This is ClickHouse roadmap 2025.

This roadmap does not cover the tasks related to infrastructure, orchestration, documentation, marketing, external integrations, drivers, etc.

See also:

Roadmap 2024: #58392
Roadmap 2023: #44767
Roadmap 2022: #32513
Roadmap 2021: #17623
Roadmap 2020: link

Data Lakes

• Automatic use of cluster functions Use cluster table functions automatically if parallel replicas are enabled #70659
• Parallel distributed INSERT SELECT by default Enable parallel distributed insert select by default #80425
• Enable Hive-style partitioning by default Enable use_hive_partitioning by default #71636
• Uniform distribution of load across cluster s3Cluster is suboptimal when the number of files is comparable to the number of servers #70190
• Partition pruning for Iceberg Iceberg Partition Pruning for time-related partition transforms #72044
• Prewhere support for Parquet [Feature] Add support for prewhere in parquet native reader #65527
• Unify different Parquet readers into one A new parquet reader that supports filter push down, which improves the total time on clickbench by 50+% compared to arrow parquet reader #70611
• Writing bloom filters for Parquet Write Parquet bloom filters #71681
• Time travel for Iceberg Add setting to query Iceberg tables as of a specific timestamp #71072
• Support for deletions in Iceberg Cannot read Iceberg table: positional and equality deletes are not supported #66588 support Iceberg equality deletes #83627
• In-memory metadata cache A cache for data lake's metadata #71579
• Integration with Delta kernel Integrate with delta_kernel #75255
• Glue catalog support Unity catalog integration #76988
• Unity catalog support Add glue catalog integration #77257
• Support for Delta Lake on Azure Support deltalake for AzureBlobStorage #74541
• Materialized views on top of data lakes
• Support arbitrary nesting in database and table names Support arbitrary nesting in database and table names. #71171
• Writing into data lakes Support for Writing to Apache Iceberg Tables in ClickHouse  #49973
• Background merges for data lakes ☁
• Subsets of a large file as a unit of work

Query Engine

• Non-constant CASE, non-constant IN Support non-constant second argument of IN #65398
• Using a partition key to optimize JOINs
• Optimization of anti-joins: LEFT JOIN ... WHERE ... IS NULL to NOT IN
• Deriving index condition from the right-hand side of INNER JOIN
• Automatic JOINs in external memory, if needed
• Optimization for certain cases of distributed JOINs
• JOIN reordering based on finer-grained statistics
• Correlated subqueries with decorrelation Analyzer: identifier resolution in parent scopes #66143 Support correlated subqueries in WHERE clause #76078
• PREWHERE to work with the FINAL clause Additional primary key scan for skip index FINAL queries #70210
• Secondary indices to work with the FINAL clause Secondary indexes should be used for queries with FINAL #70292
• Support for certain SQL functions without parentheses Allow certain functions without parentheses in SQL #52102
• Unification of table structures in Merge tables Improve usability of Merge tables #73956
• Optimizations for a single dictionary in LowCardinality [RFC] Use global LowCardinality dictionary for optimizations if it is small enough #72717
• Materialized CTE Materialized CTE #61086
• Query conditions cache Implement Query Condition Cache #69236
• Userspace page cache Userspace page cache v2 #70509
• Lazy columns Optimize performance with lazy projection to avoid reading unused columns #55518
• Block-level hints WIP Add Sorting Optimization with General Data Hints #48800
• Query cache for partial results Use query cache when a query shares most of the query pipeline. #57490
• On-disk query cache Allow using cache disk for query result cache. #52141
• Streaming queries Streaming queries model with cursors #63312

Data Storage

• JSON data type from beta to production JSON/Dynamic/Variant are production-ready #77785 See https://jsonbench.com/
• Unique key constraint RFC: Unique key constraint #70589
• Lightweight updates with patch parts Lightweight Updates with patch parts #82004
• Automatic LowCardinality columns Automatic LowCardinality representation of columns #69916
• Secondary indices by default Add minmax indices by default #72090
• ShardedMap data type Add shards to data type Map #47045
• Time data type [WIP] Implementing Time/Time64 data types #71943
• Unify statistics with secondary indices
• Production-ready full-text index Inverted Text Index v2 #86485
• Functions and codecs for quantization
• Production-ready vector search
• Allow to instantly attach backups Backup disk #64222
• Transactions for Replicated tables

Interfaces

• Random early drop on overload Reject queries when the server is overloaded #63206
• Resource scheduler by CPU CPU scheduling for workloads #77595
• Support for PromQL Basic support for the PromQL dialect #75036
• Configuration of query handlers with SQL
• Remote and Cloud database engines Remote database engine #59304
• Drivers for UDFs Drivers for User-Defined Functions #71172
• API for query construction Add combining http parameters with queries, new handler for accessing tables as files  #64336
• Even simpler data upload Even simpler data upload. #38775
• Passing input data to the url table function Support input data for reading from the URL table engine #45994
• Support for TLS in the PostgreSQL wire protocol Support TLS for Postgres wire protocol #73812
• Support for HTTP Event-Stream
• Persistent databases in clickhouse-local Persistent databases in clickhouse-local #71722
• Implicit tables in clickhouse-local clickhouse-local: if there is input, use it as a default table in SELECT instead of system.one #65023
• SSH protocol for the server SSH interface with PTY #48902
• Predictive autocomplete in the CLI Autocomplete #69641

Cleanups

• Remove old analyzer
• Remove old predicate pushdown mechanics
• Remove Object data type in favor of JSON
• Remove Window View in favor of streaming queries
• Remove Live View in favor of streaming queries
• Remove the old cache for Hive Remove obsolete cache for Hive #77795
• Remove the "send metadata" feature Remove send_metadata logic related to zero-copy replication #82508