Workflow [PR], commit [3560f7c]

Summary: ✅

Tests failures seem unrelated to the code in the PR

• 03273_ghdata_json_to_json_alter stateless test seems to be failing due to timeout
• logical error thrown in the middle of stateless tests execution seems to be related to the issue LOGICAL_ERROR: Expected JOIN table expression to be table, table function, query or union node. Actual ANY LEFT JOIN ... ON k1 = mem.k #84771
• 02443_detach_attach_partition stateless test is known to be flaky 02443_detach_attach_partition is flaky #84173
• stress test failure seems to be related to Stress test fails with Disk does not support WriteMode::Append #84669

@RuS2m ...also it would be nice to have some integration test for this client and previous server having no system.completions table, and also this version

Both test failures seem to be unrelated to the changes made in the PR:

• test_query_after_restore_db_replica[rename table-with exists table-with restart] integration test seems to be flaky due to callback in the assertion never being successful, while this raw query output compared with a string
• AST fuzzer test is most likely failing due to nonunique aliases or some kind of race condition (?) (Error "Column identifier __table2.ws_warehouse_sk is already registered").

Dear @yakov-olkhovskiy, this PR hasn't been updated for a while. You will be unassigned. Will you continue working on it? If so, please feel free to reassign yourself.

@RuS2m Can you please add a proper changelog entry, ClickHouse users will have no clue what Use system.completions table for suggestions means for them in practice.

CC: @yakov-olkhovskiy

@RuS2m Can you please add a proper changelog entry, ClickHouse users will have no clue what Use system.completions table for suggestions means for them in practice.

Just updated the entry, sorry for the churn and that the test turned out to be flaky. Potentially test_suggestions_backwards_compatibility_for_multiple_suggestions_prefix test can be dropped if it's okay with @yakov-olkhovskiy, as I presume the flakiness is coming from change in the suggestions logic, which makes regression tests on more vague autocompletion more challenging:

# In the new server suggestions logic exclude from suggestions aggregate function with "Null" combinator as well as other combinators with internal only usage
# https://github.com/ClickHouse/ClickHouse/blob/master/src/AggregateFunctions/Combinators/AggregateFunctionNull.cpp#L27

as I presume the flakiness is coming from change in the suggestions logic, which makes regression tests on more vague autocompletion more challenging

Hm, can you please elaborate? I'm not sure a I understood this. At first I was thinking that the test was flaky (here) due to it does not handle partial completion update by the client, but AFAICS it should works ok

@RuS2m this change got reverted (#87868), please resubmit it