New Features

Query & Execution

• Support function ARRAY_CROSS_PRODUCT (#64031)
• Add exponential_moving_average aggregate function (#63499)
• Support murmur_hash3_128 function (#63196)
• Add datasketches HLL sketch aggregate functions (#63143)
• Skip collecting stats for long string columns (#62686)
• Add information_schema role mappings table (#62077)
• Add FE constant folding for cosine_similarity and standardize test patterns (#60403)

Cloud Native

• Show compute group for MTMV refresh task (#63206)

Security & Authentication

• Integrate OIDC authentication, MySQL login bridge, and role mapping (#61819)

Improvement

Query & Execution

• Use real elapsed time to compute workload group metrics refresh interval (#63537)
• Support TIMESTAMPTZ in multiple aggregate and array functions (#62756)

Storage & Compaction

• Enable packed file and empty rowset optimization by default (#63475)
• Support configurable S3 credentials providers (#62788)
• Add enable_recycler config to skip recycler dynamically (#63286)

Bugfix

Index & Search

• Fix reading packed inverted index file on file cache miss (#64383)
• Fix ANN IVF/PQ recall, avoid init-time large ANN build-buffer reservation, and skip ANN index bui... (#64082)
• Guard LogicalView.computeOutput() against schema drift (IndexOutOfBoundsException) (#64007)
• Fix ANN range search state leakage and incorrect slot index tracking. (#63666)
• Clamp variant_sparse_hash_shard_count to >=1 in SHOW CREATE output (#63661)
• Bind Variant search to nested indexes (#63660)
• Reject Lucene-syntax SEARCH on columns without inverted index (#63637)
• Preserve operative slots when deep copying logical relations (#63315)
• Preserve variant subfields in view definitions to fix select view result wrong when view select h... (#62907)
• Fix tokenize function incorrect result when first argument is const (#62699)
• Allow MATCH on aliased variant subcolumns (#63772)
• Improving AI function performance (#62494)

Query & Execution

• Reject TopN in correlated scalar subquery (#64251)
• Optimize floating fmod fast path (#64161)
• Make role-mapping keywords RULE/CEL/MAPPING non-reserved (#64104)
• Prevent unsafe runtime filter pushdown through outer joins (#64102)
• Fix datediff folding for zero date (#64084)
• Align convert_tz folding with BE DST handling (#64029)
• Merge struct_element into element_at (#64027)
• Fix TopN runtime filter activation (#63969)
• Preserve negative zero sign in SIGNBIT constant folding (#63954)
• Fix assert row join pushdown alias handling (#63892)
• Fix array subscript on pruned variant subpath (#63891)
• Align COM_RESET_CONNECTION behavior with MySQL (#63884)
• Preserve NaN in numeric constant folding (#63870)
• Avoid potential OOM when reading large snapshot splits (#63833)
• Cast variant subcolumn as json in variant_hirachinal for stable output (#63828)
• Fix changed variable output in show variables (#63734)
• Fix computeDestIdToInstanceId picking wrong ExchangeNode for multi-input fragments (#63615)
• Avoid unsigned underflow in JSON modify path (#63579)
• Handle legacy DecimalV2 segments with missing precision/frac (#63569)
• Preserve TIMESTAMPTZ values in sparse path (#63522)
• Align legacy literal compareLiteral with Nereids ComparableLiteral semantics (#63481)
• Reject COUNT DISTINCT on variant arguments (#63479)
• Compare JSON numeric values by value (#63396)
• Remove unsafe JsonbWriter key overload (#63355)
• Prevent invalid alias rewrite in view definitions (#63353)
• Reject non-positive topn count argument (#63350)
• Fix json contains duplicate array candidates (#63301)
• Reject super wildcard path in json keys (#63300)
• Report TIMESTAMPTZ as string to MySQL clients (#63292)
• Add null reject compensation for join rewrite (#63268)
• Reject lone UTF-16 surrogates in JSONB literals (RFC 8259 §8.2) (#63255)
• Fix alias function with cast outermost expr and reject illegal expressions (#63254)
• Support TIMESTAMPTZ in TopN runtime predicate (#63220)
• Reject JSONB and variant distribution columns (#63211)
• Move the pruning of predicates that are always true after partition pruning into the PlanPostProc... (#63111)
• Fix coalesce function output null (#63092)
• Keep first duplicate Variant JSON path (#63082)
• Avoid unioning query-unused MV partitions (#63081)
• Fix integer typing and prefer Variable.realExpression for argument/type resolution (#62524)
• Fix redundant aggregation in agg-union query plan (#62231)
• Pass ConnectContext to canUseNereidsDistributePlanner method instead of call ConnectContext.get() (#60529)

