Workflow [PR], commit [250b67b]

Summary: ✅

• Performance Comparison: Performance dashboard

AI Review

Summary

The PR fixes parsing of DESCRIBE TABLE (SELECT ...) AS alias by routing lone parenthesized SELECT sources through ParserTableExpression, while preserving the existing DESCRIBE (SELECT ...) UNION/EXCEPT/INTERSECT ... path. The current diff addresses the previously raised set-operation regressions and adds focused stateless coverage for the reported alias case, bare DESCRIBE SELECT, top-level set operations, and alias-bearing subqueries whose body contains a set operation.

Final Verdict

Status: ✅ Approve

The failures of 00175_obfuscator_schema_inference and 00096_obfuscator_save_load in Stateless tests (amd_tsan, parallel) are NOT caused by this PR.

They were introduced by #104690 ("Add UntrackedMemory asynchronous metric"), which made clickhouse-obfuscator abort (SIGABRT) on process teardown: the query results are correct, but the Aborted line on stderr fails the test. #104690 has now been reverted (#106365).

#104690 was merged in violation of the ClickHouse team rules: its own CI already showed these two tests failing (10 times between May 12 and June 1) before it was merged.

Please update your branch to pick up the revert; the tests should pass again.

Picked up this PR, merged the latest master, and addressed the AI Review blocker.

AI Review blocker (fixed in 0915825): the previous guard pos->type != TokenType::OpeningRoundBracket routed every parenthesized DESCRIBE source through ParserTableExpression, which consumes only the first (SELECT ...) and leaves a following set operator unparsed — regressing valid queries like DESCRIBE (SELECT 1) UNION ALL (SELECT 2).

The parser now speculatively parses a SELECT first and keeps the SELECT path only for a bare SELECT or a genuine set operation (more than one element in the union); a lone parenthesized SELECT is rolled back so ParserTableExpression can pick up a trailing alias. Both branches are now covered in 04053_describe_subquery_alias (added UNION ALL).

Verified with a freshly built binary; the test output is byte-identical to the reference:

DESCRIBE (SELECT 1 AS a) AS source            -> a UInt8
DESCRIBE (SELECT 1 AS a) UNION ALL (SELECT 2) -> a UInt8
DESCRIBE (SELECT 1 AS a) INTERSECT (SELECT 2) -> a UInt8
DESCRIBE (SELECT 1 AS a) EXCEPT (SELECT 2)    -> a UInt8
DESCRIBE SELECT 1 AS a                        -> a UInt8
DESCRIBE numbers(10)                          -> number UInt64

CI: the only failures on the previous run were the Performance Comparison (arm_release) jobs, where every flagged query (entropy, rand, logical_functions_small, parallel_hash_join_various) appears in both ::old and ::new — i.e. environmental perf noise, not a regression a DESCRIBE-parser change could cause. The 00175_obfuscator_schema_inference / 00096_obfuscator_save_load failures you noted earlier are already resolved: the branch now contains the revert of #104690 (#106365).

Merged the latest master to refresh CI (the branch was ~700 commits behind and the previous run was red only on unrelated jobs). The PR code is unchanged — the three DESCRIBE source files are byte-identical to the previous head.

The red checks on the previous run are all unrelated to this DESCRIBE-parser change:

• Stress test (arm_msan) — Logical error: Bad cast from type A to B (STID: 4054-600e), an AST-fuzzer exception while building a MergeTreeIndexMinMax / KeyCondition (RPNBuilder), with no DESCRIBE involved. It is a known master-wide flake, tracked as Logical error: Bad cast from type A to B (STID: 4054-600e) #107598 (it fires on master itself and on many unrelated PRs over the last week), and the CI summary already links that issue.
• Performance Comparison (arm_release, 4/6 and 6/6) — every flagged query (hash_table_sizes_stats, array_reduce, group_by_fixed_keys, count_from_formats, aggregate_functions_of_group_by_keys, index_bulk_filtering, …) is flagged in both ::old and ::new, i.e. environmental noise on the arm runners. A change confined to ParserDescribeTableQuery cannot affect the runtime of aggregation/join workloads.

AI Review verdict is ✅ Approve, there are no unresolved review threads, and coverage of the changed lines is 25/25 (100%). Re-running CI on a current master for a clean confirmation before merge.

