This release of DuckDB is named "Cyanoptera" after Spatula Cyanoptera, a species of duck found in western North and South America.

Please also refer to the announcement blog post: https://duckdb.org/2026/10/21/announcing-duckdb-200.html

What's Changed

• Make Identifier a class instead of using string everywhere by @Mytherin in #23161
• Move to tidy-check diff again by @Mytherin in #23196
• Step 3: Make duplicate eliminator optimizer logic work for CTE rewrite by @kryonix in #23187
• Allow MERGE INTO in triggers by @Dtenwolde in #23179
• RETURNING support for tables with triggers by @krleonid in #23153
• Remove special formatting code for main CMakeLists.txt by @Mytherin in #23211
• Expand aggregate state export to work with lists (StateListType) and sort keys (StateSortKey), and add aggregate state export to list, any_value/first/last, arg_min/arg_max by @Mytherin in #23199
• Initial implementation of variant_array_length processing function by @sebastiaan-dev in #23167
• Add row group count to parquet RETURN_STATS by @pdet in #23218
• Unify ProfilerPrintFormat and ExplainFormat , remove special casing and move towards pluggable tree renderers by @Mytherin in #23222
• Rework quantile to use list infrastructure by @Mytherin in #23228
• Add decimal_average aggregate function by @Dtenwolde in #23097
• Step 4: Make unnest rewriter recognize CTE-based plans by @kryonix in #23240
• Rework ordered aggregates to use the list infrastructure by @Mytherin in #23231
• Pluggable combine types rules by @yan-alex in #23173
• Auto generate select.gram by @Dtenwolde in #23236
• Unify quantile update/combine callbacks with the list aggregate by @Mytherin in #23247
• Add JSONBench to the benchmark suite by @Mytherin in #23265
• Add support for static linking C-API extensions by @Maxxen in #22251
• Add vertex_extract function to get coordinates out of point-geometries by @Maxxen in #23252
• [Dev] Resolve DEFAULTs for INSERT/UPDATE/MERGE_INTO at bind-time rather than plan-time by @Tishj in #23033
• Consistent ordering for gen files in package_build by @staticlibs in #23267
• Transform type literals into unbound logical types by @d-justen in #23237
• Remove UnaryAggregateDestructor, and instead fold it into UnaryAggregate by @Mytherin in #23268
• Add comment policy for bots by @Mytherin in #23272
• Fix null sentinel for keys in dict vector (Arrow) by @evertlammerts in #22648
• Parquet: add expression stats pruning for arithmetic filters by @Damon07 in #23264
• Add init_local_state_finalize callback to aggregate functions which allows them to create a FunctionLocalState that is persisted across Finalize calls, and use this to speed up ordered aggregates by @Mytherin in #23270
• fix out-of-bounds read in strict time parsing by @jmestwa-coder in #23266
• Add support for string_agg to export its state by making separators part of the signature by @Mytherin in #23271
• Remove redundancies in join filter creation by @d-justen in #23250
• Add setting force_update_to_del_and_insert which forces updates to always be converted to delete and inserts by @Mytherin in #23281
• [Variant] Implement SetVectorType for ShreddedVectorBuffer by @Tishj in #23280
• Commits on top of PR #20859 by @lnkuiper in #23282
• Dead node counting fixes by @artjomPlaunov in #23289
• Add support for quantile, median, mad to aggregate state export by @Mytherin in #23285
• Optimize rollup / cube / grouping sets queries by folding them into CTEs to allow for cascading aggregation by @Mytherin in #23287
• Async Parquet Writes by @lnkuiper in #23283
• Step 5: Make Mark to Semi join conversion more stable by @kryonix in #23296
• Add setting: SET dialect_compatibility_mode='spark' by @yan-alex in #23251
• Remove stale PR workflow by @DinosL in #23301
• Fix flaky AsyncFileWriter test by @lnkuiper in #23300
• [Variant] Fix crash for variant_extract on non-object shredded column by @Tishj in #23303
• Add Parquet ASYNC I/O for local files by @pdet in #23290
• Fix flaky AsyncFileWriter test by @lnkuiper in #23309
• Simplify subquery planning for provably single-row subqueries by @lnkuiper in #23306
• Add variant_comparator which ensures variants are ordered / compared in the same manner as native DuckDB values by @Mytherin in #23307
• Add missing import / fix non-unity builds by @evertlammerts in #23308
• Pass ParserOptions in default macro generation by @nicholasjng in #23288
• [Dev] Refactor MultiFileReader::GetVirtualColumnExpression and BaseFileReader::expression_map by @Tishj in #22928
• Allow slicing of shredded vectors, and fix issue with shredded vectors being fed into StructVector::GetEntries by @Mytherin in #23312
• Auto generate transform expression.gram (Part 8) by @Dtenwolde in #23310
• Add VariantIterator which allows traversal of variant values without unshredding by @Mytherin in #23317
• Remove specialized variant comparison code by @Mytherin in #23319
• Internal #9655: Test Vector DOUBLE[3] by @hawkfish in #23320
• Add export/import callbacks for aggregates with non-standard state, and support approx_top_k / approx_quantile by @Mytherin in #23304
• Avoid unnecessary call to ToUnifiedFormat in ColumnSegment filter by @Mytherin in #23324
• CMake: Set the extension base directory to the duckdb module dir by @nicholasjng in #23328
• Handle exceptions thrown while recovering from a failed commit by @Huey97 in #23298
• Use optional semantic values for generated transformer rules by @Dtenwolde in #23332
• [Parquet] Fix corrupted file invalidating the database by @Tishj in #23331
• Report in check_peg_parser parse errors by @fallintoplace in #22975
• Step 6: Flip the switch and have delim join as CTEs by default by @kryonix in #23333
• Add invoke scalar function by @feichai0017 in #23185
• Allow empty STRUCT types by @Maxxen in #23273
• Fix uninitialized TransactionContext::invalidation_policy and auto_rollback members by @evertlammerts in #23321
• Propagate shredded stats through variant_comparator by @Mytherin in #23336
• AGENTS.md: Drop in-process by @szarnyasg in #23345
• Clean-up VariantIterator API and unify it with vector iterator infrastructure by @Mytherin in #23346
• Support AFTER INSERT FOR EACH ROW triggers by @EtgarDev in #23284
• Add shredded_vector option to debug_verify_vector by @Mytherin in #23343
• Fix state export finalize with non-zero offset, and update list of skipped tests by @Mytherin in #23349
• Propagate stats in structured type -> variant casts by @Mytherin in #23347
• Deprecate Implicit Table-Function Identifier String Conversion by @kryonix in #23350
• Support submodules in extensions patching process by @staticlibs in #23191
• Nicer variable syntax by @Dtenwolde in #21194
• Remove TransformEnum by @Dtenwolde in #23355
• Quote special object keys in json_each/json_tree fullkey and path by @truffle-dev in #23206
• New debug_local_file_system_delay_ms setting and asynchronous file initialization/finalization by @lnkuiper in ...