Storage & Compaction

• Add from-to cdc WAL-search timeout and stale-reader release (#64013)
• Validate recycle rowset key state during commit rowset (#63985)
• Reduce sparse variant parse memory (#63970)
• Avoid retaining full segment key bounds buffers (#63968)
• Release packed file writer buffer after flush (#63967)
• Normalize SC rowset graph before delete bitmap capture (#63960)
• Support cdc_client JVM opts and adopt externally-managed cdc_client (#63898)
• Remove the udf cache expiration_time property (#63897)
• Drain txn lazy committer workers before destruction (#63876)
• Filter nereidsPrunedTabletIds per partition in distributionPrune (#63851)
• Keep prefetch reader alive for async tasks (#63796)
• Avoid retrying object storage SlowDown errors (#63776)
• Remove single replica compaction (#63771)
• Cache cluster id per query and drop redundant locks on getBackendId hot path (#63636)
• Fix meta tool build on master (#63540)
• Skip stale tablet cache check for STOP_TOKEN (#63520)
• Preserve labels for histogram metrics to fix wrong metric name for prometheus (#63485)
• Normalize default HDFS paths in LocationPath (#63476)
• Truncate segment key bounds before storing segment stats (#63469)
• Fill schema change version holes before running (#63443)
• Disable dict encoding in row store columns (#63438)
• Remove pure attribute from assert_cast (#63417)
• Fix clear_file_cache right after reboot causing file cache size percent overflow (#63410)
• Verify DML and 3-replica create table when one BE is down (#63401)
• Fix pre-aggregation context leakage across join branches (#63357)
• Clean empty v3 cache dirs (#63344)
• Avoid concurrent tablet stat iteration failures (#63298)
• Allow show tablet without selected database (#63280)
• Restore split-bound Java types when reading FE-persisted CDC offset (#63219)
• Support storage vault for clone instance (#63217)
• Delay overwrite partition routing until incremental open (#63209)
• Fix simple aggregate cache after partition recycle (#63175)
• Enhance OOM error message for statistics analyze tasks (#63172)
• Fix incorrect memory availability check in RowSourceBuffer during vertical compaction (#63152)
• Async chunk splitting for cdc source job (#63079)
• Replace std::mutex/std::lock_guard with annotated wrappers for thread safety analysis (#63070)
• Avoid NPE for force-finished publish task (#63069)
• Add SubQueue abstraction and thread-safety annotations to DataQueue (#62947)
• Skip wait for async rowset warmup (#62764)
• Cache version and get tablet stats actively for RestoreJob (#62704)
• Support storage vault for create/list snapshot (#62523)
• Infer null-reject from INNER JoinEdge for multi-hop outer join MV rewrite (#62492)
• Avoid crash when late holder cleanup sees removed cache cell (#62437)
• Fix PartitionRebalancer generating invalid moves to BEs without required storage medium (#62206)
• Add system rate limit for meta-service (#61516)
• Replace Tablet references with tabletId in CloudTabletRebalancer (#61233)
• Add async lru update machanism and fix partial hit in cache reader (#61083)
• Avoid duplicated FileCache counter accumulation in NewOlapScanner (#61072)
• Show proc should display partition cached version (#60807)
• Avoid false tablet diagnosis alarms in cloud mode (#60805)

Load

• Fix nullable date literal binding and date *_diff folding (#64127)
• Keep load row metrics monotonic for auto partition (#64109)
• Fix postgres cdc multi-table publication data loss and binlog duplicate key (#64075)
• Match STRUCT sub-fields by name when loading JSON (#64011)
• Change lzo download link (#63785)
• Fix postgres historical-date timestamp handling in cdc-client (#63618)
• Add per-job routine load metrics (#63576)
• Fix int overflow in BeIdComparator causing stream load failure (#63565)
• Add txn write amplification brpc metrics for sub txn load (#63545)
• Enforce explicit compute group form for workload DDLs (#63505)
• Support user-specified mysql server_id with per-reader assignment (#63490)
• Misc fixes for typo/log/validation/visibility (#63480)
• Avoid NPE on cross-table DML during snapshot chunk read (#63435)
• Keep isCanceled set when cancel runs on terminal task (#63427)
• Drop neighbour-table rows leaked by JDBC LIKE wildcards in JdbcPostgreSQLClient (#63402)
• Optimize row-store memtable flush memory in the row-store scenario (#63342)
• Fix broken pipe risk on stream load redirect with unconsumed request body (#63332)
• Pin esdk-obs-java-bundle to 3.21.11 to fix version range resolution failure (#63278)
• Add per-job lag metric to streaming insert jobs (#63194)
• Fix NPE in routine load Kafka meta request (#63180)
• Remove dead code across core types and utilities (#62994)
• CancelTaskById should not be blocked by unrelated streaming jobs (#62940)
• Recompute derived fields after replay and ALTER (#62936)
• Persist cdc_stream TVF offset across FE checkpoint (#62902)
• Avoid mutating shared variant columns (#64092)
• Select txn insert backend from current cluster (#63634)
• Fix host not match if start fe in metadata_failure_recovery (#62748)
• Deduplicate pending one-shot warm up jobs (#62384)

Lakehouse

• Mapping iceberg varint type to unsupported type (#64331)
• Fix iceberg write special partition error. (#64225)
• Fix can't get migrated Iceberg tables format type (#64134)
• Use object store path for data location (#64028)
• Reject iceberg COW table row-level DML (#63950)
• Fix iceberg v3 row lineage count distinct error result (#63826)
• Fix DCHECK in LocalExchangeSharedState::sub_total_mem_usage (#63742)
• Add missing Iceberg field IDs for position delete files. (#63483)
• Fill Hive meta cache when loading row count for queries (#63470)
• Fix iceberg sink writer with spill report error (#62899)
• Rest & S3Table Support Iam-role (#60498)

Security & Authentication

• Bound length in MysqlProto.readLenEncodedString (#63604)
• Include HDFS connection in file handle cache key (#63516)
• Fix arrow flight client ip auth (#63506)
• Preserve narrowing datetimev2 casts in simplify in predicate (#63343)
• Fix arrow::Status inline static empty msg core (#63191)
• Support multi-root auth plugin loading and normalize OIDC access token auth (#62159)
• Improve masking of user's password for ALTER USER and CREATE USER commands in audit logs (#62141)
• Improve LDAP authentication resiliency and diagnostics (#61673)
• Integrate authentication chain and simplify fallback config (#61362)