Behavior Changes

The SHOW PARTITIONS command no longer supports Iceberg tables. Please directly use Iceberg's $partitions system table to view partitions. #56985

New Features

Added the mmh64_v2 function to generate hash results consistent with those from other third-party libraries. #57180
Added the json_hash function to generate hash values for jsonb types. #56962
Added the binary data type, along with a series of related functions: length, from_base64_binary, to_base64_binary, and sub_binary. #56648
Added the sort_json_object_keys/normalize_json_numbers_to_double function for sorting the keys of jsonb.
Added several MySQL-compatible time functions: UTC_DATE, UTC_TIME, and UTC_TIMESTAMP. #57443
Added support for MaxCompute Schema hierarchy. #56874 Documentation: https://doris.apache.org/docs/3.x/lakehouse/catalogs/maxcompute-catalog#hierarchical-mapping
The JSON_OBJECT function now supports using * as a parameter. #57256

Added support for phrase queries, wildcard queries, and regular expression queries to the SEARCH function. #57372 #57007
Extended the SEARCH function with 2 new parameters: an optional default_field parameter (default column) and a default_operator parameter (specifying the boolean operator for multi-column queries as "and" or "or"). #57312
The SEARCH function now supports searching sub-columns of the variant type, allowing direct search of specific fields in JSON paths via dot notation (e.g., variantColumn.subcolumn:keyword).
Upgraded the default storage format of the inverted index from V2 to V3. #57140
Enhanced support for custom tokenizer pipelines by adding the char_filter component; introduced two built-in tokenizers (basic tokenizer and ICU tokenizer) in the analyzer framework; added aliases for built-in tokenizers and supported component configuration with the same name to optimize the unified analyzer framework. #56243 #57055

Added the session variable merge_io_read_slice_size_bytes to address the issue of severe read amplification in external table Merge IO under certain conditions. Documentation: https://doris.apache.org/docs/3.x/lakehouse/best-practices/optimization#merge-io-optimization

Others

Optimized the size of Runtime Filter serialization information in physical plans. #57108 #56978

Fixed the issue with search query results for non-tokenized fields and enabled the execution of the search function on MOW tables. #56914 #56927
Fixed the calculation error in the inverted index when filtering with the IS NULL predicate. #56964

Fixed the issue where predicate pushdown could not use the Parquet Page Index under certain conditions. #55795
Fixed the issue of missing shard reads in external table queries under certain conditions. #57071
Fixed the issue where modifying Catalog properties did not take effect when the Hadoop file system cache was enabled under certain conditions. #57063
Fixed the issue of metadata replay failure due to connection property validation during upgrades from older versions under certain conditions. #56929
Fixed the issue of FE thread deadlock caused by Refresh Catalog under certain conditions. #56639
Fixed the issue of being unable to read Iceberg tables converted from Hive. #56918
Fixed the issue of BE crashes caused by collecting Query Profiles under certain conditions. #56806

Fixed the incorrect result of datetime type during timezone-related cast operations under boundary conditions. #57422
Fixed the incorrect precision derivation of results for some datetime-related functions. #56671
Fixed the core dump issue when inf is used as a predicate condition for float types. #57100
Fixed the core dump issue of the explode function with variable parameters. #56991
Fixed the instability issue when casting decimal256 to float types. #56848
Fixed the potential core dump issue caused by repeated scheduling during spill disk operations. #56755
Fixed the occasional incorrect adjustment of the order between mark join and other joins. #56837
Fixed the issue where some commands were not correctly forwarded to the master frontend for execution. #55185
Fixed the occasional incorrect generation of partition topn by window functions. #56622
Fixed the potential query error when synchronized mv definitions contain keywords. #57052