This is a bug fix release for various issues discovered after we released v1.5.5

What's Changed

• Use relative includes in unity build sources for cache reuse by @NiclasHaderer in #23972
• bump vortex by @samansmink in #24089
• [variegata] Remove Julia client in favor of duckdb/DuckDB.jl by @evertlammerts in #24127
• [variegata] Replace the C API spec and code generation by @evertlammerts in #24131
• TopNWindowElimination Fixes by @d-justen in #24080
• Fix active query race by @NiclasHaderer in #24123
• Fix UNNEST Pushdown by @d-justen in #24119
• [Parquet][Variant] Don't produce VARIANT metadata with sorted_strings set by @Tishj in #23591
• Fix LIMIT pushdown through a volatile projection with an OFFSET by @DinosL in #24240
• Don't push filters on volatile groups through aggregates by @DinosL in #24239
• fix(adbc): return a dense union from ConnectionGetInfo by @eitsupi in #24196
• Backport 24038 by @artjomPlaunov in #24326
• Unify C-API symbol versioning for clients/extensions, stabilize all v1 API's by @Maxxen in #24362
• backport 24332 by @DinosL in #24427
• Fix #24363: Preserve NULLs in Top-N window elimination when ORDER BY expression has no column references by @sdk011 in #24399
• backport 24260 by @artjomPlaunov in #24447
• Fix ICU strptime leaking time zone state between rows by @NiclasHaderer in #24438
• Moved thread-local work outside the global lock in PhysicalInsert::Combine by @sebastiaan-dev in #24516
• Bump httpfs by @mlafeldt in #24580
• bump vortex by @samansmink in #24650
• Fix 1.5 release by @smvv in #24687
• Allow duplicates in index in ReplayRowGroupData by @artjomPlaunov in #24744
• Dispatch Go nightly build by @mlafeldt in #24720
• Fix join projection map bug in TopNWindowElimination by @d-justen in #24592
• Fix Go nightly workflow dispatch by @mlafeldt in #24802
• Fix Top-N window elimination for nullable ordering expressions by @matthiasgoergens in #24551
• Backport 23964 & 24336 by @DinosL in #24791
• Fix off-by-one in union type by @Maxxen in #24843
• Fix inner join deduplication in TopNWindowElimination by @d-justen in #24586
• fix geometry row group pruning with nulls/empty (on v1.5) by @Maxxen in #24845
• Backport DuckDB 2.0 API spellings to 1.5 by @JelteF in #24852
• Reserve column_ids.size() in CollectionScanState::Initialize by @jeewonhh in #24888
• Fix pathological scan memory for NULLs trailing an overflow string (#24897) by @mdlavin in #24918
• Use absolute unity build includes on Windows to avoid MSVC MAX_PATH errors by @NiclasHaderer in #24978
• Fix mixed path separators for extension/loader in amalgamation (v1.5 backport) by @mlafeldt in #25026
• Backport #25101: Fix crash in Top-N with LIMIT 0 (#24794) by @RagamuffinCat in #25103
• Track and expose Pipeline's intra-MetaPipeline dependencies by @jeewonhh in #24494
• Bump httpfs by @DinosL in #25202
• Skip top-n window elimination when the limit exceeds the min/max n cap by @AbhinavMir in #24741
• Add v1.5.6 storage by @carlopi in #25208
• TryScanIndex fixes by @artjomPlaunov in #25209
• Backport #22678: avoid per-row DATE->TIMESTAMP casts in comparisons by @peterboncz in #25139
• Harden temp file reads by @carlopi in #25283
• Avoid deduplicating UNION twice by @ywelsch in #25261
• Backport #24878: Always quote identifiers in error messages by @JelteF in #25214
• Fixup TryFlushCachingOperators to avoid doubly emitted chunks by @carlopi in #25331
• Close the main WAL handle before renaming over it during WAL recovery by @carlopi in #25490
• [1.5] Report the real storage version when opening a v2.0+ database file by @DinosL in #25573
• Do not assert on a scan of a row group collection without committed row groups by @bleskes in #25668
• Backport dead node counting fixes to v1.5 by @artjomPlaunov in #25693
• Fix silent truncation of very long integer literals into HUGEINT by @DinosL in #25714
• Auto-rollback failed implicitly-wrapped multi-statements on all paths by @jraymakers in #24065
• Backport #25690: Fix MAX on hive partition column after file pruning by @avakili-voleon in #25766
• Bump scanners for 1.5 by @staticlibs in #25671
• Backport #23958 (Fix TIME_NS read/write with parquet) by @avakili-voleon in #25728
• Fix deprecated CLI highlighting dot commands suggesting .render_color by @DinosL in #25689
• Backport PR #25189: Reject invalid UTF8 from printf formatting by @artjomPlaunov in #25808
• Update vortex extension by @samansmink in #25791
• Backport fix for wrong results from common_subplan with UNION ALL arms sharing a join subtree (#25819) by @Kunal8954 in #25831
• Backport #25660: Use read-only transaction to bind indexes in Checkpoint by @Flogex in #25858
• Backport 22373 by @DinosL in #25844
• Internal #10791: v1.5.6 TimeZone 2026c by @hawkfish in #25745
• Keep the delete blocks of an unchanged table in use without rewriting them by @ywelsch in #25971
• Update MySQL and ODBC scanners in 1.5 by @staticlibs in #25945
• Internal #10927: v1.5.6 TimeZone 2026d by @hawkfish in #26039
• backport 25752 by @DinosL in #26027
• Bump DuckLake for v1.5.6 by @pdet in #26009
• Backport #24557: Resolve partition columns for non-pushdown table functions by @rustyconover in #25998
• Issue #25995: Streaming Window UNIONs by @hawkfish in #26104
• Bump multiple extensions by @maiadegraaf in #26072
• Backport 35ce409 to v1.5-variegata by @carlopi in #26137
• bump iceberg by @Tmonster in #25899
• Bump: ducklake by @maiadegraaf in #26151
• Backport #24194: Add enable_optimistic_write setting by @mbrobbel in #26102
• bump delta + unity_catalog extensions for v1.5.6 by @benfleis in #26204
• Bump lance to 291316938c4ab06dc9fec5b92820f496662ef24a by @carlopi in #26203
• Bump vortex to b86c6ab223b53b445ec4c96e74f26dcb616d9c4f by @carlopi in #26202
• [Parquet] Backport of #25891 by @Tishj in #26162

New Contributors

• @AbhinavMir made their first contribution in #24741
• @mbrobbel made their first contribution in #26102

Full Changelog: v1.5.5...v1.5.6

This is a bug fix release for various issues discovered after we released v1.5.4.

