Skip to content

astrofrog/numtraits

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

72 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Numerical properties for Python objects

Travis Build Status Coverage Status

Please note: this package is experimental and may still see some changes to the API. If you have any suggestions for improving the API, please open an issue!

About

This simple module defines a descriptor class that can be used to define numerical properties (scalar and n-dimensional arrays) on classes and provide a way to validate these. Thus, instead of writing something like:

class Sphere(object):

    @property
    def radius(self):
        return self._radius

    @radius.setter
    def radius(self, value):
        if value <= 0:
            raise ValueError("Value should be strictly positive")
        if not np.isscalar(value):
            raise TypeError("Value should be a scalar")
        if not np.isreal(value):
            raise TypeError("Value should be numerical")
        self._radius = value

for each property you want to define, you can simply do:

from numtraits import NumericalTrait
from traitlets import HasTraits

class Sphere(HasTraits):

    radius = NumericalTrait(domain='strictly-positive', ndim=0)

The NumericalTrait class is implemented on top of the traitlets module. Any class using NumericalTrait for the definition of a property must derive from the traitlets.HasTraits class.

Support is also included for checking the dimensionality and shape of arrays (which includes converting tuples and lists to arrays on-the-fly), as well as checking the units of quantities for the astropy.units, pint, and quantities unit frameworks.

Installing

This package is compatible with Python 2.7, 3.3 and later, and requires numpy and traitlets. If you are interested in doing unit validation, you will also need astropy, pint, or quantities, depending on which unit framework you normally use.

To install, you can do:

pip install numtraits

You can also bundle numtraits.py into your package if you want to avoid using an external dependency, but please be sure to keep the copyright and the license in the file.

Using

To create self-validating numerical properties on a class, use the NumericalTrait class:

from traitlets import HasTraits
from numtraits import NumericalTrait

class Sphere(HasTraits):

    radius = NumericalTrait(domain='strictly-positive', ndim=0)
    position = NumericalTrait(shape=(3,))

When a property is set, it will be validated:

>>> s = Sphere()
>>> s.radius = 1.
>>> s.radius = -3
...
TraitError: radius should be strictly positive
>>> s.radius = [1,2]
...
TraitError: radius should be a scalar value
>>> s.position = (1,2,3)
>>> s.position = 3
...
TraitError: position should be a 1-d sequence
>>> s.position = (1,2,3,4)
...
TraitError: position has incorrect length (expected 3 but found 4)

The following arguments to NumericalTrait are available:

  • ndim: restrict the values to arrays with this number of dimension
  • shape: restrict the values to arrays with this shape. If specified, ndim does not need to be given.
  • domain: restrict the values to a particular domain - can be one of positive, strictly-positive, negative, strictly-negative, or a tuple representing a range of values.
  • default: the default value to return, if not specified (defaults to None)
  • convertible_to: restrict the values to ones with units that would be convertible to a specific set of units (see section below)

Note that tuples and lists will automatically get converted to Numpy arrays, if they are considered valid.

Physical units

While NumericalTrait can be used for plain scalars and Numpy arrays, it can also be used for scalars and arrays which have associated units, with support for three popular unit handling units: astropy.units, pint, and quantities.

To restrict a NumericalTrait to quantities with a certain type of unit, use the convertible_to option. This option takes units from any of these three unit packages, and will ensure that any value passed has units equivalent (but not necessarily equal) to those specified with the convertible_to option.

If the units passed to convertible_to are astropy.units units, then any value passed to the property should then be an astropy.units quantity. If the units passed to convertible_to are pint units, then any quantity passed to the property should be a pint property. And finally if the units passed to convertible_to are quantities units, then any quantity passed to the property should be a quantities quantity.

astropy.units Quantity example

The following example shows how to restrict the radius property to be an astropy.units quantity in units of length:

from astropy import units as u

class Sphere(HasTraits):
    radius = NumericalTrait(convertible_to=u.m)

will then behave as follows:

>>> s = Sphere()
>>> s.radius = 3. * u.m
>>> s.radius = 4. * u.cm
>>> s.radius = 4. * u.s
...
TraitError: radius should be in units convertible to m

pint Quantity example

The following example shows how to restrict the radius property to be a pint quantity in units of length:

from pint import UnitRegistry
ureg = UnitRegistry()

class Sphere(HasTraits):
    radius = NumericalTrait(convertible_to=ureg.m)

will then behave as follows:

>>> s = Sphere()
>>> s.radius = 3. * ureg.m
>>> s.radius = 4. * ureg.cm
>>> s.radius = 4. * ureg.s
...
TraitError: radius should be in units convertible to meter

quantities Quantity example

Finally, the following example shows how to restrict the radius property to be a quantities quantity in units of length:

import quantities as pq

class Sphere(HasTraits):
    radius = NumericalTrait(convertible_to=pq.m)

will then behave as follows:

>>> s = Sphere()
>>> s.radius = 3. * pq.m
>>> s.radius = 4. * pq.cm
>>> s.radius = 4. * pq.s
...
TraitError: radius should be in units convertible to m

Planned support

  • Linking of properties (e.g. a property should have the same dimensions as another)

About

Sanity checking for numerical properties/traits 🔢

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages