Nani is a Python library that provides an alternative approach to defining and viewing NumPy's arrays.
Upon getting started with NumPy, the rules to define
tend to quickly become confusing. Not only different syntaxes can create a
same data type, but it also seems arbitrary and hence difficult to remember
that sub-array data types can only be defined as tuples while structured data
types exclusively require lists made of field tuples, and so on.
To address this point, Nani takes the stance of offering one—and only one—way
numpy.dtype objects. Although this syntax makes the code
more verbose, it also makes it easier to read and to reason about.
Nani's approach allows type introspection which brings additional benefits
in the form of dynamically generated default values and view types.
Default values facilitate the definition of new array elements while view types
are useful to encapsulate interactions with NumPy and to expose a different
public interface to your library users instead of the one provided with
- explicit syntax for defining
- generates default values and view types.
- allows for type introspection.
>>> import numpy >>> import nani >>> color_type = nani.Array( ... element_type=nani.Number(type=numpy.uint8, default=255), ... shape=3, ... view=None) >>> dtype, default, view = nani.resolve(color_type, name='Color') >>> a = numpy.array([default] * 2, dtype=dtype) >>> v = view(a) >>> for color in v: ... print(color) [255, 255, 255] [255, 255, 255]
color_type above defines an array of 3
numpy.uint8 elements having
each a default value of
255. The resulting
objects are used to initialize a new NumPy array of 10 color elements, while
view type is used to wrap that array into a standard collection
See the Tutorial section from the documentation for more detailed examples and explanations on how to use Nani.
Read the documentation online at nani.readthedocs.io or check its source in
Projects using Nani include:
Christopher Crouzet <christophercrouzet.com>