What's Changed

• backport the out-of-bounds security fixes made in #23100 by @Tishj in #23197
• Remove checked_array_iterator from fmt dep (1.4) by @carlopi in #23261
• Remove checked_array_iterator from fmt dep (1.4) by @staticlibs in #23239
• [Dev] Add LLDB scripts to help with debugging by @Tishj in #23297
• Add support for test config extending by @NiclasHaderer in #23145
• bump iceberg again by @Tmonster in #23311
• Correctly get typed value stats for fully shredded variants by @Mytherin in #23325
• bump unity to include backfill fix by @benfleis in #23326
• Add request body length to http logs by @DinosL in #23316
• Show transport errors in HTTP log by @DinosL in #23327
• Dispatch Rust nightly build by @mlafeldt in #23352
• parquet: reject DATA_PAGE_V2 pages with inconsistent compressed_page_size by @benfleis in #23279
• Fix deadlock in TemporaryMemoryManager by @lnkuiper in #23351
• Register oid of dependencies in the DependencyManager so that we can track if an object was re-created with the same name by @Mytherin in #23348
• [Dev] Fix recursion issue with scripts/lldb/pointer_print by @Tishj in #23362
• [MultiFileReader] Prevent NULL MAP keys because of (likely missing) default value by @Tishj in #23354
• Propagate lambda bindings into try operator by @Mytherin in #23375
• Bump Julia to v1.5.4 by @carlopi in #23398
• Add additional guards to DICT_FSST to prevent exception during compression with small block size by @sebastiaan-dev in #23341
• fix out-of-bounds read in json path ReadKey lookahead by @jmestwa-coder in #23371
• fix out-of-bounds read in string to struct cast by @jmestwa-coder in #23384
• Issue #23383: Common Aggregate CTEs by @hawkfish in #23409
• Use correct start time when verifying dependencies by @Mytherin in #23430
• Fix false RLE corruption error by @DinosL in #23458
• [v1.5] No cache if file doesn't contain version information by @dentiny in #23434
• Improve rle corruption error messages by @DinosL in #23480
• Issue #23457: Ordered FIRST_VALUE Framing by @hawkfish in #23499
• Enabled ALP and ALP_RD for storage version v1.5.0 and up when using smaller block sizes by @sebastiaan-dev in #23483
• [v1.5] Backport PR #23444: fix eviction node memleak when external fi… by @dentiny in #23479
• [v1.5] Fix min/max aggregate stats when row groups filtered by @dentiny in #23517
• Backport #23468 and #23510 by @artjomPlaunov in #23529
• Fix arrow type extension bugs by @evertlammerts in #23534
• Issue #23500: JSON TIMESTAMP_TZ Formatting by @hawkfish in #23513
• fix out-of-bounds read in dictionary string decompression by @jmestwa-coder in #23549
• Merge v1.4 into v1.5 by @carlopi in #23560
• [ADBC] Add support for duckdb:// URI scheme in URI option by @amoeba in #21293
• Fix/capi scalar bind subquery crash by @krleonid in #23566
• [variegata] Bump extensions by @carlopi in #23562
• Fix filter combiner not pruning unsatisfiable bounds by @DinosL in #23563
• fix out-of-bounds read on empty byte array decimals by @jmestwa-coder in #23567
• [v1.5] Quickfix ALTER TABLE ADD COLUMN IF NOT EXISTS ... DEFAULT regression by @yan-alex in #23507
• Bump version map to include v1.5.5 by @carlopi in #23629
• backport #23548 to v1.5 by @myrrc in #23587
• Issue #23641: CUME_DIST Underflow by @hawkfish in #23651
• Fixed unsafe iteration when parent is NULL in string cast by @sebastiaan-dev in #23593
• Bump HTTPFS by @lnkuiper in #23648
• backport #23573 by @artjomPlaunov in #23672
• fix(adbc): support the ADBC Statistics API by @eitsupi in #23230
• Issue #23664: RANGE ZERO by @hawkfish in #23710
• Fix DROP COLUMN corrupting per-column metadata block bookkeeping by @ywelsch in #23714
• Include extension header in libduckdb archives by @mlafeldt in #23752
• Correctly promote SUGGEST_NEW to REQUIRE_NEW for variant / geometry columns by @Mytherin in #23763
• Fix lance build on Rust 1.97: patch its Cargo.lock to ethnum 1.5.3 by @carlopi in #23770
• Bump Postgres, MySQL and SQLite in 1.5 by @staticlibs in #23780
• Fix segfault in external hash aggregate when radix bits grow after going external by @carlopi in #23757
• Add HTTPUtil::ShouldRetry(request, response) retry-classification hook by @carlopi in #23793
• Bump azure, unity_catalog extensions to v1.5-variegata HEAD by @benfleis in #23800
• Bump extensions in preparation for v1.5.5 by @carlopi in #23794
• bump iceberg by @lnkuiper in #23735
• Fix for malformed JSON when rendering via duckbox by @sebastiaan-dev in #23803
• [v1.5] Backport a few fix PRs for v1.5.5 release by @dentiny in #23804
• bound alp/alprd exception positions to vector size by @jmestwa-coder in #23778
• Fix swapped min/max for multi-row-group 128-bit DECIMAL in RETURN_STATS by @sfc-gh-okalaci in #23693
• [variegata] Bump quack to 8e715ebb by @carlopi in #23820
• backport #22198 by @artjomPlaunov in #23810
• [v1.5] Backport two more fixes by @dentiny in #23841
• [DependencyManager] Fix ALTER dependency preservation by @Tishj in #23808
• TryLookupEntry now uses default schema as fallback by @okruitho in #23790
• Fix concurrent ALTER and INSERT crash by @Tishj in #23861
• Bump Postgres extension for 1.5 by @staticlibs in #23930
• [v1.5 backport] Fix: reset empty_range in the TIMESTAMP range() table function by @wmTJc9IK0Q in #23879
• Bump extensions for v1.5.5: quack, httpfs and aws by @carlopi in #23892
• Expose destructors, so they are callable module boundaries by @carlopi in #23953
• Bump DuckLake by @pdet in #23962
• Fix OSX ci by @Tmonster in #23963
• Bump lance pin to upstream ethnum fix, drop local patch by @carlopi in #23969
• Add missing statement types to capi by @hrl20 in #23407
• Bump httpfs by @lnkuiper in #24003

Full Changelog: v1.5.4...v1.5.5

This is a bug fix release for various issues discovered after we released v1.5.3.

Please also refer to the announcement blog post: https://duckdb.org/2026/06/17/announcing-duckdb-154.html

What's Changed

• backport 22094 by @artjomPlaunov in #22136
• backport 21815 & 21427 by @artjomPlaunov in #22137
• remove CleanupState destructor and remove noexcept on DuckCleanupInfo… by @artjomPlaunov in #22186
• Backport revert append fix to v1.4 by @Mytherin in #22414
• Backport windows-32bit test fix to 1.4 by @staticlibs in #22639
• Fix TemporaryFileManager reported size to reflect live blocks by @DinosL in #22767
• Set query text on PIVOT MultiStatement sub-statements at construction by @bleskes in #22769
• Fix double free and memory leak in Arrow GeoArrow CRS serialization by @rustyconover in #21854
• Render MAP values as valid SQL in Value::ToSQLString() by @carlopi in #22815
• Remove time-out waiting for terminal background color by @Mytherin in #22838
• Fix WindowSelfJoinOptimizer ignore exception by @qsliu2017 in #22800
• fix progress bar output and crash when piping SQL by @DinosL in #22836
• Internal #9438: TopN Window Projections by @hawkfish in #22851
• Fix: check if ParseFormatSpecifier returns unrecognized format by @DinosL in #22850
• Issue #22791: Window Self-Join Limits by @hawkfish in #22844
• [Backport] Column-level metadata loading and serialization by @ywelsch in #22768
• Fix json_keys with wildcard paths by @feichai0017 in #22855
• Fix: JSON add list type check by @DinosL in #22862
• Use committed row count in RowGroupReorderer by @lnkuiper in #22884
• Fix geometry stats checkpointing when no changes are detected by @Maxxen in #22882
• Check for nullptr expressions in deserialized JSON by @lnkuiper in #22906
• Add vacuum_rebuild_indexes as an (experimental) ATTACH option by @evertlammerts in #22690
• Use non-deleted row count in RowGroupReorderer by @lnkuiper in #22911
• Validate width/scale in duckdb_create_decimal_type and duckdb_create_decimal by @mlafeldt in #22905
• Bump httpfs to fix CI by @sebastiaan-dev in #22817
• Update vortex extension by @samansmink in #22930
• Fix alias propagation when replacement scan is wrapped in SubqueryRef by @truffle-dev in #22852
• Rowgroup index append by @artjomPlaunov in #22940
• Bump Julia to v1.5.3 by @carlopi in #22804
• Make url a value, not a const ref by @NiclasHaderer in #22953
• fix(adbc): support StatementExecuteSchema of ADBC 1.1.0 by @eitsupi in #22965
• Fix partial column metadata reuse bug by @ywelsch in #22994
• bump avro hash for 1.4.5 by @djouallah in #22952
• Add avro and unity_catalog to extension list by @d-justen in #22948
• Fix case-insensitive column match in INSERT ... SELECT ON CONFLICT by @remsky in #22825
• Normalize db_type to lowercase on ATTACH, apply extension aliases on compare by @carlopi in #22758
• MERGE INTO: only consider target table when binding WHEN NOT MATCHED and source table when binding WHEN NOT MATCHED BY TARGET by @Mytherin in #23014
• In the optimistic writer always start a new row group after merging by @Mytherin in #22997
• Merge Into: Avoid recursively binding in the ProjectionBinder by @Mytherin in #23022
• Rename emscripten action in v1.5 by @smvv in #23044
• Rename emscripten action in v1.4 by @smvv in #23045
• Fix VARIANT cast reading wrong rows under a filter by @mlafeldt in #23031
• fix(adbc): implement ADBC 1.1.0 Rich Error Metadata API by @eitsupi in #23073
• Add v1.4.5 to storage version map by @carlopi in #23081
• Merge v1.4-andium into v1.5-variegata, and add storage versions v1.4.5 and v1.5.4 by @carlopi in #23082
• Backport PR #22348 to andium by @carlopi in #23084
• Backport variegata bugfixes into andium by @carlopi in #23085
• Bump vortex in andium by @carlopi in #23091
• Bump lance in variegata by @carlopi in #23094
• Bump vortex in variegata by @carlopi in #23096
• Bump extensions to v1.4-andium by @carlopi in #23090
• v1.5-variegata: bump extensions by @carlopi in #23089
• Fix Parquet thrift byte order on windows + relax geometry stats pruning by @Maxxen in #23095
• fix(adbc): fill metadata of GetObjects by @eitsupi in #23110
• backport GetLocalFileSystem improvements to andium by @ccfelius in #23130
• Reject NULL json key by @dentiny in #23116
• Fix windows last modification timestamp by @dentiny in #23136
• Fix: ignore_errors silently accepting invalid json by @DinosL in #23137
• [v1.5] Backport external file cache fix by @dentiny in #23132
• Internal #9375: PRAGMA enum NULL by @hawkfish in #23146
• Clarify BIGNUM C API data is big endian by @mlafeldt in #23127
• Allow array type for array_to_json by @dentiny in #23129
• Fix json argument order affecting result by @DinosL in #23144
• Fix #21931: avoid trying to bind an expression that doesn't exist in UNPIVOT by @Mytherin in #23156
• Fix RowGroup assertion by @ywelsch in #23155
• Upgrade ducklake for release by @pdet in #23182
• Update Postgres, SQLite and ODBC in 1.5 by @staticlibs in #23172
• Fix native geometry parquet stats pruning and add OPERATOR_ROW_GROUPS_SCANNED to parquet reader by @Maxxen in #23140
• Expose bytes to parquet variant function by @Tmonster in #23057
• Minor fixes by @DinosL in #23162
• Bump aws extension by @ccfelius in #23170
• Merge v1.4-andium into v1.5-variegata by @carlopi in #23171
• Backport to variegata some isolated fixes by @carlopi in #23175
• Fix: guard againt null row group reorder stats by @DinosL in #23189
• bump iceberg by @Tmonster in #23192
• Replace ARTConflictType::TRANSACTION with fatal exception by @artjomPlaunov in #23193
• [Dev] Fix variant shredding analysis logic discrepancy with shredded writing by @Tishj in #23194
• Initialize all BaseStatistics members and zero stats_union by @krlmlr in #23202
• Initialize TransactionContext::invalidation_policy and auto_rollback by @krlmlr in #23203
• Fix selection vector use in Arrow extension callbacks by @evertlammerts in #23190
• Add hardening to many DuckDB/Parquet decompression/deserializing paths by @Tishj in #23100
• Out-of-line SelectionData destructor to silence g++-16 -Warray-bounds by @krlmlr in #23204
• Bump quack to 9ac6521f712812cc6f2e58815ef0a6c85c5e06e0 by @carlopi in #23178
• Fix crash when storage path is not set by @guillesd in #23174
• Zero the inlined buffer in string_t's length-only constructor by @krlmlr in #23201
• [Dev] Fix "environment variable already defined" error in sqllogictest when test_env is used by @Tishj in #21305
• parquet: initialize ParquetReader::rows_read by @krlmlr in #23205
• Bump quack no patches by @carlopi in #23210
• bump delta & unity for 1.5.4 by @benfleis in https://github....

