Parse PostgreSQL event triggers, REINDEX, SECURITY LABEL and schema elements

[minleejae]

Purpose

Parse PostgreSQL administration statements that fail in every configuration today. pg_dump output and maintenance scripts contain them, for example:

• CREATE EVENT TRIGGER etg ON ddl_command_start WHEN TAG IN ('CREATE TABLE') EXECUTE FUNCTION abort_any_command()
• ALTER EVENT TRIGGER etg ENABLE ALWAYS
• REINDEX INDEX CONCURRENTLY idx
• SECURITY LABEL FOR selinux ON TABLE t1 IS 'system_u:object_r:sepgsql_table_t:s0'
• CREATE SCHEMA hollywood CREATE TABLE films (…) CREATE VIEW winners AS SELECT …

Current design (As-is)

• The statement grammar has no alternative for REINDEX or SECURITY LABEL.
• CREATE EVENT … and ALTER EVENT … reach only the MySQL event productions, which fail at the trigger name.
• CREATE SCHEMA handles schema elements only partly:
  • It accepts one CREATE TABLE element, and also a VIEW … element without CREATE. A second CREATE element fails.
  • toString() and StatementDeParser drop all elements and the PATH clause.
  • The parser rewrites each CREATE TABLE element's name to carry the schema name.
  • TablesNamesFinder throws for every CREATE SCHEMA.

Target design (To-be)

• CreateEventTrigger (statement.create.trigger) records:
  • the name and the event name;
  • the WHEN filters in source order, each a variable with its string values;
  • the invoked function. It reuses CreateTrigger.ExecuteKeyword and the Function reference of CREATE TRIGGER, with an empty argument list.
  • Rejected: arguments, OR REPLACE, and empty or non-string tag lists.
• AlterEventTrigger (statement.alter):
  • ENABLE [REPLICA | ALWAYS] and DISABLE reuse RelationAlterAction.EnableState and the PostgreSqlEnableState() production from Share PostgreSQL trigger and rule enable states #2718.
  • It also covers OWNER TO and RENAME TO.
  • Each non-null setter selects its action and discards the values of the other actions.
• Telling event triggers from MySQL events: EVENT TRIGGER starts a PostgreSQL event trigger only when the name is followed by ON (CREATE) or by an action keyword (ALTER). A MySQL event named trigger still parses to CreateEvent/AlterEvent.
• Reindex (new statement.reindex package) records:
  • the Target (INDEX, TABLE, SCHEMA, DATABASE, SYSTEM) and the CONCURRENTLY keyword;
  • the parenthesized CONCURRENTLY, TABLESPACE and VERBOSE options, with values as written;
  • the name. Only DATABASE and SYSTEM may omit it.
  • REINDEX becomes a non-reserved token. A statement that starts with a plain identifier adds choice conflicts at several optional trailing identifiers, so it needs a token, like NOTIFY in Support PostgreSQL rewrite rules with reusable statement actions #2723. The word remains usable as an identifier, and KeywordsTest covers it.
• SecurityLabel (new statement.securitylabel package) records the provider as written and one target. getObjectType() covers all 21 object kinds that PostgreSQL documents:
  • named objects as Table;
  • a qualified COLUMN;
  • FUNCTION/PROCEDURE/ROUTINE/AGGREGATE signatures, using the shared RoutineReference;
  • LARGE OBJECT OIDs.
  • IS NULL is a null label.
  • It does not reuse CommentTarget, which has no TABLE/COLUMN/VIEW kinds and is being extended by the COMMENT ON targets PR. Qualified names reuse CommentObjectName().
• CREATE SCHEMA elements:
  • A new SchemaElement() production reuses Create() and Grant(). It accepts PostgreSQL's element kinds: CREATE TABLE, VIEW, INDEX, SEQUENCE, TRIGGER and GRANT.
  • It rejects materialized views, other CREATE statements, the VIEW form without CREATE, and elements after IF NOT EXISTS.
  • Behavior change: elements keep their names as written. The parser no longer writes the schema name into CREATE TABLE elements.
  • toString() and StatementDeParser render PATH and the elements. TablesNamesFinder and StatementVisitorAdapter traverse the elements.
  • Inside an element, GRANT is never read as an implicit alias (… FROM films GRANT SELECT …). Elsewhere grant remains an alias.
• Wiring:
  • StatementVisitor defaults, the adapter, StatementDeParser, TablesNamesFinder, StatementValidator and StatementFeatureVisitor.
  • Feature analysis: schema change; REINDEX also reads data.
  • With unsupported statements enabled, the statement-start check recognizes REINDEX and SECURITY LABEL instead of capturing them as opaque.
  • New features createEventTrigger, alterEventTrigger, reindex, reindexConcurrently (PostgreSQL 12) and securityLabel. They are in PostgresqlVersion, and the event trigger features are also in FeaturesAllowed.CREATE/ALTER.
  • The usage guide, README and index list the statements.

Validation

• 100 new tests:
  • PostgreSqlEventTriggerTest 23, ReindexTest 27, SecurityLabelTest 39, and 11 more in CreateSchemaTest.
  • Every form: parse, deparse and reparse with Dialect.POSTGRESQL, plus the AST fields.
  • Rejected forms, builder and setter round trips, and table discovery.
  • MySQL events named trigger, and the new words used as identifiers.
  • Statements among unsupported statements, validation capabilities and statement features.
  • The existing CREATE SCHEMA tests pass unchanged.
• Java 17 offline Gradle spotlessApply check: 10,477 tests, 10,452 passed, 25 skipped, zero failures or errors. The grammar choice conflict check passes.
• Compared with master (bdf8ac31) on a corpus of 635 MySQL 8.4 and PostgreSQL 18 DDL/DCL statements, each accepted by the server, plus variants:
  • Fixed with Dialect.POSTGRESQL and without a dialect: CREATE EVENT TRIGGER, REINDEX INDEX CONCURRENTLY, REINDEX TABLE, SECURITY LABEL and CREATE SCHEMA with schema elements. Each renders identically to the source.
  • 0 regressed and 0 changed across the MySQL and PostgreSQL corpora, the keyword sweeps (reindex, security and label as names), the variants and the scripts.
• MySQL 8.4 and PostgreSQL 18.3: every corpus statement this branch parses ran as written on one server and as rendered by toString() on a second, identical server, in the same order. All 205 MySQL and 332 PostgreSQL statements give the same result (success, or the same error).

Integration validation

Applied #2806–#2819 together on master (bdf8ac31), with the stacked PRs on top of their bases (#2806 → #2807 → #2808, #2806 → #2809 → #2810 and #2811 → #2812).

• Each branch merges cleanly into master. Between the branches, conflicts are limited to adjacent additions (README/index rows, usage sections, imports, visitor and feature lists, one enum line and one alias guard).
• The integrated build passes Java 17 offline Gradle spotlessApply check: 11,494 tests, 11,469 passed, 25 skipped, zero failures/errors. The grammar choice conflict check passes.
• On a corpus of 635 MySQL 8.4 and PostgreSQL 18 DDL/DCL statements, each accepted by the server, the integrated build parses 265 of 271 MySQL statements (master: 206) and 363 of 364 PostgreSQL statements (master: 328). Every MySQL 8.4 non-reserved word and PostgreSQL 18 unreserved or column-name keyword parses as a column and table name.
• On paired MySQL 8.4 and PostgreSQL 18.3 servers, all 263 MySQL and 362 PostgreSQL statements rendered by the integrated build give the same results as the originals.