Release Notes

Version 2025.6 (May 19, 2025)

Bug fixes:

  • Fixed #76 Bichromatic models appear now in ‘Peak model’ combobox.

Version 2025.5 (April 28, 2025)

New features:

  • #55 Added Fit Statistics (PySide).

Enhancements:

  • #67 Added Peaks Highlighting when selecting a Peak in the Peaks Table (PySide).

Bug fixes:

  • Fixed #53 and #54 SpinBoxes now uses scientific notation and dynamic step.

  • Fixed #66 Save bichromatic settings between sessions

  • Fixed #68 Can’t preview model when no spectrum are loaded in pyside 2025.4

  • Fixed #69 Peaks dissapear from plot

  • Fixed #72 Peaks table’s columns can be too short

  • FWHM default value when changing model in Peaks Table is now the same as when adding a new peak

Version 2025.4 (April 2, 2025)

New features:

  • Parameters bounds (min and max) can now be defined interactively in the GUI with click buttons (interactive bounds)

  • ‘PowerLaw’ has been added in the list of predefined background models

  • Log axes are now available for profiles visualization

  • Weights can be supplied to the fit model via the input data files (through a 3rd and optional column in .txt file)

  • .hdf input data files following the NXcanSAS standard can now be read

  • Optional bichromatic models have been added in “More Settings”

Enhancements:

  • The “Semi-Auto” baseline algorithm is now ~x60-90 faster

Bug fixes:

  • Fixed #52: Input file with special characters can not be read

  • Fixed #56: Model pre-visualization is not adapted

Version 2025.3 (March 14, 2025)

Enhancements:

  • #49 Add possibility to view/edit .json model without having to load Spectrum

  • Adds multi-row edit

  • Adds dx0 and dfwhm settings

  • List items foreground color now based on background luminance

Bug fixes:

  • Fixed #45: Preview mode prevents from scrolling in peaks table

  • Fixed #46: Disable widgets edit on wheel scroll

  • Fixed #47: Map marker size is not tied to pixel size

  • Fixed #48: Can’t edit peaks after fit

  • Fixed #50: RuntimeWarning: All-NaN slice encountered

  • Fixes “NotImplementedError: cannot remove artist”

Version 2025.2 (February 24, 2025)

Enhancements:

  • Added python version in About window (PySide).

  • Added reinit_spectra method to Appli class for spectra reinitialization (PySide).

Bug fixes:

  • Fixed #42: Do not reinitialize params when ‘vary’ is False

  • Fixed #43: Fit fails when Xmin/Xmax contains “inf” or “NaN”

  • Fixed #44: Peak Table values not in the correct columns

Version 2025.1 (February 17, 2025)

New features:

  • A PySide application has been added (MAJOR CHANGE), aiming to be more flexible, easier to maintain (using the MVC approach), and more user-friendly (featuring Drag-and-Drop functionality, etc…).

  • Consequently, the project structure has been significantly reorganized: Spectrum, Spectra, and related modules have been moved into the fitspy.core directory, and fitspy.app has been renamed to fitspy.apps (with separate subdirectories for tkinter and pyside).

  • Input data can now be saved directly in the model (as a .json file), allowing it to be replayed as-is, without relying on file path dependencies.

Enhancements:

  • Spectrum objects can be created directly by passing ‘fnames’ to Spectra(fnames=fnames)

  • a ‘model.json’ fname can be passed directly as ‘model’ to apply_model(model=model.json)

  • Peak label visibility has been improved (PySide).

  • Colormap settings have been added for peaks and 2D maps (PySide).

  • User warnings have been introduced for incorrect bounds in model parameters (PySide).

  • The ‘baseline’ subtraction feature has been extended to include the ‘background’.

  • Spectra subsampling has been introduced in the semi-automatic baseline approach to reduce calculation time when applied to very large spectra.

Bug fixes:

  • create the related dirname_fig in save_figures() if the directory does not exist

Version 2024.5 (September 4, 2024)

New features:

  • A semi-automatic approach for the baseline determination has been added

  • The spectra can now be displayed before and after the baseline subtraction

Enhancements:

  • The progressbar (with the number of CPUS used during the calculations) has been integrated in the processing windows.

