Add support for CREATE OR REPLACE MATERIALIZED VIEW and TTL GROUP BY

[marre]

Closes #315

Both statement shapes were validated against a real ClickHouse server (26.7.4.58) before and after implementation.

CREATE OR REPLACE MATERIALIZED VIEW

• parseDDL guard now accepts MATERIALIZED after CREATE OR REPLACE.
• CreateMaterializedView gains an OrReplace field, wired through the parser, FormatSQL, and the visitor.

TTL GROUP BY ... SET action

• tryParseTTLPolicy parses a GROUP BY action without a preceding keyword: the keys are read as a plain expression list (matching ClickHouse, which rejects query-level GROUP BY modifiers such as WITH TOTALS/GROUPING SETS inside a TTL, but treats GROUP BY ALL and CUBE(...) as ordinary key expressions), followed by optional SET <col> = <agg expr>[, ...] assignments.
• Reuses the existing GroupByClause and UpdateAssignment AST nodes, so no new node types were introduced.
• TTLPolicyRule gained GroupBy *GroupByClause and Set []*UpdateAssignment; the redundant TTLPolicy.GroupBy field was removed.

Tests

• New fixtures: parser/testdata/ddl/create_or_replace_materialized_view.sql, parser/testdata/ddl/create_table_ttl_group_by.sql (single SET, SET-less, multi-key/multi-SET variants), goldens regenerated.
• TestParser_InvalidSyntax covers the engine-rejected TTL forms (WITH TOTALS/ROLLUP/CUBE, GROUPING SETS).
• Verified: make test, golangci-lint v1.53.3 (CI pin), gofmt.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 83181bcdba

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Copilot]

Pull request overview

Adds ClickHouse parser support for CREATE OR REPLACE MATERIALIZED VIEW and TTL GROUP BY ... SET actions.

Changes:

• Extends AST, parsing, formatting, and traversal support.
• Adds valid and invalid syntax tests plus regenerated goldens.
• Adds formatting fixtures for both statement forms.

Review findings:

• parser/ast.go:2480 — Moderate, 2 votes: TTL span calculations omit Policy.End().
• parser/parser_table.go:1259 — Moderate, 2 votes: GROUP BY ALL SET ... is rejected.
• parser/parser_table.go:1278 — Moderate, 2 votes: Commas before subsequent TTL rules are consumed as assignments.

Reviewed changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Reviewed change
parser/walk.go	Traverses TTL grouping and assignments.
parser/testdata/dml/output/alter_table_modify_ttl_multiple.sql.golden.json	Updates multiple-TTL AST golden data.
parser/testdata/ddl/output/create_table_with_ttl_policy.sql.golden.json	Updates existing TTL policy golden data.
parser/testdata/ddl/output/create_table_ttl_group_by.sql.golden.json	Adds TTL GROUP BY ... SET AST golden data.
parser/testdata/ddl/output/create_or_replace_materialized_view.sql.golden.json	Adds materialized-view replacement AST golden data.
parser/testdata/ddl/output/create_mv_with_order_by.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_mv_with_not_op.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_with_refresh.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_with_gcs.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_with_empty_table_schema.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_with_definer.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_with_comment_before_as.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_rmv_engine_with_columns.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_rmv_depends_on_multi.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/create_materialized_view_basic.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/output/bug_001.sql.golden.json	Updates materialized-view golden data.
parser/testdata/ddl/format/create_table_ttl_group_by.sql	Adds TTL formatting fixture.
parser/testdata/ddl/format/create_or_replace_materialized_view.sql	Adds replacement-view formatting fixture.
parser/testdata/ddl/format/beautify/create_table_ttl_group_by.sql	Adds beautified TTL formatting fixture.
parser/testdata/ddl/format/beautify/create_or_replace_materialized_view.sql	Adds beautified replacement-view fixture.
parser/testdata/ddl/create_table_ttl_group_by.sql	Adds TTL grouping and assignment fixtures.
parser/testdata/ddl/create_or_replace_materialized_view.sql	Adds replacement materialized-view fixture.
parser/parser_view.go	Propagates OR REPLACE state.
parser/parser_test.go	Adds invalid TTL syntax coverage.
parser/parser_table.go	Parses the new DDL and TTL syntax.
parser/format.go	Formats replacement views and TTL actions.
parser/ast.go	Adds materialized-view and TTL AST fields.

Suppressed comments (3)

parser/parser_table.go:1259

• This generic comma-separated list greedily consumes commas that belong to the surrounding parseTTLClause(..., true). For example, in TTL ... GROUP BY id, created + INTERVAL 2 DAY DELETE, the second TTL expression is parsed as another group key and the trailing DELETE remains unexpected. Existing CREATE/ALTER TTL syntax supports multiple comma-separated rules, so the new GROUP BY rule needs a boundary-aware parser rather than parseColumnExprList.

		keys, err := p.parseColumnExprList(p.Pos())

parser/parser_table.go:1245

• This branch is also reachable while parsing a column definition: parseTableColumn calls tryParseTTLClause(..., false) and that reuses tryParseTTLPolicy. A GROUP BY ... SET action is a table TTL operation, so input such as x UInt64 TTL created + INTERVAL 1 DAY GROUP BY id SET x = sum(x) is accepted here even though ClickHouse does not allow row-grouping actions in a column TTL. Pass the table/column context through the TTL parser (or gate this branch for table TTLs) rather than enabling it for both callers.

	case p.matchKeyword(KeywordGroup):

parser/parser_table.go:1259

• parseColumnExprList consumes an initial DISTINCT into HasDistinct, so this TTL-specific path accepts and re-formats ... TTL expr GROUP BY DISTINCT id. DISTINCT is a SELECT/list modifier, not a valid TTL GROUP BY key-list modifier (ClickHouse rejects this form); use an expression-list parser that does not consume DISTINCT here while still allowing ordinary key expressions such as ALL or CUBE(...).

		keys, err := p.parseColumnExprList(p.Pos())

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 39740e0784

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 46c2a9c2c1

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Copilot]

Pull request overview

Copilot reviewed 27 out of 27 changed files in this pull request and generated no new comments.

Suppressed comments (1)

parser/parser_table.go:1259

• The new GROUP BY branch falls through to the shared tryParseWhereClause, so the parser now accepts and formats TTL ... GROUP BY id SET x = sum(x) WHERE id > 0. ClickHouse's TTL GROUP BY action ends after the optional SET list; WHERE belongs to DELETE actions. Return this policy immediately so a trailing WHERE remains unconsumed and is rejected by the statement parser.

	case p.matchKeyword(KeywordGroup):
		groupBy, err := p.parseTTLPolicyGroupBy(pos)
		if err != nil {
			return nil, err
		}
		rule = groupBy

[marre]

Responding to the Copilot review pullrequestreview-4992522436 about TTL ... GROUP BY id SET x = sum(x) WHERE id > 0:

Confirmed and fixed in fa6ab50. Verified on a full ClickHouse server (26.7.4.58) that WHERE is rejected after GROUP BY and RECOMPRESS as well as TO DISK/VOLUME — it belongs only to the DELETE action. tryParseTTLPolicy now parses WHERE only in the DELETE branch and leaves it unconsumed for the statement parser otherwise, so all four ... WHERE forms (GROUP BY SET, GROUP BY, RECOMPRESS, TO VOLUME) are rejected, while DELETE WHERE and every existing fixture still round-trip. Regression tests added to TestParser_InvalidSyntax.

[git-hulk]

LGTM

[git-hulk]

@marre Thank you!