GitLab

Increased version number to 3.0.4

Set default dtype for FieldArray

Fixed the bug that the distribution strategy of the return field of power_analyze was the original but not the natural one.

- Use attrs for storing metadata
- Remove pickle usage from Field and PowerSpace

Fixed a few small bugs.

Added keyword arguments mean and std to Field.power_synthesize. Bug fixes in Field, SmoothingOperator, LMSpace and Space.

Removed NIFTy 2 files that have not been ported so far.

-> A Field on a PowerSpace can now be initialized with a spectral function.

Started to consolidate the code base of LM <-> GL/HP transformations.

- Refactor _smooth_helper
- Remove inplace property
- Rename method compute_k_array to distance_array in RGSpace
- Fix compute_k_array usage

Modified power_analyze in Field for complex power spectra.

Fixed violation of LSP for implemented keyword/property.

- Create skeleton for TransformationOperator
- Minor bug fixes in Field, FFTTransform and RGRGTransformations and
  __init__.py for endomorphic_operator

Restructured new operator classes -> Introduced endomorphic_operator.py in order to avoid violation of Liskov's SP for square- and diagonal operator.
Removed paradict from operators.
Added from_random as static method to Field.

Fixed furthermore:
- Let's undo the changes in the RGSpace.__init__ (see the comment on gitlab)
- The HPSpaceParadict still has `distances` in the list.
- HPSpace.copy: Is this method really necessary? The one from Space should be sufficient, right?
- HPSpace.total_volume is missing
- Unnecessary imports of `random`, `LMSpace`,... in hp_space.py
- I removed the plotting functionality from the space classes.
- GLSpace.copy should also not be necessary
- GLSpace.dim is missing
- GLSpace.vol should be called `total_volume`. Also its implementation is defect: there is no self.distances for GLSpace. The total volume is just 4*pi.
- Use nifty_utilities.cast_axis_to_tuple in order to sanitize `axes` input in GLSpace.weight
- typo in `new_shape = np.ones(x.shape)`