This is a bug fix release for various issues discovered after we released v1.4.4.

Please also refer to the announcement blog post: https://duckdb.org/2026/06/17/announcing-duckdb-145.html

What's Changed

• Internal #7212: FILL Empty Chunk by @hawkfish in #20657
• Secret Manager: Ensure secrets are created exactly once by @Mytherin in #20686
• Use FileSystem APIs in QueryProfiler::Write by @Mytherin in #20727
• Fix escape handling in AddEscapes function by @llukito in #20760
• Fixup test/extension/update_extensions_ci.test [v1.4-andium] by @carlopi in #20777
• Backport various race condition fixes to v1.4 by @Mytherin in #20804
• bump ref to azure for v1.4.4 (rebuild) by @benfleis in #20888
• Internal #7584: AsOf Simple Joins by @hawkfish in #21090
• Do not write ccache to github cache for v1.4-andium by @smvv in #21109
• Disable extension caching for v1.4 to free github cache space by @smvv in #21150
• Disable save_cache for v1.4-andium by @smvv in #21167
• Backport #19232: allow * Not similar to 'pattern' by @Dtenwolde in #21177
• backport zstd PR to 1.4 by @artjomPlaunov in #21178
• Fix unbounded row group growth for indexed tables on repeated load+insert cycles by @abramk in #21316
• Backport #21271 by @artjomPlaunov in #21326
• [lance] Add for v1.4-andium by @carlopi in #21410
• add lance for v1.4 by @samansmink in #21113
• Correctly detect overflows when decoding integers from storage by @Mytherin in #21482
• Bump Julia to v1.4.4 on v1.4-andium by @taniabogatsch in #21589
• Fix for CSV reader buffer-boundary value read by @Mytherin in #21577
• Backport __EMSCRIPTEN__ fix by @Mytherin in #21581
• counting fix by @artjomPlaunov in #21733
• backport 22094 by @artjomPlaunov in #22136
• backport 21815 & 21427 by @artjomPlaunov in #22137
• remove CleanupState destructor and remove noexcept on DuckCleanupInfo… by @artjomPlaunov in #22186
• Backport revert append fix to v1.4 by @Mytherin in #22414
• Backport windows-32bit test fix to 1.4 by @staticlibs in #22639
• Bump httpfs to fix CI by @sebastiaan-dev in #22817
• bump avro hash for 1.4.5 by @djouallah in #22952
• Rename emscripten action in v1.4 by @smvv in #23045
• Add v1.4.5 to storage version map by @carlopi in #23081
• Backport PR #22348 to andium by @carlopi in #23084
• Backport variegata bugfixes into andium by @carlopi in #23085
• Bump vortex in andium by @carlopi in #23091
• Bump extensions to v1.4-andium by @carlopi in #23090
• backport GetLocalFileSystem improvements to andium by @ccfelius in #23130
• backport the out-of-bounds security fixes made in #23100 by @Tishj in #23197
• Remove checked_array_iterator from fmt dep (1.4) by @carlopi in #23261
• Remove checked_array_iterator from fmt dep (1.4) by @staticlibs in #23239

Full Changelog: v1.4.4...v1.4.5

This is a bugfix release for various issues discovered after we released v1.5.2.

Please also refer to the announcement blog post: https://duckdb.org/2026/05/20/announcing-duckdb-153

