Deprecations
- The
.G()
has been removed as planned.
Maintenance
- Switched to using
setup.cfg
instead ofsetup.py
and various other tooling updates.
Enhancements
- Add ability to perform hankel transforms on complex functions.
Enhancements
- Limit calculation for k=0 provided (resulted in nan before)
- Alternative hankel kernel now selectable: sqrt(x) * J(nu, x)
- Caching of series factors for faster calculations
- Various infrastructure upgrades (setuptools_scm, pre-commit, black)
Bugfixes
- Saver calculation of xrange_approx
Deprecations
- Methods
G
anddeltaG
now namedfinal_term_amplitude
andslope_of_last_term
.
Version corresponding to JOSS paper release.
Bugfixes
- Fixed SymmetricFourierTransform default N to be dynamic, like HankelTransform.
Enhancements
- Swapped to pytest from nose for all testing
- Removed warnings of overflow for cosh and sinh functions
- Improved documentation in many areas.
- Swapped out default 3.2/h for pi/h, which sounds way cooler :-)
Bugfixes
- Fixed an error in tests such that the wrong analytic function was being tested.
- Removed a check for deltaG in get_h, which sometimes fails spectacularly.
Bugfixes
- Fixed issue in Py3 in which dim=2 would not run the fast function (thanks @MuellerSeb)
Enhancements
- Added support for transforms with
nu=-0.5
(thanks @MuellerSeb)
Bugfixes
- Fixed Python 3 support from v0.3.4
Features
- Added get_h function to aide in determining optimal h value for a given transformation.
Enhancements
- Added _get_series method to quickly retrieve the summed series for the integration.
- Two updated notebook examples.
Bugfixes
- Moved setting of N to avoid error.
Features
- Some additional tools to determine accuracy -- quick calculation of last term in sum, and evaluated range.
Enhancements
- Default setting of N=3.2/h, which is the maximum possible N that should be chosen, as above this, the series truncates due to the double-exponential convergence to the roots of the Bessel function.
Bugfixes
- Fixed error in cumulative sum when k is not scalar.
Enhancements
- Documentation! See it at https://hankel.readthedocs.io
- Two new jupyter notebook demos (find them in the docs) by @francispoulin
Bugfixes
- Fixed relative import in Python 3 (tests now passing), thanks to @louity
- Fixed docstring of SymmetricFourierTransform to have correct Fourier convention equation
- Fixed bug in choosing alternative conventions in which the fourier-dual variable was unchanged.
Bugfixes
- Fixed normalisation for inverse transform in
SymmetricFourierTransform
.
Features
- Ability to set Fourier conventions arbitrarily in
SymmetricFourierTransform
.
Features
- New class
SymmetricFourierTransform
which makes it incredibly easy to do arbitrary n-dimensional fourier transforms when the function is radially symmetric (includes inverse transform). - Addition of
integrate
method to base class to perform Hankel-type integrals, which were previously handled by thetransform
method. This latter method is now used for actual Hankel transforms. - Documentation!
Enhancements
- Addition of many tests against known integrals.
- Continuous integration
- Restructuring of package for further flexibility in the future.
- Quicker zero-finding of 1/2-order bessel functions.
- This changelog.
- Some notebooks in the devel/ directory which show how various integrals/transforms behave under different choices of integration steps.
Enhancements
- Compatibility with Python 3 (thanks to @diazona)
- Can now use with array-value functions (thanks to @diazona)
Bugfixes
- Fixed pip install by changing readme --> README
Enhancements
- updated docs to show dependence on mpmath
Features
- Non-integer orders supported through mpmath.
- First working version. Only integer orders (and 1/2) supported.