Skip to content

Release Notes

1.1.0 (2024-12-25) - Major Refactoring

Key Changes to FitResult:

  • FitResult is no longer a subclass of tuple. To achieve similar behavior and unpack res and res_func, use the new method fit_result.res_and_func().
  • All parameters are now directly accessible as attributes. Accessing parameters via fit.res is still possible but can be deprecated in the future.
  • The res as np.ndarray is now accessible via fit.res_array.
  • The FitArrayResult structure has been refactored and is now unified with FitResult.
  • Results in FitResult are guaranteed to never be None. If the fit fails, each parameter will return np.nan, and the success attribute will be set to False.

Other Changes:

  • array_fit can now be applied to any axes, offering greater flexibility.
  • Each implemented fit function now has a dedicated result class with predefined slots and custom methods.
  • The leastsq method has been removed. Without specifying an exact function, the FitResult for classical leastsq becomes useless. Use curve_fit instead, specifying method='leastsq'.

Normally these changes only affect you if you unpack the FitResult before. We apologize for any inconvenience, but we believe these changes will improve the package's usability in the long run and should be made as soon as possible.

1.0.0 (2024-11-16)

  • Automatic documentation generation for each function.
  • Automated testing for all functions.
  • New guide available for creating custom functions.

0.2.0 (2024-08-28)

  • Beta version of the package.