What's Changed

• Use DB serialization compatibility for json_serialize_sql by @kryonix in #22004
• DuckLake Bump by @pdet in #22014
• fix: resolve current catalog in ADBC Ingest to avoid temp table shadowing by @eitsupi in #22020
• Fix TIMESTAMPFORMAT being ignored for TIMESTAMPTZ columns in copy to json by @DinosL in #21992
• Provide BWC support for join filter pushdowns by @ywelsch in #22029
• Add iceberg copy function autoload by @NiclasHaderer in #22037
• Fix DISABLE_EXTENSION_LOAD by @mlafeldt in #22019
• Fix INSERT OR REPLACE BY NAME regression by excluding conflict columns from SET list by @DinosL in #22049
• Git-ignore generated extension loader by @mlafeldt in #22056
• Fix constant struct args in lateral table in-out functions by @AlfieJones in #21827
• Avoid handling Ctrl+C during shutdown (state might be already gone) by @carlopi in #22059
• Add DISABLE_BUILTIN_HTTPLIB option by @carlopi in #22054
• Set query field for statements in ALTER TABLE ... ADD COLUMN ... DEFAULT ... workaround by @yan-alex in #22057
• Coorporative tasks might lead to busy spinning in TaskExecutor::WorkOnTasks by @NiclasHaderer in #22092
• Row group append by @evertlammerts in #22060
• Fix empty parquet child schema by @dentiny in #22105
• Issue #22096: TopN Window Casts by @hawkfish in #22098
• Add support for reading VARIANT using C API by @Dtenwolde in #22065
• Bump Julia to v1.5.2 by @taniabogatsch in #22121
• fix commit iteration offset bug + relax RemoveFromIndexes assertion by @artjomPlaunov in #22094
• Internal #8812: From TIMESTAMPTZ Casts by @hawkfish in #22000
• Only build plan_serializer when building the main DuckDB library by @rustyconover in #22100
• Fix Row Group Pruner Distinct Bug by @d-justen in #22132
• Correctly skip preprocessing PIVOT MultiStatements by @yan-alex in #22141
• fix(adbc): report the table name if the table doesn't exist when appending by @eitsupi in #22146
• [v1.5 patch] Attempt to fix cache read by @dentiny in #22126
• Also execute auto-rollback on CLI ClientContext::Query() query by @yan-alex in #22159
• Fix window self join optimizer by @kryonix in #22164
• Account for ROW_GROUP_SIZE when deciding whether to append to an existing row group by @evertlammerts in #22109
• Use the latest storage version for temp storage by @Mytherin in #22169
• Fix: Add pg_catalog.pg_collation compatibility view for SQLAlchemy 2.0.45 reflection by @alonfaraj in #22160
• Fix CSV escape by @dentiny in #22176
• Fix union_by_name remap for non-nested parquet columns by @feichai0017 in #22177
• User-facing enable_caching_operators setting by @lnkuiper in #22191
• Fix bignum sum Combine to correctly take over memory ownership of state by @Mytherin in #22209
• CompressedFile::Close -> calls Close on its child_handle by @benfleis in #22149
• ISSUE-22061: Fix JSON shell output: emit BOOLEAN as true/false, not strings by @herbenderbler in #22073
• [Bugfix] Reset pg_err_pos in pg_parser_init to prevent stale error position leaking by @p1p1bear in #22239
• Bump httpfs to 3139e40a by @carlopi in #22248
• Correctly use new row group when checkpointing, and avoid incorrectly re-using metadata when targeting older storage versions and row ids have changed by @Mytherin in #22253
• Fix: release ParquetReader when a file is marked SKIPPED in multi-file scan by @J-Meyers in #22261
• Fix UTC±NN00 cannot be parsed in SQL by @tianjq16 in #22244
• Defer Bloom Filter Pushdown until it's done by @lnkuiper in #22218
• [v1.5] Backport ADBC memleak fix on error path by @dentiny in #22216
• Allow package builds to choose linked extensions by @mlafeldt in #22305
• Fix UTC+HHMM time zone was parsed incorrectly by @tianjq16 in #22297
• Bump httpfs by @carlopi in #22312
• Exception format: accept string literals by @carlopi in #22314
• Skip schema analysis even if no shredding for rowgroup by @rgernhardt in #21937
• Use batch limit for table scans with filters by @Mytherin in #22315
• Also redacting bearer token for HTTP secrets by @hannes in #22323
• First initialize system, then load extensions (to peek at file to be opened) by @carlopi in #22341
• Node Handle Scoping fix by @artjomPlaunov in #22344
• Enable windows_amd64 for lance extension by @carlopi in #22367
• Internal #9003: TIMETZ Parsing Limit by @hawkfish in #22378
• Downcasting decimal fix incorrect out of range error by @Dtenwolde in #22386
• GetLocalFileSystem improvements by @carlopi in #21983
• Fix double decrement of evicted_data_per_tag in .block read-back by @krleonid in #22394
• RowGroup Operator metrics: sequentially scanned row groups + total row groups + cumulative counterparts by @artjomPlaunov in #22339
• PostgreSQL compatability: in pg_catalog.pg_database simulate columns datallowconn, datistemplate by @muayyad-alsadi in #22302
• Fix eviction size metrics report by @dentiny in #22452
• Fix enable_logging() silently resetting logging_storage by @guillesd in #22475
• Avoid 3 instances of idx_t - idx_t > 0, and avoid unnecessary check on zLine by @carlopi in #22518
• Fix RESET my_global_extension_setting to actually be GLOBAL by @carlopi in #22520
• Fix BlockAllocator invalid memory access by @dentiny in #22503
• Patch httplib by making ThreadPool constructor more solid on pthread_create failures by @carlopi in #22516
• Fix duck fuzz #4430 by @dentiny in #22435
• Test runner: avoid running clean-up routine if there is no database to run it in by @Mytherin in #22540
• Fix bare numeric interval parsing at end of string on v1.5 by @M1llerF in #22534
• Fix parquet metadata cache validation by @dentiny in #22547
• ExtensionInstall: Remove use of IsHTTP to IsRemoteFile by @carlopi in #21900
• Fix variant write small decimal by @dentiny in #22544
• Move http_proxy setting to global setting, and use GetEnvVariable('HTTP_PROXY') as default by @carlopi in #22541
• Bump httpfs and remove patches by @carlopi in #22556
• Bump Postgres, MySQL and ODBC by @staticlibs in #22579
• Bump avro, azure, delta, ducklake, spatial, unity_catalog and vortex by @carlopi in #22554
• Fix variant selection vector index by @dentiny in #22573
• Fix for SIGABRT in setting size on zero-capacity vector by @ccfelius in #22571
• Fix incorrect profiling results when using LIMIT by @sebastiaan-dev in #22561
• Move Jemalloc into core by @Maxxen in #22558
• Allow json ts format variation across columns - issue 22103 by @benfleis in #22559
• Limit parallel linker jobs to avoid out-of-memory errors by @smvv in #22588
• ARTOperator::Delete return false if rowid not found in nested ART leaf by @artjomPlaunov in #22591
• Enable/disable jemalloc linking through BUILD/SKIP_EXTENSIONS by @Maxxen in https://gi...

