148 lines
5.4 KiB
Plaintext
148 lines
5.4 KiB
Plaintext
|
Metadata-Version: 1.1
|
||
|
|
Name: np
|
||
|
|
Version: 0.2.0
|
||
|
|
Summary: np = numpy++: numpy with added convenience functionality
|
||
|
|
Home-page: https://github.com/k7hoven/np
|
||
|
|
Author: Koos Zevenhoven
|
||
|
|
Author-email: koos.zevenhoven@aalto.fi
|
||
|
|
License: BSD
|
||
|
|
Description: np -- create numpy arrays as ``np[1,3,5]``, and more
|
||
|
|
====================================================
|
||
|
|
|
||
|
|
``np`` = ``numpy`` + handy tools
|
||
|
|
|
||
|
|
For the numerical Python package ``numpy`` itself, see http://www.numpy.org/.
|
||
|
|
|
||
|
|
The idea of ``np`` is to provide a way of creating numpy arrays with a compact syntax and without an explicit function call. Making the module name ``np`` subscriptable, while still keeping it essentially an alias for numpy, does this in a clean way.
|
||
|
|
|
||
|
|
Any feedback is very welcome: `koos.zevenhoven@aalto.fi`.
|
||
|
|
|
||
|
|
Getting Started
|
||
|
|
===============
|
||
|
|
|
||
|
|
Requirements
|
||
|
|
------------
|
||
|
|
|
||
|
|
* Works best with Python 3.5+ (Tested also with 3.4 and 2.7)
|
||
|
|
* numpy (you should install this using your python package manager like `conda` or `pip`)
|
||
|
|
|
||
|
|
Installation
|
||
|
|
------------
|
||
|
|
|
||
|
|
np can be installed with pip:
|
||
|
|
|
||
|
|
.. code-block:: bash
|
||
|
|
|
||
|
|
$ pip install np
|
||
|
|
|
||
|
|
|
||
|
|
or directly from the source code:
|
||
|
|
|
||
|
|
.. code-block:: bash
|
||
|
|
|
||
|
|
$ git clone https://github.com/k7hoven/np.git
|
||
|
|
$ cd np
|
||
|
|
$ python setup.py install
|
||
|
|
|
||
|
|
Basic Usage
|
||
|
|
===========
|
||
|
|
|
||
|
|
Even before the `np` tool, a popular style of using ``numpy`` has been to import it as ``np``:
|
||
|
|
|
||
|
|
.. code-block:: python
|
||
|
|
|
||
|
|
>>> import numpy as np
|
||
|
|
>>> my_array = np.array([3, 4, 5])
|
||
|
|
>>> my_2d_array = np.array([[1, 2], [3, 4]])
|
||
|
|
|
||
|
|
The most important feature of ``np`` is to make the creation of arrays less verbose, while everything else works as before. The above code becomes:
|
||
|
|
|
||
|
|
.. code-block:: python
|
||
|
|
|
||
|
|
>>> import np
|
||
|
|
>>> my_array = np[3, 4, 5]
|
||
|
|
>>> my_2d_array = np[[1, 2], [3, 4]]
|
||
|
|
|
||
|
|
As you can see from the above example, you can create numpy arrays by subscripting the `np` module. Since most people would have numpy imported as ``np`` anyway, this requires no additional names to clutter the namespace. Also, the syntax ``np[1,2,3]`` resembles the syntax for ``bytes`` literals, ``b"asd"``.
|
||
|
|
|
||
|
|
The `np` package also provides a convenient way of ensuring something is a numpy array, that is, a shortcut to ``numpy.asanyarray()``:
|
||
|
|
|
||
|
|
.. code-block:: python
|
||
|
|
|
||
|
|
>>> import np
|
||
|
|
>>> mylist = [1, 3, 5]
|
||
|
|
>>> mylist + [7, 9, 11]
|
||
|
|
[1, 3, 5, 7, 9, 11]
|
||
|
|
>>> np(mylist) + [7, 9, 11]
|
||
|
|
array([8, 12, 16])
|
||
|
|
|
||
|
|
As an experimental feature, there are also shortcuts for giving the arrays a specific data type (numpy dtype):
|
||
|
|
|
||
|
|
.. code-block:: python
|
||
|
|
|
||
|
|
>>> np[1, 2, 3]
|
||
|
|
array([1, 2, 3])
|
||
|
|
>>> np.f[1, 2, 3]
|
||
|
|
array([ 1., 2., 3.])
|
||
|
|
>>> np.f2[1, 2, 3]
|
||
|
|
array([ 1., 2., 3.], dtype=float16)
|
||
|
|
>>> np.u4[1, 2, 3]
|
||
|
|
array([1, 2, 3], dtype=uint32)
|
||
|
|
>>> np.c[1, 2, 3]
|
||
|
|
array([ 1.+0.j, 2.+0.j, 3.+0.j])
|
||
|
|
|
||
|
|
Changelog
|
||
|
|
=========
|
||
|
|
|
||
|
|
0.2.0 (2016-03-29)
|
||
|
|
------------------
|
||
|
|
|
||
|
|
- Quick types are now np.i, np.f, np.u, np.c, or with the number of /bytes/ per
|
||
|
|
value appended: np.i4 -> int32, np.u2 -> uint16, np.c16 -> complex128, ...
|
||
|
|
(still somewhat experimental)
|
||
|
|
- Removed the old np.i8 and np.ui8 which represented 8-bit types, which
|
||
|
|
was inconsistent with short numpy dtype names which correspond to numbers of
|
||
|
|
bytes. The rest of the bit-based shortcuts are deprecated and will be removed
|
||
|
|
later.
|
||
|
|
- Handle Python versions >=3.5 better; now even previously imported plain numpy
|
||
|
|
module objects become the exact same object as np.
|
||
|
|
- Tests for all np functionality
|
||
|
|
- Ridiculously slow tests that runs the numpy test suite several times to
|
||
|
|
make sure that np does not affect numpy functionality.
|
||
|
|
- Remove numpy from requirements and give a meaningful error instead if numpy
|
||
|
|
is missing (i.e. install it using your package manager like conda or pip)
|
||
|
|
- Better reprs for subscriptable array creator objects and the np/numpy module.
|
||
|
|
|
||
|
|
0.1.4 (2016-01-26)
|
||
|
|
------------------
|
||
|
|
|
||
|
|
- Bug fix
|
||
|
|
|
||
|
|
0.1.2 (2015-06-17)
|
||
|
|
------------------
|
||
|
|
|
||
|
|
- Improved experimental dtype shortcuts: np.f[1,2], np.i32[1,2], etc.
|
||
|
|
|
||
|
|
0.1.1 (2015-06-17)
|
||
|
|
------------------
|
||
|
|
|
||
|
|
- PyPI-friendly readme
|
||
|
|
|
||
|
|
0.1.0 (2015-06-17)
|
||
|
|
------------------
|
||
|
|
|
||
|
|
- First distributable version
|
||
|
|
- Easy arrays such as np[[1,2],[3,4]]
|
||
|
|
- Shortcut for np.asanyarray(obj): np(obj)
|
||
|
|
- Experimental dtype shortcuts: np.f64[[1,2],[3,4]]
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
Platform: UNKNOWN
|
||
|
|
Classifier: Development Status :: 5 - Production/Stable
|
||
|
|
Classifier: Natural Language :: English
|
||
|
|
Classifier: Operating System :: OS Independent
|
||
|
|
Classifier: Programming Language :: Python :: 2.7
|
||
|
|
Classifier: Programming Language :: Python :: 3
|