Merged the latest master to refresh CI (919e667b46a..4c5fee8ce95). The branch was 553 commits behind and the previous run was red only on unrelated, master-wide flakes. The PR code is unchanged — the merge net diff vs master is exactly the three DESCRIBE files (src/Parsers/ParserDescribeTableQuery.cpp and the 04053_describe_subquery_alias test/reference), all byte-identical to the previous head. master touched none of these files.

The previous run's failures are not caused by this PR:

• test_refreshable_mat_view_replicated/test.py::test_circular_dependencies_survive_restart (max_t not unique) — a refreshable-materialized-view flake that fails across many PRs and on master itself (PR=0), ~2–6 times/day; nothing to do with the DESCRIBE parser.
• Stress test (arm_msan) → Hung check failed, possible deadlock found — a chronic master-wide burst (≈45–50 PRs/day on 2026-06-16…19), tracked in Hung check failed, possible deadlock found #107941. The stack trace is in the AST fuzzer (executeASTFuzzerQueries), unrelated to this change.
• BuzzHouse (arm_asan_ubsan) — the run aborted on the injected fault point tcp_handler_fail_connection_setup (MEMORY_LIMIT_EXCEEDED), i.e. a fault-injection harness artifact, not a real crash and not related to the parser.

AI Review is ✅ Approve, all review threads are resolved, and changed-line coverage is complete. The remaining blocker is a human approval.

Merged the latest master to refresh CI (4c5fee8ce95..250b67baacf). The PR code is unchanged — master touched none of the PR's three files (src/Parsers/ParserDescribeTableQuery.cpp and the 04053_describe_subquery_alias test/reference), so the merge net diff vs master is exactly those three files (96 insertions / 2 deletions), all byte-identical to the previous head.

This merge is not just to refresh drift — it brings in a fix for the only red on the previous run. That run's sole failure was a job-level cancellation of Integration tests (amd_asan_ubsan, db disk, old analyzer, 5/6): test_storage_iceberg_concurrent/test_concurrent_reads.py::test_concurrent_reads hung for ~50 minutes, pushing the shard past its time budget (##[error]The operation was canceled.). The other tests reported as failed in that shard (test_huge_concurrent_restore, test_azure_disk_unreachable) were merely in flight at the moment of cancellation, not real failures — the shard's own test report shows 0 failures. The hang is the known-flaky test fixed by #108075, which was merged after the previous merge-base, so this merge picks up the fix.

AI Review is ✅ Approve, all review threads are resolved, and changed-line coverage is complete. The remaining blocker is a human approval.

LLVM Coverage Report

Changed lines: Changed C/C++ lines covered by tests: 25/25 (100.00%) | Lost baseline coverage: none · Uncovered code

Full report · Diff report

Found via ClickGap automated review. Please close or comment if this is incorrect or needs adjustment.

TL;DR

Fix DESCRIBE TABLE (...) AS ... syntax

ClickGap verified: reproduced it with the script below on 26.5; tested 6 version(s) — 5 affected, 1 clean (see Affects); bisected to a candidate range — a single introducing PR could not be pinned (see Bisect).

Reproduction

-- Works on all builds (no trailing alias):
DESCRIBE TABLE (SELECT 1 AS a);
-- Pre-fix: rejected with 'Code: 62 ... Syntax error ... (AS): AS source' (SYNTAX_ERROR);
-- post-fix: returns `a UInt8`:
DESCRIBE TABLE (SELECT 1 AS a) AS source;

▶ Run on ClickHouse Fiddle

Bisect

• The bisect boundary was rejected — PR Set ZooKeeper component in unfreezePartitionsByMatcher #99630 only modifies MergeTree partition freezing logic in ZooKeeper handling, which is completely unrelated to SQL parser changes needed to fix DESCRIBE TABLE syntax. — a coincidental boundary, not the cause; no introducing PR pinned.

Affects

• ✓ Fix is on master — this PR's change fixes the bug on master. Branches below still have the bug and need the backport.
• Reproduces on: 26.5, 26.4, 26.3, 26.2, 26.1
• Does NOT reproduce on: master
• Backport needed: 26.5, 26.4, 26.3, 26.2, 26.1

Suggested next step

The bisect boundary above narrows the search in comp-sql-syntax — treat the candidate range as the starting point; the single introducing PR is unconfirmed.

Component: comp-sql-syntax
Severity: P2

cc fix PR author: @yariks5s
cc fix PR reviewer: @alexey-milovidov

ClickGap · Finding: phase_d_pr100205

Triage metadata

• Primary component: comp-sql-syntax
• Secondary components: comp-testing
• Component owners (comp-sql-syntax): @alexey-milovidov @vitlibar