This is a bug fix release for various issues discovered after we released v1.5.1

What's Changed

• Backport various race condition fixes to v1.4 by @Mytherin in #20804
• bump ref to azure for v1.4.4 (rebuild) by @benfleis in #20888
• Internal #7584: AsOf Simple Joins by @hawkfish in #21090
• Do not write ccache to github cache for v1.4-andium by @smvv in #21109
• Disable extension caching for v1.4 to free github cache space by @smvv in #21150
• Disable save_cache for v1.4-andium by @smvv in #21167
• Backport #19232: allow * Not similar to 'pattern' by @Dtenwolde in #21177
• backport zstd PR to 1.4 by @artjomPlaunov in #21178
• Fix unbounded row group growth for indexed tables on repeated load+insert cycles by @abramk in #21316
• Backport #21271 by @artjomPlaunov in #21326
• [lance] Add for v1.4-andium by @carlopi in #21410
• add lance for v1.4 by @samansmink in #21113
• Correctly detect overflows when decoding integers from storage by @Mytherin in #21482
• Correctly revert dictionary size when reverting string appends by @Mytherin in #21489
• Add descriptions for vortex and lance by @carlopi in #21500
• Fixup Write bytes are counted as BYTES_WRITTEN by @carlopi in #21501
• Fix memory leak when reusing PreparedStatement (#21089) by @wzharies in #21104
• Make some MultiStatements and PRAGMAs Transactional by @yan-alex in #21171
• Reduce concurrent thread count in test by @Mytherin in #21511
• Fixing integer overflow in list_resize by @hannes in #21515
• Fix #21512: correctly render empty results in .mode json by @Mytherin in #21517
• Fix parsing test path to skip by @jeewonhh in #21495
• Add setting for limiting the number of threads launched concurrently in the test runner (max_test_threads) by @Mytherin in #21520
• Unify adding suffixes to path in Path:: AddSuffixToPath - fix temp directory split bug by @Mytherin in #21527
• Correctly deal with negative values in GetPosixVersionTag, and fix constant NULL struct scans after recent fix by @Mytherin in #21549
• Fix missing extension static libs in Windows MinGW bundle by @mlafeldt in #21559
• PEG grammar fixes: Update extension and allow numeric struct keys by @Dtenwolde in #21331
• [v1.5-variegata] Fix #21514: ASOF join empty right by @Pieczasz in #21553
• Fix data path in test/sql/copy/parquet/parquet_no_stats.test by @jeewonhh in #21561
• TopNWindowElimination Column Binding Fix by @d-justen in #21564
• lance: bump lance-duckdb to 4d9ecab by @Xuanwo in #21572
• Bump Julia to v1.5.0 by @taniabogatsch in #21588
• Bump Julia to v1.4.4 on v1.4-andium by @taniabogatsch in #21589
• Simplify the way we determine which row groups to checkpoint during checkpoints by @Mytherin in #21574
• Fix for CSV reader buffer-boundary value read by @Mytherin in #21577
• CLI: Avoid division by zero when formatting a large result with a non-wide shell by @Mytherin in #21591
• Backport __EMSCRIPTEN__ fix by @Mytherin in #21581
• Fix missing SetSizeAndFinalize in BIGNUM Add for zero result case by @pdet in #21465
• Segfault due to unchecked malloc/realloc, proposed fix to #21593 by @wasade in #21594
• Reduce arg_min_max_n heap preallocation by @feichai0017 in #21467
• Fixed an issue where the describe statement did not work correctly in markdown output mode(ISSUE:#21579) by @ArNine in #21611
• fix unpivot serialization by @Maxxen in #21595
• Windows shell: enable VT100 processing on startup by @staticlibs in #21615
• Make PEG Parser use strict mode in CI by @Dtenwolde in #21590
• Fix #21623: flatten input chunk in TopNHeap::CheckBoundaryValues by @Mytherin in #21629
• Issue #21592: Window Self-Join Framing by @hawkfish in #21628
• Bump Julia to v1.5.1 by @taniabogatsch in #21637
• Merge v1.4-andium into v1.5-variegata by @Mytherin in #21639
• Bump storage version to v1.5.2 by @taniabogatsch in #21638
• bump delta and unity_catalog ext refs in v1.5-variegata by @benfleis in #21640
• Fix cancellation order between pipelines and tasks in CancelTasks by @carlopi in #21642
• Fix shell completion enter handling by @atulagrwl in #21552
• fix(adbc): err use after free by @gishor in #21605
• Fix some Parquet fuzzer issues by @lnkuiper in #21635
• Re-organize WAL replay slightly, and correctly deal with empty checkpoint WAL files in WAL recovery by @Mytherin in #21645
• Warn instead of error when trying to persist geometry columns with CRS in old storage format by @Maxxen in #21649
• Avoid throwing an error when failing to bind views in duckdb_columns by @Mytherin in #21658
• Infer timestamps with timezone in read_json_auto by @leo-altertable in #21660
• Windows: remove prefix from canonical paths by @staticlibs in #21652
• CLI: Add .help shortcuts by @carlopi in #21662
• Internal #8553: Window TopN Except by @hawkfish in #21671
• Fix issue with struct filter on missing structs by @Mytherin in #21676
• Internal #7568: ASOF SEMI Test by @hawkfish in #21683
• Fix type check in st_crs by @Maxxen in #21688
• Disable regular updates for geometry by @Maxxen in #21641
• Fix stoi crash in Arrow format string parsing for w: and +w: types by @yharby in #21692
• Test runner: Support replacement without dollar ({i} instead of ${i}) in loop iterators by @Mytherin in #21708
• Internal #8657: IEJoin Filter Sides by @hawkfish in #21721
• Add missing test for delta byte array by @ccfelius in #21714
• Fix variant shredding consistency issue by @Tishj in #21715
• TopNWindowElimination fixes by @d-justen in #21663
• Ignore NULL/__HIVE_DEFAULT_PARTITION__ when detecting types by @lnkuiper in #21731
• fix path test warnings by @benfleis in #21711
• Allow SET DEFAULT / DROP DEFAULT for tables that have dependencies by @Mytherin in #21729
• Add clickbench by @c-herrewijn in #21730
• Fix prepared temp-table INSERT invalidation after DROP by @wordhardqi in #21712
• Allow join filter pushdown through integral up/down casts by @lnkuiper in #21743
• Allow join filter pushdown for NOP collations by @lnkuiper in #21742
• counting fix by @artjomPlaunov in #21733
• [Variant] Re-add the removed variant_legacy_encoding setting by @Tishj in #21710
• Fix update plans when deserializing if type no longer supports regular updates by @Maxxen in #21718
• Re-instantiate dependencies of tables for ALTER TABLE ... DROP COLUMN and ALTER TABLE .. SET DEFAULT by @Mytherin in #21752
• Use correct error message for name conflicts between table and views by @JelteF in #21760
• Bump spatial by @taniabogatsch in #21781
• PEG parser strict mode: followup fixes and improvements by @Dtenwolde in #21709
• Fix DELETE RETURNING for rows inserted in the same transaction by @rustyconover in #21541
• Disable bloom filter pushdown through casts by @lnkuiper in #21792...

This is a bug fix release for various issues discovered after we released v1.5.0.

Please also refer to the announcement blog post: https://duckdb.org/2026/03/23/announcing-duckdb-151

What's Changed

• [Extensions] Preserve require order in InterpretedBenchmark by @Tishj in #21214
• CLI: Make help entry uppercase by @szarnyasg in #21224
• Restore view's bind_state when binding fails by @jeewonhh in #21193
• [Regression v1.5] Fix INTERNAL Error in UnnestRewriter for deeply nested struct UNNEST by @guizmaii in #21209
• fix(storage): prevent TrimFreeBlocks from zeroing concurrently alloca… by @llersch in #21146
• Windows shell: UTF-8 mode when writing to pager by @staticlibs in #21202
• [Dev] Add internal exception for misuse of MultiFileReader method by @Tishj in #21046
• Before loading libraries in TryAutoLoadExtension check AutoloadKnownExtensions Setting by @pdet in #21051
• fix: buffer overread in Arrow dictionary conversion with NULLs - fixes crash. by @rustyconover in #21083
• Avoid misleading GetRequestInfo constructor that takes endpoint + path by @carlopi in #21151
• Enable column pruning for MATERIALIZED CTEs by @kryonix in #21169
• Add framework to test two DuckDB versions by @Y-- in #21184
• Internal #7691: 2026a Time Zones by @hawkfish in #21204
• Fix DuckFuzz #4325 by @kryonix in #21215
• Fix decorrelation delim index bug by @kryonix in #21233
• Fix two minor shell bugs: Avoid mutating rendered strings and avoid p… by @carlopi in #21234
• Fix wrong cast in RecursiveDependentJoinPlanner by @kryonix in #21235
• Use atomic load in bloom filter look-up by @Mytherin in #21238
• Move BWC tests to nightly for now by @Mytherin in #21266
• Internal #7740: AsOf Tidy Fix by @hawkfish in #21255
• Re-add accidentally removed -jsonlines command line parameter by @Mytherin in #21263
• Switch MAIN_BRANCH_VERSIONING to False for v1.5-variegata branch by @carlopi in #21239
• Lazy call mmap in BlockAllocator by @lnkuiper in #21276
• Dynamically determine radix bits for external aggregation by @lnkuiper in #21274
• Avoid potential division by 0 in PhysicalHashJoin::PrepareFinalize by @lnkuiper in #21271
• Do not use MSYS shell for MinGW builds by @staticlibs in #21296
• add structured Path objects, update joinpath: normalize, support url/… by @benfleis in #21277
• When stress testing FORCE_ASYNC, add the relevant error messages by @carlopi in #21267
• Fix invalid JSON when casting from certain types by @Maxxen in #21280
• Fix .open of Parquet files etc. in CLI by @hannes in #21269
• Bump storage version to v1.5.1 by @taniabogatsch in #21287
• Implement line number reporting in errors for PEG parser by @Dtenwolde in #21157
• Support new v1.5 syntax for PEG grammar by @Dtenwolde in #21172
• [Fix] Memory error when transforming to v1.0.0 ART storage by @taniabogatsch in #21270
• Add missing expression traversal in merge into binding by @kryonix in #21283
• Add UUID to settings cache by @Mytherin in #21290
• Fix Terminal::TryGetBackgroundColor when stdin is reading from a pipe but stdout is still writing to the terminal by @Mytherin in #21288
• Fix #21248: correctly expand type count in CSV header detection by @Mytherin in #21292
• Add batch index support to parquet_metadata and friends to allow it to be executed in parallel also with order preservation by @Mytherin in #21314
• Make overriding HTTPUtil thread safe by @Mytherin in #21210
• Revert #21067 - and fix WAL corruption issue through calling MarkBlockAsCheckpointed on WAL blocks by @Mytherin in #21285
• Fix sampling for UB in musl libc by @carlopi in #21318
• TryGetCurrentSetting: Fixup paramater passing for FileOpenerInfo by @carlopi in #21301
• Fix Row Group Reorderer Bug by @d-justen in #21282
• Windows shell: fix UTF-8 to UTF-16 conversion by @staticlibs in #21319
• Issue #21244: AsOf Unordered LIMIT by @hawkfish in #21250
• Throw error on incorrect use of ParseExpressionList by @Dtenwolde in #21306
• Issue #20481: Prepare Window Parameters by @hawkfish in #21323
• CLI: Explicitly clean up shell state instead of letting static destructor order determine when it gets destroyed by @Mytherin in #21315
• Fixup 401bcf0: "When stress testing … by @carlopi in #21329
• CLI: Make bail on error the default in more situations by @Mytherin in #21344
• Improve estimate when first Parquet file is almost empty, and get exact cardinality count when we are using union by name by @Mytherin in #21345
• Fix infinite recursion in FileOpener::TryGetCurrentSetting by @Mytherin in #21356
• fix argument check for .multiline and .singleline by @jaywgraves in #21336
• Windows shell: remove .utf8 dot command by @staticlibs in #21343
• When writing unsupported Parquet variant types to Parquet, try to convert them to INT64 by @Mytherin in #21357
• Use cached Parquet metadata to improve cardinality estimates over Parquet files by @Mytherin in #21358
• Parquet Reader: Allow merging of prefetch column ranges for columns that do not have table filters by @Mytherin in #21373
• In the parquet reader, if we are globbing over a directory, use file sizes from the glob to estimate cardinalities instead of relying only on the first parquet file by @Mytherin in #21374
• MbedTLS: Actually throw exception by @hannes in #21365
• Well-defined environment for MinGW builds by @staticlibs in #21371
• Fix #21335: correctly add checkpoint and recovery WAL to allowed paths when launching an initial db with enable_external_access set to false by @Mytherin in #21379
• automate clangd/compile_commands.json gen via builds by @benfleis in #21361
• Parquet: Ignore optional filters when deciding whether or not to do a lazy fetch by @Mytherin in #21383
• parquet: avoid corrupting define buffers during skips by @Mikubill in #21298
• CLI: Fix .tables rendering for large database names by @Mytherin in #21389
• Fix column pruning for CTEs by @kryonix in #21275
• add test for #21271 by @artjomPlaunov in #21398
• Grab correct row group id in ValidityColumnData::UpdateWithBase by @Mytherin in #21397
• Add lance extension by @carlopi in #21406
• add lance for v1.5 by @samansmink in #21111
• Fix #21289: correctly determine render width of UTF8 string with grapheme clusters by @Mytherin in #21409
• Avoid calling PushCollation in create_sort_key by @Mytherin in #21412
• Flush remaining operators in pipeline by @DinosL in #21405
• Correctly limit auto-detection in read_json for files that don't have string columns by @Mytherin in #21411
• Fix INSERT OR REPLACE updates on non-unique indexed columns by @feichai0017 in #20962
• fix(adbc): support concurrent statements on the same connection by @eitsupi in #21415
• Fix invalid common subplan CTE reuse for issue #21372 by @feichai0017 in #21386
• Use regular pushdown path for semi/anti/left delim joins by @Mytherin in #21416
• Predicate factoring by @lnkuiper in #21418
• Bump SQLite by @staticlibs in https://github.com/duckdb/duckdb/pull...

