Hi all,

I would like to propose starting the release process for Apache TsFile 2.4.0.

Compared with the previous release version 2.3.1, this release includes a set 
of user-visible
improvements, performance optimizations, and bug fixes. The main scope includes 
improvements to Python 
DataFrame/dataset support, C++ read/write performance optimizations, metadata 
loading improvements, and several compatibility and stability fixes.


The proposed release highlights include:

1. Improved Python DataFrame / dataset support


This release improves the Python TsFileDataFrame and dataset layer, including 
support for tree-model TsFiles, better unified access to table-model and 
tree-model files.

It also adds metadata index cache support to improve repeated initialization of 
TsFileDataFrame,
with additional test coverage.

2. C++ read/write performance optimizations

This release includes multiple optimizations in the C++ reader/writer path, 
including
batch read, SIMD-related improvements, aligned chunk reader improvements, table 
batch
read improvements, time filter batch evaluation, and parallel write-related 
improvements.

3. Metadata loading improvements
This release optimizes get_timeseries_metadata() to reduce N+1 I/O during 
metadata loading,
significantly improving metadata loading efficiency. It also fixes related 
table size accounting issues.


4. Encoding and compatibility fixes
This release fixes TS_2DIFF compatibility issues for float/double pages and 
improves C++
 encoding/decoding test coverage.

5. Large file and writer correctness fixes
This release fixes C++ TsFile writer flush offset issues for multi-GB files, 
adds large file tests, 
and introduces Tablet.serializedSize() together with more comprehensive size 
validation.


6. Build, CI, and usability improvements
This release also includes Windows CI / Python wheel build fixes, upgrades the 
minimum
JDK requirement to 17, adjusts C++ build options, and includes several 
Maven/CMake/workflow improvements.


Please let us know if you have any concerns about the release scope, or if 
there are any
blockers or important fixes that should be included before the release.

Best,
Colin












Reply via email to