],
       [2, 3],
       [2, 4]])

Easily multiply many arrays without the cubersome dot syntax. It’s also much faster than dot because it selects the fastest evaluation order. (This is part of numpy 1.10.0.)

>>> from nphelper import multi_dot
>>> A = np.random.random((10000, 100))
>>> B = np.random.random((100, 1000))
>>> C = np.random.random((1000, 5))
>>> D = np.random.random((5, 333))
>>> # Sick of this?
>>> np.dot(np.dot(np.dot(A, B), C), D)  # doctest: +SKIP
>>> # Or this?
>>> A.dot(B).dot(C).dot(D)  # doctest: +SKIP
>>> # Use multi_dot
>>> multi_dot([A, B, C, D])  # doctest: +SKIP

• TODO along, maxalong, minalong, sumalong, meanalong, stdalong, varalong

Dev

Run the tests

Run tox to run the tests for python 2.7, 3.4, and 3.5:

tox

You might have to install addiotional dependencies to run the tests:

- ``py.test``,
- ``nose``,
- ``python2.7-dev``,
- ``python3.4-dev``, and
- ``python3.5-dev``.

Build the Docs

cd doc
make html

You might have to install addiotional dependencies:

pip install sphinx sphinx_rtd_theme