v5.2.0
🚀 Features
- Introduce the block version ref key (#1969)
- Add AWS STS authentication support (#1884)
- Python 3.12 and Python 3.13 support (#1945) (#2016)
- Numpy 2 support (#2050)
- Reliable storage lock (#2014)
- Storage mover (#2039)
- Add S3 STS proxy support (#2072)
- Implement origin for pandas resampling (#1962)
- Refactor to storages to support async reads (#2012)
- Chunk up incomplete segments by rows when they are staged (#2117)
read_batch
performance improvements - now up to 10 times faster
🐛 Fixes
- Fix decoding of fields with >2^16 blocks (#2089)
- Handle very old normalization metadata RangeIndexes (#2118)
- Finalize staged data memory use improvements (#2013)
- Fix handling of empty DF in pandas 1.0 (#2010)
- Fix string reference count leak (#1998)
- Fix version release not attaching symbols for debug (#2018)
- Delete staged segments after writing vref key (#2037)
- Improve the performance of update by parallelising reads. Implement internal async update method. (#2087)
- Performance regression when requesting a timestamp before the earliest version (#2076)
- Using compact incomplete on a library with dynamic schema with a named index can result in an unreadable index (#2116)
- Notimplemented handling (#2108)
- Fix race between list_versions and delete_snapshot on NFS (#2092)
Full Changelog: v5.1.3...v5.2.0
The wheels are on PyPI. Below are for debugging: