Skip to main content

New Features

Storage & Compaction​

  • Add adaptive batch size for scan path (#62835)
  • Add CompactionTaskTracker with system table and HTTP API (#61696)
  • Add value predicate pushdown control for MOR tables (#60513)

Other​

  • Add json_object_flatten scalar function (#62825)

Improvement

Storage & Compaction​

  • Skip zero-length files in FileSplitter to avoid sending empty splits to BE (#62482)
  • Avoid single-point read/write during sequentially reading key (#62476)
  • Add stampede protection for AnnIndexIVFListCache (#62442)
  • Limit packed file writes to rowset first segment (#62342)
  • Unify current query runtime statistics and expose task progress (#60567)

Load​

  • Make thirdparty jar download mirror configurable (#62376)

Cloud Native​

  • Add HashCRC32Return32 and make join_hash_table use 32-bit hash value (#62512)

Other​

  • Add current_database builtin alias (#62591)

Bugfix

Index & Search​

  • Fix deep nested complex type subtype validation bypass (#63208)
  • Fix potential NPE in getDeleteBitmapUpdateLock (#62809)
  • Fix ANN query vector extraction to handle all constant expression types (#62637)
  • Split monolithic function_string.h into domain-specific files (#62262)
  • Fix OOB crash in partition column generation for Iceberg/Paimon tables (#62177)
  • Fix null pointer exception in sessionVariables after upgrade (#61959)
  • Optimize the compilation time of aggregate_function_orthogonal_bitmap.cpp (#61606)
  • Drop index should set db_id (#58401)

Query & Execution​

  • Fix remote Flight SQL result receiver initialization (#63136)
  • Keep LogicalOlapScan for short-circuit point query on empty table (#62948)
  • Fix DST spring-forward gap handling in timestamptz literal (#62945)
  • Fix UB and param order in assert_num_rows_operator (#62800)
  • Preserve TIMESTAMPTZ in LEAD/LAG by adding TimeStampTzType signature matching (#62779)
  • Fix exchange receiver dependency race (#62777)
  • Add fast path for canonical format datetime parse (#62757)
  • Fix isCountStar incorrectly treating count(null) as count(*) (#62548)
  • Fix view columns losing colUniqueId in lazy materialization (#62533)
  • Align extra-join elimination safety check (#62527)
  • Fix Doris query service fails after the file handles on the BE node are used up (#62393)
  • Scale num_nulls in col stats when partition pruned (#62265)
  • Fix prepared statement QPS metrics not counted when audit log disabled (#61621)
  • Update license-maven-plugin to 2.1.0 (#58951)

