April 2026: Implemented Winsorized mean support for TDigest in Velox across scalar, API, and aggregate forms, backed by targeted refactors and documentation. Delivered a single-pass approx_winsorized_mean to replace the common two-pass pattern, improving latency on large datasets. Refactored core TDigest code by extracting shared helpers and consolidating the TDigestAccumulator into a common header to reduce duplication and improve maintainability. Added robustness enhancements (NaN handling and parameter validation) and comprehensive documentation to enable correct usage and faster adoption.

October 2025: Implemented Student's t-distribution support in both Presto and Velox, expanding in-SQL statistical capabilities and enabling robust hypothesis testing directly within queries. Delivered t_cdf(df, value) and inverse_t_cdf(df, p) with input validation and high-accuracy implementations. - prestodb/presto: Added t_cdf and inverse_t_cdf using Apache Commons Math3 TDistribution; included input validation (df > 0, p in [0,1]); updated math documentation. - oap-project/velox: Added C++ implementations of t_cdf and inverse_t_cdf using Boost.Math's students_t_distribution; registered functions, added unit tests, and updated documentation; aligned with existing probability function patterns. - Achieved cross-repo parity and consistent error handling across Java and C++ implementations, improving developer experience and reliability. - Strengthened quality with comprehensive tests covering edge cases and documentation updates to reflect new capabilities.