This release of DuckDB is named "Variegata" after the paradise shelduck Tadorna Variegata, also known as the paradise duck, a species of Shelduck (a group of goose-like ducks), endemic to New Zealand.

Please also refer to the announcement blog post: https://duckdb.org/2026/03/09/announcing-duckdb-150.html

What's Changed

• Optimize LAST aggregate to iterate backward within batches by @xe-nvdk
• Allow user type with schema defined to have array bounds by @Dtenwolde in #18342
• Remove case-insensitive character from grammar files by @Dtenwolde in #18948
• Fix query() function error message for PIVOT statements without explicit IN clauses by @shivampr in #18900
• Merge v1.4-andium into main by @c-herrewijn in #18952
• [Dev] Some light code cleanup in RemoveUnusedColumns::VisitOperator by @Tishj in #18964
• Support renaming a database - ALTER DATABASE <> RENAME TO <> by @abramk in #18970
• V1.4 andium by @Mytherin in #18982
• Fix field id uniqueness for alter database by @abramk in #18988
• Fix unused variable warnings by @mlafeldt in #18913
• Avoid adding a TupleDataChunkPart with only one row. by @lance5401 in #18904
• Tiny fix for ColumnDataAllocator::ColumnDataAllocator by @baolinhuang in #18990
• Fixes for ALTER DATABASE - use serialization framework, and fix stringification of identifiers by @Mytherin in #18994
• Expected errors 2053 by @hmeriann in #18960
• Fix union semantics in recursive CTEs by @cryoEncryp in #19017
• Multifile parquet_metadata and friends parallelism by @J-Meyers in #18854
• [Optimize] use StorageLock for CompressionFunction by @baolinhuang in #19028
• PEG grammar updates by @Dtenwolde in #19031
• Rework Compression Function registration to avoid grabbing locks by @Mytherin in #19045
• Merge v1.4 into main by @Mytherin in #19048
• [C-API] default to adding overload on conflict when registering functions by @Maxxen in #19036
• Destroy row groups in parallel by @Mytherin in #19053
• Switching core extension upload to dedicated credentials by @hannes in #19060
• CI: Increase stale bot timeout to 1 year by @szarnyasg in #19000
• Revert "set default value of MAIN_BRANCH_VERSIONING to false" by @c-herrewijn in #19057
• Fix extension load by @dentiny in #19079
• Change core extensions bucket name by @hannes in #19082
• Merge v1.4 into main by @Mytherin in #19098
• Allow directory override for extension build by @Mytherin in #19110
• Skip serialization of row_start in DataPointer when targeting latest storage by @Mytherin in #19111
• [C-API] Initial support for accessing file system by @Maxxen in #19086
• Rework UNION (ALL) operators (SetOperationNode/LogicalSetOperation/PhysicalUnion) to have multiple children by @Mytherin in #19109
• pushdown filter when doing cross or lateral join with unnest by @DinosL in #19085
• Fix unholy interaction between CREATE VIEW and CTE definitions by @kryonix in #19116
• Add support for read_duckdb - a function that allows reading / globbing DuckDB database files by @Mytherin in #19108
• Fix typo by @szarnyasg in #19130
• fix: check tree depth and flatten dep join for custom operator extension by @Wal8800 in #19094
• During optimistic writing, gather N row groups and write them at the same time column-at-a-time by @Mytherin in #19118
• Remove unused code by @szarnyasg in #19134
• Add support for late materialization and filter pushdown to read_duckdb by @Mytherin in #19145
• Move storage block prefetching to an option instead of baking in testing during ALTERNATIVE_VERIFY + add a test config for this option by @Mytherin in #19146
• Introduce parser_override_function_t by @Dtenwolde in #19126
• Perform deduplication based on join_keys when building ht by @tianjq16 in #19097
• Print help to stdout by @staticlibs in #19170
• Validate filesystem registration by @dentiny in #19168
• Fix missing replacement scan code for CTEs by @kryonix in #19159
• Correctly set row start in WAL replay by @Mytherin in #19192
• Use a PartialBlockManager per column, instead of sharing one over the entire row group by @Mytherin in #19194
• Internal #5380: AsOf Join Pipeline by @hawkfish in #19169
• Cross and lateral join pushdown with unnest or json - update by @DinosL in #19152
• Missing error messages for test cases by @hmeriann in #19129
• [Profiling] Move EXTRA_INFO into the metrics map by @maiadegraaf in #19124
• Common Subplan Elimination by @lnkuiper in #19080
• GEOMETRY Rework: Part 1 - Logical Type by @Maxxen in #19136
• [Parquet] Implement (unshredded) VARIANT support for COPY TO parquet by @Tishj in #19125
• Add more missing error messages to test cases by @hmeriann in #19210
• [CI] - Fix failing Extension updating test NightlyTests job by cleaning up some space on the runner by @hmeriann in #19214
• Inconsistent semicolon handling by @Dtenwolde in #19199
• Internal #5383: Range Join Sorting by @hawkfish in #19174
• [Profiling] Add QueryContext to more functions, for Reads by @maiadegraaf in #18726
• Merge v1.4-andium into main by @Maxxen in #19225
• [shell] Fix tilde parsing for .once statement by @dentiny in #19254
• Partially remove BoundQueryNode - have Bind directly emit a BoundStatement by @Mytherin in #19289
• Binder clean-up by @Mytherin in #19293
• test: add test for CREATE TABLE with my_schema.my_enum[] by @NickCrews in #19296
• Fix uncaught exception in test_empty_profiling_settings.test by @maiadegraaf in #19286
• GEOMETRY Rework: Part 2 - Statistics by @Maxxen in #19203
• Internal #5380: AsOf Sorting Conversion by @hawkfish in #19278
• Add option for opt-in to parser override by @Dtenwolde in #19181
• Remove BoundTableRef and make binding a TableRef directly return a BoundStatement by @Mytherin in #19294
• Rewrite Grouped Top-N Window Functions To Aggregations by @d-justen in #19280
• Perform optimistic reads while filling gaps in CachingFileSystem by @carlopi in #19122
• Issue #19027: AsOf Join Predicates by @hawkfish in #19316
• Enable experimental metadata re-use by default by @Mytherin in #19317
• Binder: BoundQueryNode cleanup by @Mytherin in #19322
• Remove old sort code by @lnkuiper in #19321
• [Parquet] Implement shredded VARIANT support for COPY TO parquet by @Tishj in #19219
• Merge andium into main by @Tishj in #19325
• [C API] Expose column count and type for table description by @taniabogatsch in #19334
• Relocate the python sqllogic code to the dedicated repository by @Tishj in #19340
• Binder: Rework the way set operations work so that we can use the standard binding flow by @Mytherin in #19342
• Rework CTE binding: remove CTENode, and bind CommonTableExpressionMap directly instead by @Mytherin in #19351
• Avoid keeping shared ptr to ClientContext in QueryContext by @Mytherin in #19353
• Internal #6190: AsOf Threading by @hawkfish in #19328
• [Indexes] Buffer Managed Indexes Part 4: Segment Handles f...

