[3.39.0] - 2025-06-13

Enhancements

• Feature as Predictor (#6852)
• K-means: clusters can be inferred for new data (#7010)
• Parameter Fitter: more responsive when changing settings (#7027)
• Paint Data: Slight optimization (#6990)
• Numpy 2.0 compatibility (#6850)

Bugfixes

• Order available PostgreSQL tables (#7099)
• tableview: Avoid unnecessary calls to setSortingEnabled (#7109)
• Update check and notifications: avoid compression (#7097)
• tSNE: Add compute_value to tSNE variables (#7066)
• ScoringSheetViewer computing target class values from base values (#7025)
• ScoringSheetViewer slider style and tooltip (#7024)
• Disable dynamic signals where applicable (#7045)
• Make ReliefF and RReliefF results replicable (#7026)
• ScoringSheet deterministic results (#7023)
• io_util: Detect utf-8-sig when usingfile utility (#7006)
• Fix error bars for threshold averaged ROC curves (#7014)
• Save Data: fix when saving to different drives on Windows (#7019)
• owhierarchicalclustering: Fix IndexError (#6989)

[3.38.1] - 2024-12-23

Enhancements

• Scatter Plot: Error Bars (#6934)

Bugfixes

• 1 line headers: only allow single-letter types and flags (#6959)
• Group By: fix categorical aggregations (#6958)
• Orange can be imported without Qt (fixed importing localization) (#6955)
• Fix documentation in packages: fix sphinx build command (#6950)
• httpx 0.28 support: use mounts instead of proxies (#6943)
• owpredicttions: Remove special magic value 2 (#6933)

[3.38.0] - 2024-11-15

Enhancements

• Parameter Fitter: Basic implementation (#6921)
• Datasets: Add domain field; respect "Unlisted" (#6920)
• Datasets: Let the filter override domain and language (#6930)
• ScoringSheet and ScoringSheetViewer widgets (#6817)
• Multilingual package and installation (#6828)

Bugfixes

• Calibrated Learner: Prevent in place modification of base learner (#6917)
• ListView: Fix selection (#6823)
• plotutils: Fix scene layout tracking in AnchorItem (#6859)
• oweditdomain: Hide "categories mapping" warning on change (#6865)
• Logistic regression: fix penalty argument for no regularization (#6886)
• Avoid slowdowns by caching Domain.eq (#6764)
• SVM: degree has to be an integer (#6866)
• Only deepcopy the .attributes for the outermost Table transformation (#6849)
• Line plot: compatibility with scipy 1.14 (#6845)
• PCA: ensure tests pass on sklearn 1.4 and 1.5, which can return different results (#6821)
• Remove Orange implementation of randomized PCA for sparse data (#6815)

Release 3.34.0