Storage & Compaction​

  • Avoid nested task attach in cloud snapshot manager (#63189)
  • Materialize NG compaction regular paths (#63104)
  • Rebuild broker load storage properties after Gson replay (#63094)
  • Skip statistics cache for system dbs (#63089)
  • Fix variant flat-leaf root read plan (#63086)
  • Resolve variant sub-column indexes for score() (#62992)
  • Avoid deleting nonexistent delete bitmap files (#62967)
  • Skip skewed warmup rowset latency samples (#62941)
  • Return unknown stats for system tables (#62913)
  • Mark internal query failures as ERR in audit log (#62908)
  • Fix warm up don't retry due to error message change (#62886)
  • Reject oversized block size in add_cell (#62878)
  • Move #include directives outside namespace blocks to avoid ODR violations (#62871)
  • Bound cdc_client RPCs with per-category timeouts (#62870)
  • Add NestedGroup path metadata support (#62848)
  • Reduce point-query network overhead by resending query context on cache miss (#62836)
  • Skip full footer scan when constructing VariantStatsCaculator (#62819)
  • Avoid NPE and clear stale cache on warmup job cancel/expire (#62805)
  • Hold table write lock across first-time dynamic partition setup to prevent CREATE MV race (#62755)
  • Support specifying compute_group for StreamingJob (#62747)
  • Attach MemTracker to cache background threads to fix orphan crash (#62739)
  • Treat empty cancel alter job list as all rollup jobs (#62712)
  • Remove enable_nereids_load switch (#62703)
  • Add some Sanity-check for avoid npe (#62691)
  • RestoreJob cannot deserialize Tablet when upgrade (#62673)
  • The profile hitcache not correct in force refresh and sink id not correct (#62645)
  • Aggregate non-MOW segment key bounds to reduce rowset meta size (#62604)
  • Support ADMIN COMPACT TABLE type='FULL' and enable it in cloud mode (#62596)
  • Refine revocable memory accounting for spill (#62581)
  • Fix compaction failure on no-key table with variant column uid=0 (#62571)
  • Propagate ALTER source/target properties to runtime and persistence (#62553)
  • Avoid per-row Field temporaries in TypedZoneMapIndexWriter (#62544)
  • Fix incorrect for-loop (#62517)
  • Normalize legacy single-part dot-key subcolumn paths on read (#62409)
  • Avoid BE crash when finalize misses local cache writer (#62389)
  • Fix InsertLoadJob memory leak caused by jobs permanently stuck in PENDING state (#62282)
  • Add SC_COMPACTION_CONFLICT error code to retry cross-V1 compaction failures (#62272)
  • Add per-job metrics for streaming insert jobs (#62224)
  • Reduce compile time of aggregate_function_reader_replace.cpp (#62047)
  • Reduce HybridSet/InListPredicate template bloat and fix create_string_value_set bug (#61999)
  • Deprecate TYPE_LAMBDA_FUNCTION in PrimitiveType enum (#61941)
  • Add get_version and get_tablet_stats case (#61915)
  • Migrate old VExpr execute interface to new execute_column (Part 1) (#61912)
  • Reject upload snapshots on broken storage path (#61251)
  • Submit manual full compaction task to thread pool instead of detached thread (#61222)
  • Add fe-foundation module with zero-dependency JDK-only utilities (#61175)
  • Add show_segment_data operation to meta_tool (#60608)
  • Avoid SIGSEGV in background LRU update when clear cache (#60533)
  • Fix IOContext Use-After-Free (#59947)
  • Prioritize scheduling the most recently active tablets in cloud (#59539)
  • Support config hot reload of file cache microbench in running state (#58922)
  • Skip agg delete bitmap v1 to ms when v2 (#57990)
  • Check delete bitmap version (#57989)
  • Exposes cloud balance metrics (#57200)

Load​

  • Start counting task max interval after the first record is received (#63141)
  • Load JNI log4j2 properties config (#63063)
  • Fix broker load silently loaded only the first file when parsing multiple files path (#62969)
  • Poll packed file async close without blocking (#62938)
  • Avoid copying ANN search results (#62924)
  • Fix filteredRows always 0 on single-table S3 streaming (#62816)
  • Report physical file count in LoadStatistic.FileNumber (#62804)
  • Use CPU metrics delta for CPU pressure detection in adaptive flush controller (#62744)
  • Reject silent-no-op ALTER keys and unsupported load.* properties (#62680)
  • Roll sample window before accumulating in checkDataQuality (#62636)
  • Use per-table publication instead of ALL TABLES for PostgreSQL CDC (#62526)
  • Support specifying offset for StreamingInsertJob create and alter (#62490)
  • Fix S3 offset and job statistics lost after FE checkpoint restart (#62449)
  • Fix NPE in StreamingInsertJob.replayOnCommitted during EditLog replay (#62416)
  • Cap auto-resume attempts and expose structured FailureReason (#62345)
  • Fix INSERT job statistics lost in show load after FE restart (#62331)
  • Fix invalid String.format pattern in NereidsStreamLoadTask (#62225)
  • Update flink cdc version for streaming job (#62212)
  • Return early for non-master stream load precommit (#62109)
  • Validate stream load content length before group commit (#62110)
  • Mask sensitive headers in stream load logs (#62108)
  • Reject invalid stream load tokens on commit and rollback (#62111)
  • Using memtable memory instead of load memory to calculate adaptive write buffer size (#62104)
  • Fix GenericPool reopen blocking 18min on stale TLS connections (#61951)
  • Fix IllegalMonitorStateException in routine load afterAborted when coordinate BE restarts (#61881)
  • Reduce template instantiation in scan_operator to improve compile time (#61227)
  • Support adaptive memtable write buffer size (#56948)
  • Remove NDEBUG for groupcommit's memory tracker (#56577)