This is a bug fix release for various issues discovered after we released 1.4.3. Please excuse the erroneous release earlier.

What's Changed

• Cast Fix: Correctly handle negative exponent with a number with a decimal in VARCHAR -> INTEGER cast by @Mytherin in #20098
• [Storage] Fix NULL filter check for constant segments by @Tishj in #20103
• fixing staged upload for install.duckdb.org - hopefully by @hannes in #20104
• Remove undefined loop by @artjomPlaunov in #20105
• [Test] Adjust concurrent attach-detach test to expect write-write conflict by @taniabogatsch in #20108
• Fix ALIAS function in filter pushdown to preserve column aliases(#20008) by @henry8128 in #20106
• Headers missing from http logs by @samansmink in #20030
• Bumping httpfs to include duckdb/duckdb-httpfs#174 by @carlopi in #20145
• Issue #20136: Secondary IGNORE NULLS by @hawkfish in #20153
• Fix minio nightly tests by @c-herrewijn in #20132
• Fix use-after-free in mode aggregate Combine function by @victor-ab in #20146
• Only pushdown varchar if the arrow type is not a string view by @evertlammerts in #20165
• Quote filters in adbc_get_objects by @evertlammerts in #20172
• Issue #20156: Streaming Window Unions by @hawkfish in #20191
• MergeInto: correctly clean-up buffer and handle non trivial GetData by @carlopi in #20163
• Fix extentension-ci-tools to latest release version in the v1.4-andium branch by @carlopi in #20227
• [Chore] Clean up CI nightly run to prevent time outs by @taniabogatsch in #20150
• Bump Julia to v1.4.3 by @maiadegraaf in #20248
• Internal #6881: 2025c Time Zones by @hawkfish in #20258
• Split statements by semicolon by @Dtenwolde in #20174
• [chore] Remove numeric_cast.hpp import and other tidy stuff by @taniabogatsch in #20278
• [chore] Reduce latency of VARCHAR index creation test by @taniabogatsch in #20277
• Reuse correct table-level metadata during checkpoints by @ywelsch in #20267
• Fix view resolution not being stable if the referenced table lives in a different schema by @Tishj in #20260
• ConstantOrNullFunction input validity mask overwrite bugfix by @artjomPlaunov in #20283
• dbgen: use TaskExecutor framework by @Mytherin in #20284
• fix(adbc): return error when setting an empty sql query by @gishor in #20071
• Bump iceberg, and add wasm platforms! by @carlopi in #20205
• _extension_distribution: Pin to v1.4-andium branch of extension-ci-tools by @carlopi in #20294
• Optimize prepared statement parameter lookups by @EtgarDev in #20252
• Update VectorType in ComputePartitionIndices by @lnkuiper in #20343
• Add some defensive programming in RadixPartitionedHashTable::Combine and RadixPartitionedHashTable::Finalize by @lnkuiper in #20342
• Increase reserved size for paths in SetPathsInternal by @Flogex in #20340
• Use UTF-16 console output in Windows shell (1.4) by @staticlibs in #20339
• Force repartitioning in RadixPartitionedHashTable::Combine by @lnkuiper in #20357
• Fixup comparison to wrong iterator on abdc's driver by @carlopi in #20360
• [chore] dsdgen generation signed integer overflow fix by @taniabogatsch in #20279
• Give preference do variables set in config by @pdet in #20396
• Add QueryContext also to Write(void *buffer, idx_t nr_bytes), so that BytesWritten are updated also there by @carlopi in #20393
• Fix Issue #20233: fix function chain in qualify by @ArNine in #20302
• Backport client data cleanup by @taniabogatsch in #20403
• Add v1.4.4 to Storage Version by @maiadegraaf in #20404
• Parquet Reader: Ignore invalid UTF8 in string stats, instead of throwing an error by @Mytherin in #20405
• Don't add a semicolon to final query when splitting statements by @Dtenwolde in #20401
• Fixup BRANCHES_TO_BE_CACHED, vars are not available on PRs, so env it is by @carlopi in #20411
• [Fix] Defensive infinite loop guard and UTF-8 check in C API by @taniabogatsch in #20348
• Add ducklake, httpfs and iceberg tests so they are run in CI by @carlopi in #20319
• [Stats] date_trunc stat propagation fix by @Tishj in #20421
• Refactor allowed path sanitize by @hannes in #20346
• Issue #20413: ASOF SEMI/ANTI Bindings by @hawkfish in #20433
• Fix #20410: fix for RIGHT SEMI/ANTI - cannot fully label chain as found if there are non-equality predicates present in the join condition by @Mytherin in #20435
• [Fix] Misaligned size in ART prefix count by @taniabogatsch in #20455
• Wrap ccache in own action by @carlopi in #20466
• [Chore] Add ORDER BY to AS OF test by @taniabogatsch in #20465
• Fix segfault in hive partitioning with NULL values by @Schwarf in #20468
• Nightly test encryption fixes by @ccfelius in #20461
• [C API] Fix error data creation by @taniabogatsch in #20451
• Skip concurrent_encrypted_attach due to race in autoloading httpfs by @carlopi in #20471
• CI fixup: Comparisons needs to be done via strings (since input via workflow_call is a string) by @carlopi in #20464
• Add unity_catalog to extensions built by duckdb/duckdb by @carlopi in #20445
• Invert the setup of ccache and cleanup_runner by @carlopi in #20429
• bump spatial for andium by @Maxxen in #20479
• Reclaim disk space on MacOS runners by @lnkuiper in #20493
• Adding secure clear functions by @ccfelius in #20285
• rewrite unaligned scan by @artjomPlaunov in #20474
• Issue #20470: TIMESTAMPTZ to DATE by @hawkfish in #20498
• Reset cached dictionaries in TRY expression by @Maxxen in #20452
• [Copy] Fix #20324 partition_by option binding by @Tishj in #20509
• Bump multiple extensions by @maiadegraaf in #20504
• update azure ref for v1.4-andium by @benfleis in #20506
• [Fix] Directly retrieve the logical column index during MERGE INTO binding by @taniabogatsch in #20503
• Sanitize pragmas by @samansmink in #20514
• Bump httpfs, ducklake and postgres by @carlopi in #20518
• Revert vortex bump for v1.4.4 by @taniabogatsch in #20527
• Bump VSS by @taniabogatsch in #20542
• Bump Excel by @maiadegraaf in #20540
• Fix Issue #20233: Fix function chain in having and merge to v1.4 by @ArNine in #20532
• bump iceberg by @Tmonster in #20549
• Bump Iceberg v1.4 by @Tmonster in #20604
• Fix KeyValueSecretReader init by @NiclasHaderer in #20620
• Fail fast extensions by @carlopi in #20605

New Contributors

• @victor-ab made their first contribution in #20146
• @gishor made their first contribution in #20071

Full Changelog: v1.4.3...v1.4.4