astropy.nddata.support_nddata(_func=None, accepts=<class 'astropy.nddata.nddata.NDData'>, repack=False, returns=None, keeps=None, **attribute_argument_mapping)[source]#

Decorator to wrap functions that could accept an NDData instance with its properties passed as function arguments.

_funccallable(), None, optional

The function to decorate or None if used as factory. The first positional argument should be data and take a numpy array. It is possible to overwrite the name, see attribute_argument_mapping argument. Default is None.

acceptsclass, optional

The class or subclass of NDData that should be unpacked before calling the function. Default is NDData

repackbool, optional

Should be True if the return should be converted to the input class again after the wrapped function call. Default is False.


Must be True if either one of returns or keeps is specified.

returnsiterable, None, optional

An iterable containing strings which returned value should be set on the class. For example if a function returns data and mask, this should be ['data', 'mask']. If None assume the function only returns one argument: 'data'. Default is None.


Must be None if repack=False.

keepsiterable. None, optional

An iterable containing strings that indicate which values should be copied from the original input to the returned class. If None assume that no attributes are copied. Default is None.


Must be None if repack=False.


Keyword parameters that optionally indicate which function argument should be interpreted as which attribute on the input. By default it assumes the function takes a data argument as first argument, but if the first argument is called input one should pass support_nddata(..., data='input') to the function.

decorator_factory or decorated_functioncallable()

If _func=None this returns a decorator, otherwise it returns the decorated _func.


If properties of NDData are set but have no corresponding function argument a Warning is shown.

If a property is set of the NDData are set and an explicit argument is given, the explicitly given argument is used and a Warning is shown.

The supported properties are:

  • mask

  • unit

  • wcs

  • meta

  • uncertainty

  • flags


This function takes a Numpy array for the data, and some WCS information with the wcs keyword argument:

def downsample(data, wcs=None):
    # downsample data and optionally WCS here

However, you might have an NDData instance that has the wcs property set and you would like to be able to call the function with downsample(my_nddata) and have the WCS information, if present, automatically be passed to the wcs keyword argument.

This decorator can be used to make this possible:

def downsample(data, wcs=None):
    # downsample data and optionally WCS here

This function can now either be called as before, specifying the data and WCS separately, or an NDData instance can be passed to the data argument.