  • The spectra input data formats have been extended to .dm3, .dm4, .emd, .hspy, .nxs files and many other formats thanks to the rosettasciio readers integration

Bug fixes:

  • Allow fit processing without any peak or background model (just for baselines visualization for instance)

Version 2024.4 (February 27, 2024)

New features:

  • noise is now estimated and a noise level criteria enables to automatically deactivate peak models in noisy regions.

  • optional outlier detection has been added, enabling the disregarding of outliers during baseline calculation and fitting processes.

  • ‘xtol’ fitting parameter has been added (associated with ‘leastsq’ and ‘least_square’ fitting methods).

  • all fitted parameters from all loaded spectra can now be saved in a single file named ‘results.csv’.

Enhancements:

  • multi-threaded calculations return now the complete fit report.

  • a new boolean argument, ‘reinit_guess’, has been added in Spectrum.fit(), enabling the adjustment of initial values for ‘ampli’ and ‘fwhm’ to the current spectrum. This helps circumvent “ill-conditioned” peak models (with fwhm~0) resulting from previous ‘Fitspy’ model calculations.

  • lists are now written to .json files on a single line.

  • parameters and statistics displaying have been separated.

Bug fixes:

  • the index count used to label the models was not reset to 1 when reloading a model

  • 2d-maps with a single X or Y coordinate were not supported

  • the Lorentzian model could previously return Nan values when dealing with fwhm=0

  • the threads execution used by apply_model() to display the different progress bars did not quit correctly in some cases

Version 2024.2 (January 29, 2024)

New features:

  • Add a progressbar in the terminal and in the GUI to follow the calculation progression during the fit processing

  • Enable the loading of user-defined models through dedicated ‘Load’ buttons in the GUI.

Enhancements:

  • Display the models parameters on the figure whatever the fit status

  • Highlight/reduce the peak model curve linewidth according to the fit success status

  • Reproduce the same view when reloading a model, considering the fit success status

  • Take into account the ‘fit_kws’ through the ‘kwargs’ of Spectrum.fit() to be passed to the lmfit Model.fit()

Code changes:

  • Baseline format/use revisited to disable multiple baselines subtract capability (the ‘baseline_history’ spectrum attributes has been removed)

  • Spectrum attributes has benn changed from:

    • ‘peaks’ to ‘attractors’

    • ‘peaks_params’ to ‘attractors_params’

    • ‘models’ to ‘peak_models’

    • ‘models_labels’ to ‘peak_labels’

    • ‘models_index’ to ‘peak_index’

Bug fixes:

  • Fixed bug when reloading ‘old’ model.json in which bkg_model was absent

  • Fixed bug when using user-defined model in multithreading

  • Fixed bug in bkg_model plotting

  • Fixed bug in removing ‘old’ Spectra 2D-map when reloading a ‘new’ one (the ‘old’ Spectra 2D-map figures are now closed)

Version 2024.1 (January 16, 2024)

New features:

  • Add user-defined model capability from external files (in a ‘.txt’ file located in %HOMEUSER%/Fitspy for models creation from literal expressions or in a ‘.py’ file from python scripting)

  • Add Fitspy static HTML Sphinx documentation in https://cea-metrocarac.github.io/fitspy

  • in 2D-map, in addition to intensity, the model parameters can now be displayed and the corresponding 2D field can be exported in a .csv file

  • Spectra, Spectrum and SpectraMap class have now their own dedicated .py modules

Enhancements:

  • Add background visualization and display the corresponding parameters in the tabview

  • Display only the used parameters models in the tabview

  • Add xmin and xmax (optional) arguments to spectrum.load_profile() to ease the x-range setting when loading profiles by python scripts

Bug fixes:

  • Fixed bug for system identification that enables clipboard copy on Windows only

  • Fixed bug on data paths in the examples

  • Make the application exit correctly and add a widget to confirm it

  • Fixed bug: make the fitting possible for a standalone background model

  • Fixed bug: enable the background models to be saved and reloaded

Version 2023.x

First releases.