Beta 和 Experimental 功能

ClickHouse 提供 Beta 和 Experimental 功能。本文档页将介绍其定义。

由于 ClickHouse 是开源项目，除了 ClickHouse 员工之外，社区也会贡献大量内容。这些贡献的开发进度往往不尽相同；某些功能可能需要较长的原型验证阶段，或需要更多时间来获得充分的社区反馈并持续迭代，之后才能被视为正式可用 (GA) 。

由于功能在何时会被归类为正式可用并不总是明确，我们将其分为两类：Beta 和 Experimental。

Beta 功能由 ClickHouse 团队正式支持。Experimental 功能则是由 ClickHouse 团队或社区推动的早期原型，不提供正式支持。

下文将明确说明 Beta 和 Experimental 功能的特性：

​

Beta 功能

• 仍在积极开发中，目标是达到正式发布 (GA) 状态
• 可在 GitHub 上跟踪主要已知问题
• 功能后续可能会发生变化
• 可能已在 ClickHouse Cloud 中启用
• ClickHouse 团队为 Beta 功能提供支持

以下是在 ClickHouse Cloud 中被视为 Beta，且可在您的 ClickHouse Cloud 服务中使用的功能。

注意：请务必使用当前版本的 ClickHouse compatibility 设置，以便使用最近引入的功能。

​

Experimental 功能

• 可能永远不会 GA
• 可能会被移除
• 可能会引入破坏性变更
• 功能行为可能会发生变化
• 需要显式启用
• ClickHouse 团队不支持实验性功能
• 可能缺少重要功能和文档
• 无法在 Cloud 中启用

请注意：除上文列为 Beta 的功能外，ClickHouse Cloud 不允许启用任何其他实验性功能。

​

Beta 设置

名称	默认值
geotoh3_argument_order	lat_lon
enable_lightweight_update	1
allow_experimental_correlated_subqueries	1
parallel_replicas_count	0
parallel_replica_offset	0
parallel_replicas_custom_key	“
parallel_replicas_custom_key_range_lower	0
parallel_replicas_custom_key_range_upper	0
parallel_replicas_filter_pushdown	0
parallel_replicas_allow_view_over_mergetree	0
allow_experimental_database_iceberg	0
allow_experimental_database_unity_catalog	0
allow_experimental_database_glue_catalog	0
session_timezone	“
low_priority_query_wait_time_ms	1000
allow_experimental_delta_kernel_rs	1
allow_insert_into_iceberg	0
enable_join_runtime_filters	1

​

Experimental 设置

名称	默认值
query_plan_optimize_join_order_randomize	0
enable_join_transitive_predicates	0
ast_fuzzer_runs	0
ast_fuzzer_any_query	0
allow_fuzz_query_functions	0
enable_materialized_cte	0
analyzer_inline_views	0
allow_experimental_kafka_offsets_storage_in_keeper	0
allow_experimental_delta_lake_writes	0
automatic_parallel_replicas_mode	0
automatic_parallel_replicas_min_bytes_per_replica	1048576
distributed_index_analysis	0
allow_experimental_materialized_postgresql_table	0
allow_experimental_funnel_functions	0
allow_experimental_nlp_functions	0
allow_experimental_hash_functions	0
allow_experimental_time_series_table	0
allow_experimental_unique_key	0
allow_experimental_codecs	0
throw_on_unsupported_query_inside_transaction	1
wait_changes_become_visible_after_commit_mode	wait_unknown
implicit_transaction	0
grace_hash_join_initial_buckets	1
grace_hash_join_max_buckets	1024
join_to_sort_minimum_perkey_rows	40
join_to_sort_maximum_table_rows	10000
allow_experimental_join_right_table_sorting	0
allow_experimental_json_lazy_type_hints	0
allow_experimental_window_view	0
window_view_clean_interval	60
window_view_heartbeat_interval	15
wait_for_window_view_fire_signal_timeout	10
stop_refreshable_materialized_views_on_startup	0
allow_experimental_database_materialized_postgresql	0
allow_experimental_nullable_tuple_type	0
allow_experimental_query_deduplication	0
allow_experimental_database_hms_catalog	0
allow_experimental_kusto_dialect	0
allow_experimental_prql_dialect	0
allow_experimental_polyglot_dialect	0
polyglot_dialect	“
enable_adaptive_memory_spill_scheduler	0
allow_experimental_cleanup_old_data_files_compaction	0
allow_experimental_iceberg_compaction	0
allow_iceberg_remove_orphan_files	0
iceberg_orphan_files_older_than_seconds	259200
allow_experimental_expire_snapshots	0
write_full_path_in_iceberg_metadata	0
iceberg_metadata_compression_method	“
make_distributed_plan	0
distributed_plan_execute_locally	0
distributed_plan_default_shuffle_join_bucket_count	8
distributed_plan_default_reader_bucket_count	8
distributed_plan_force_exchange_kind	“
distributed_plan_max_rows_to_broadcast	20000
distributed_plan_prefer_replicas_over_workers	0
allow_experimental_ytsaurus_table_engine	0
allow_experimental_ytsaurus_table_function	0
allow_experimental_ytsaurus_dictionary_source	0
distributed_plan_force_shuffle_aggregation	0
join_runtime_filter_exact_values_limit	10000
join_runtime_bloom_filter_bytes	524288
join_runtime_bloom_filter_hash_functions	3
join_runtime_filter_pass_ratio_threshold_for_disabling	0.7
join_runtime_filter_blocks_to_skip_before_reenabling	30
join_runtime_bloom_filter_max_ratio_of_set_bits	0.7
rewrite_in_to_join	0
allow_experimental_time_series_aggregate_functions	0
promql_database	“
promql_table	“
promql_evaluation_time	auto
allow_experimental_alias_table_engine	0
allow_experimental_paimon_storage_engine	0
paimon_target_snapshot_id	-1
max_consume_snapshots	0
use_paimon_partition_pruning	0
allow_experimental_object_storage_queue_hive_partitioning	0
query_plan_optimize_join_order_algorithm	greedy
allow_experimental_database_paimon_rest_catalog	0
webassembly_udf_max_fuel	100000
webassembly_udf_max_memory	134217728
webassembly_udf_max_input_block_size	0
webassembly_udf_max_instances	32
allow_experimental_ai_functions	0
ai_function_request_timeout_sec	60
ai_function_max_retries	0
ai_function_retry_initial_delay_ms	1000
ai_function_throw_on_error	1
ai_function_max_input_tokens_per_query	1000000
ai_function_max_output_tokens_per_query	500000
ai_function_max_api_calls_per_query	0
ai_function_throw_on_quota_exceeded	1
allow_commit_order_projection	0
allow_experimental_replacing_merge_with_cleanup	0
allow_experimental_reverse_key	0
allow_remote_fs_zero_copy_replication	0
distributed_index_analysis_min_indexes_bytes_to_activate	1073741824
distributed_index_analysis_min_parts_to_activate	10
enable_replacing_merge_with_cleanup_for_min_age_to_force_merge	0
force_read_through_cache_for_merges	0
merge_selector_algorithm	Simple
merge_selector_enable_heuristic_to_lower_max_parts_to_merge_at_once	0
merge_selector_heuristic_to_lower_max_parts_to_merge_at_once_exponent	5
notify_newest_block_number	0
part_moves_between_shards_delay_seconds	30
part_moves_between_shards_enable	0
remote_fs_zero_copy_path_compatible_mode	0
remote_fs_zero_copy_zookeeper_path	/clickhouse/zero_copy
remove_rolled_back_parts_immediately	1
shared_merge_tree_virtual_parts_discovery_batch	1