Changelog — kikuchipy 0.14.1 documentation

kikuchipy is a library for processing, simulating and indexing of electron backscatter diffraction (EBSD) patterns in Python, built on the tools for multi-dimensional data analysis provided by the HyperSpy library: https://kikuchipy.org.
All user facing changes to this project are documented in this file. The format is based on Keep a Changelog, and this project tries its best to adhere to Semantic Versioning.
List entries are sorted in descending chronological order. Contributors to each release were listed in alphabetical order by first name until version 0.7.0.
A file written with EBSDDetector.save() with kikuchipy < 0.12.0 and NumPy >= 2 can store a detector with shape (60, 60) as “(np.int32(60), np.int32(60)”, which could not be read before. Now it can. (#834)
Loading an EBSD detector that was saved with EBSDDetector.save(), or an EBSD signal that was saved in the kikuchipy h5ebsd format, with kikuchipy < 0.12.0 now correctly negates the azimuthal angle, as the interpretation of the angle changed in 0.12.0. (#831)
EBSDDetector.euler now matches the detector-sample geometry used for projection when both the detector tilt and azimuthal angles are non-zero. (#831)
The detector tilt, azimuthal, and twist angles are now all derived from the detector orientation in Oxford Instruments h5ebsd (H5OINA) files, so that the detector-sample geometry matches the one in the file. Previously, only the detector tilt was read, and only approximately. (#831)
The sample tilt is now read from Bruker h5ebsd files. Previously, it was always set to 0 degrees. (#831)
The detector azimuthal angle read from EDAX h5ebsd files is now negated, so that the detector-sample geometry is the same as with kikuchipy < 0.12.0. (#831)
Loading an EBSD detector with EBSDDetector.load() raises an error if the file does not contain a valid kikuchipy version, and so does loading a kikuchipy h5ebsd file with an invalid version. (#831)
EBSDDetector.to_stereo() for projecting one or more patterns from the detector onto the sphere, returned as a master pattern in the stereographic projection with both hemispheres. Intensities from multiple patterns, and optionally from all symmetrically equivalent orientations, are averaged. (#807)
Optimization of processing parameters of a single pattern against a reference pattern, scored by the normalized cross-correlation: pattern.optimize_remove_dynamic_background(), pattern.optimize_adaptive_histogram_equalization(), and pattern.optimize_bandpass_filter() optimize one processing step each, while pattern.optimize_pattern_processing() optimizes any of these steps together, in any order. Parameters are found by Bayesian optimization, which requires the optional dependency bayesian-optimization, or by local optimization with the Nelder-Mead method with restarts. A signal mask can be passed to exclude pixels from the score. (#816)
draw.plot_pattern_processing_result() for plotting the pattern after each step of a pattern processing optimization, annotated with image quality and normalized cross-correlation. (#816)
filters.bandpass_fft_filter() for creating a band-pass filter transfer function for use with EBSD.fft_filter(), e.g. with parameters from pattern.optimize_bandpass_filter(). (#816)
pattern.get_signal_mask() for getting a signal mask of pixels without Kikuchi diffraction in a pattern by intensity thresholding, for use in e.g. refinement or pattern processing optimization. (#816)
Project master pattern from the Lambert to the stereographic projection. (#807)
pattern.remove_dynamic_background() no longer fails for patterns with an integer data type, and no longer overwrites patterns with a floating point data type. (#816)
Hough indexing, dictionary indexing, and refinement of a single pattern (without navigation axes) no longer create a crystal map with coordinates of 50 points instead of one. (#816)
Loading an EBSD detector that was saved with EBSDDetector.save() with kikuchipy < 0.12.0 now works. (#830)
The missing hemisphere of a master pattern of a centrosymmetric phase with only one hemisphere is now the present hemisphere rotated 180 degrees about the z axis, not a copy of it. A copy is only correct if the point group also has a 2-fold axis along z, and was wrong for point groups -1, -3, and -3m. This affects EBSDMasterPattern.get_patterns(), orientation and projection center refinement, plot_spherical(), draw.EBSDDetectorPlotter.set_master_pattern(), and reading an ebsdsim master pattern file of a centrosymmetric phase with hemisphere="both". (#837)
normalize_intensity() of all signals and pattern.normalize_intensity() ignore NaN values when calculating the mean and standard deviation. (#807)
Minimum version of NumPy is 1.25 and of Numba is 0.58, since the optional dependency bayesian-optimization requires NumPy >= 1.25, which Numba supports from version 0.58. (#816)
The normalized cross-correlation metric “ncc” returns a score of zero instead of NaN for flat patterns. (#816)
Trying to project a master pattern in the Lambert projection with as_lambert() now raises a ValueError instead of just warning. (#835)
All parameters to the master pattern method as_lambert() must be given as keyword arguments. (#835)
Energy axis of master patterns read from ebsdsim files written by ebsdsim >= 0.1.6 are now correct. Files written by older versions of ebsdsim still read correctly. (#820)
Increased minimal version of ebsdsim >= 0.1.4. (#820)
More master patterns downloadable to a local cache via data.ebsd_master_pattern(): Alternative sigma (Fe-Cr-Mo), Cr2N, R (Fe-Cr-Mo), Pi (Fe-Mo), Al6Mn, and alpha-AlMnSi. (#796)
Function to clear the kikuchipy data cache directory. (#796)
Optional dependency on the ebsdsim Python package. (#809)
Reader for master patterns generated by the ebsdsim Python package. (#809)
Overflow error when loading EDAX binary up1/up2 files greater than ~4 GB. (#804)
Invert the rotation EBSDDetector.sample_to_detector so that it transforms vectors in the correct sense, i.e. rep