SamM755 opened a new pull request, #18809:
URL: https://github.com/apache/iotdb/pull/18809

   ## Description
   
   Target: `apache/iotdb:research/cluster-compress`.
   
   Enable the companion TsFile research implementation of native aligned and 
scalar Cluster-compressed numeric pages by updating the TsFile dependency to 
`2.1.0-SNAPSHOT`. Ordinary SQL can create ACLUSTER/KCLUSTER series, insert 
numeric records, flush, and query typed values. The existing parser already 
recognizes these encodings, so this change does not add SQL grammar or alter 
server timestamp configuration.
   
   ### Companion TsFile dependency
   
   Depends on [apache/tsfile#1000](https://github.com/apache/tsfile/pull/1000), 
targeting `apache/tsfile:research/cluster-compress`. It contains the joint 
column selection, multidimensional ACluster/KCluster, ALP companion-column 
encoding, and native page readers/writers. Install its matching Java `common` 
and `tsfile` artifacts before building this IoTDB branch. Resolving an 
unrelated `2.1.0-SNAPSHOT` artifact is insufficient.
   
   Timestamp handling remains as implemented in that companion: exact timestamp 
keys follow the same tuple reordering; they are not sent through ALP. The 
aligned shared time column retains its native time encoding. The separate 
ClusterTableSession BLOB API and its tests are retained alongside the native 
SQL path.
   
   ### Validation
   
   - ClusterTableSessionTest: 3 tests passed.
   - DataNode's 3,266 Java sources compiled against the companion TsFile 
dependency.
   - ClusterAlignedSqlIT passed on an isolated IoTDB 2.0.2-SNAPSHOT server at 
the research baseline with the newly built TsFile/common jars. For each of 
ACLUSTER and KCLUSTER, it verified 1,200 aligned rows and 400 scalar rows, 
exact timestamp/value associations, projection, time/value predicates, 
descending scans, and count/sum/first/last aggregates.
   - After a forced DataNode restart, read-only checks passed again, including 
recovery of 200 unflushed aligned rows per method from WAL.
   - ClusterTableSessionIT verified the retained BLOB API with 1,200 complete 
records per method before and after restart on the same isolated server.
   - The companion TsFile contribution passed 72 distinct targeted 
unit/regression cases plus Checkstyle, Spotless, and license checks.
   
   Validation was performed on 2026-10-08. This is the research-branch 
integration; it does not claim completion of the full repository test suite or 
comprehensive validation of compaction, replication, schema evolution, and 
full-scale performance.
   
   This PR has:
   
   - [x] added documentation for the retained BLOB API and its relationship to 
the native SQL path.
   - [x] added Session unit tests.
   - [x] added opt-in live SQL and complete-record integration checks.
   - [x] been tested on an isolated IoTDB service, including flush and 
forced-restart recovery.
   
   ##### Key changed/added classes
   
   - ClusterTableSession: the retained complete-record BLOB interface.
   - ClusterAlignedSqlIT: ordinary SQL checks for aligned/scalar native Cluster 
pages, including read-only restart verification.
   - ClusterTableSessionTest and ClusterTableSessionIT: unit and live checks of 
the independent BLOB path.
   - Root and Session POMs: companion research TsFile dependency.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to