Cloud Native​

  • Preserve correct DST fold branch to go cross the transition point (#63034)
  • Clear warmup jobs on available backends (#62931)
  • Extract snapshot integration hooks (#62859)
  • Refresh event warmup backends (#62839)
  • Avoid missing result packet for query during master switch (#62721)
  • Fix redundant copy in collect_set merge (#62640)
  • Support IAM role auth in S3 filesystem (#62584)
  • Fix typo in different_serialize test data directory introduced by #59489 (#62480)
  • Fix race condition in Set operator runtime filter processing (#62434)
  • Enforce final on CRTP aggregate function derived classes to enable compiler devirtualization (#62433)
  • Support IAM Role for s3 client ut test (#62303)
  • Allocate fresh ExprId for constants when pushing project into Union (#62296)
  • Hide KV_TXN_MAYBE_COMMITTED from clients (#62244)
  • Treat JuiceFS (jfs://) as HDFS-compatible in FE/BE (#61706)
  • Merge ms and recycler http skeleton (#61502)

Lakehouse​

  • Decouple JDBC catalog metadata name (#62806)
  • Fix iceberg & maxcompute p2 case. (#62483)
  • Educe template instantiations in datetime floor/ceil functions (#61515)

Security & Authentication​

  • Restore _applied_rf_num update in late arrival path (#62872)
  • Support SSL and align MySQL CDC source with PG (#62700)
  • Fix empty workload_group value __internal_schema.audit_log (#62651)
  • Fix Ranger column-level privilege bypass when CTE combined (#61741)
  • Fix HTTP API authentication framework for admin operations (#60761)

Other​

  • Fix TIMESTAMPTZ elapsed-time semantics to use UTC (#63161)
  • Correct Arrow UTF8/String size limit (#63137)
  • Remove useless if in is_acting_on_a_slot (#63095)
  • No longer throws exceptions when parse datetime failed in from_olap_string (#63035)
  • Remove datetime transfomers check in release mode (#63003)
  • Avoid formatting generated insert errors (#62982)
  • Doris compose supports docker compose v2 (#62851)
  • Preserve sign for negative sub-hour TIMESTAMPTZ offsets (#62823)
  • Fix -Wtype-limits error in VariantStatsCaculator (#62632)
  • Fix FE startup argument forwarding (#62587)
  • Fix wrong error message of insert overwrite (#62555)
  • License check (#62534)
  • Move report logic to pipeline fragment context to remove callback parameter from ctor (#62500)
  • Extract isDynamicScheduleTable method to reduce code duplication (#62477)
  • Set parallel_pipeline_task_num explicitly in test_hive_compress_type_large_data. (#62423)
  • Remove classhisto*=trace in JAVA_OPTS_FOR_JDK_17 to prevent printing class histogram in full gc (#62422)
  • Fix RuntimeFilter selectivity sampling_frequency lost during VExprContext recreation (#62355)
  • Close snapshot readers on re-prepare and harden REST error handling (#62337)
  • Remove FixedContainer optimization for StringValueSet (#62243)
  • Require exact match in DateTimeV2Type.acceptsType (#62201)
  • Remove io_helper helper indirection from BE cast paths (#62179)
  • Reduce hash join build template instantiations (#61349)
  • Optimize the compilation time of cast. (#61276)
  • Reduce window function template instantiations (#61232)
  • Reduce the compilation time of aggregate functions. (#61179)
  • FE startup fail due to websocket startup (#60369)
  • Support max/min agg functions for type_array (#58490)
  • Fix AdminCreateClusterSnapshotCommand (#58119)
  • Add a lambda functor version for array_sort (#57828)
  • Clone snapshot support absolute path file (#57685)
  • Show the snapshot count (#56491)
  • Move watcher.stop() into locked code block (#56462)
  • Process auto partition name when its length exceeds 50 (#56304)
Report issue
ASF
Resources
Community
Join the community
Copyright © 2026 The Apache Software Foundation,Licensed under the Apache License, Version 2.0. Apache, Doris, Apache Doris, the Apache feather logo and the Apache Doris logo are trademarks of The Apache Software Foundation.