PostgreSQL: Support CREATE FOREIGN TABLE

[fmguerreiro]

Parses CREATE FOREIGN TABLE [IF NOT EXISTS] name (columns) SERVER server [OPTIONS (...)] (docs).

Split out of #2317 at @LucaCappelletti94's request. That PR now carries CREATE FOREIGN DATA WRAPPER only.

Table-level OPTIONS reuses the list parser pulled out of parse_pg_create_server, which #2317 also moves; whichever lands second needs a rebase there.

Per-column OPTIONS (...), INHERITS, PARTITION OF and ALTER FOREIGN TABLE are out of scope; the rustdoc notes that.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 69.38776% with 30 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.01%. Comparing base (9352883) to head (b1f9ffe).

Files with missing lines	Patch %	Lines
src/ast/ddl.rs	48.48%	13 Missing and 4 partials ⚠️
src/parser/mod.rs	79.24%	0 Missing and 11 partials ⚠️
src/ast/spans.rs	81.81%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2515      +/-   ##
==========================================
- Coverage   81.02%   81.01%   -0.02%     
==========================================
  Files          42       42              
  Lines       33455    33541      +86     
  Branches    33455    33541      +86     
==========================================
+ Hits        27108    27173      +65     
- Misses       2789     2800      +11     
- Partials     3558     3568      +10     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[LucaCappelletti94]

You should not pin known-wrong behaviour in tests. This span stops at (schema_name because parse_identifier gives quoted identifiers an empty span.

Suggested change

	#[test]
	fn test_create_foreign_table_span_includes_option_keys() {
	let dialect = &crate::dialect::PostgreSqlDialect {};
	let sql = "CREATE FOREIGN TABLE ft (a INT) SERVER s OPTIONS (schema_name 'public')";
	let mut test = SpanTest::new(dialect, sql);
	// Ends at the option key, not the statement: a quoted option value is an
	// Ident with an empty span, so it contributes nothing to the union.
	let stmt = test.0.parse_statement().unwrap();
	assert_eq!(
	test.get_source(stmt.span()),
	"ft (a INT) SERVER s OPTIONS (schema_name"
	);
	}

[fmguerreiro]

shoot, you're right! fixed in 4207b27

[LucaCappelletti94]

You should assert the option spans directly instead of the statement span. The statement span still leaves out CREATE FOREIGN TABLE and the closing ), so this test pins known-wrong behaviour again. The version below fails if parse_identifier drops the span of a quoted identifier.

Suggested change

	#[test]
	fn test_create_foreign_table_span_includes_quoted_option_value() {
	let dialect = &crate::dialect::PostgreSqlDialect {};
	let sql = "CREATE FOREIGN TABLE ft (a INT) SERVER s OPTIONS (schema_name 'public')";
	let mut test = SpanTest::new(dialect, sql);
	// Span reaches through the quoted value but not the closing paren.
	let stmt = test.0.parse_statement().unwrap();
	assert_eq!(
	test.get_source(stmt.span()),
	"ft (a INT) SERVER s OPTIONS (schema_name 'public'"
	);
	}
	#[test]
	fn test_create_foreign_table_option_spans() {
	let dialect = &crate::dialect::PostgreSqlDialect {};
	let sql = r#"CREATE FOREIGN TABLE ft (a INT) SERVER s OPTIONS ("schema_name" 'public')"#;
	let mut test = SpanTest::new(dialect, sql);
	let options = match test.0.parse_statement().unwrap() {
	Statement::CreateForeignTable(stmt) => stmt.options.unwrap(),
	stmt => panic!("expected CREATE FOREIGN TABLE, got {stmt:?}"),
	};
	assert_eq!(test.get_source(options[0].key.span), r#""schema_name""#);
	assert_eq!(test.get_source(options[0].value.span), "'public'");
	}

[fmguerreiro]

done in 88cbb5d, asserting the key and value spans directly.

[LucaCappelletti94]

Thanks for splitting CREATE FOREIGN TABLE out of #2317 and finishing up the PR, @fmguerreiro.

If you have some time, we would really appreciate your help with reviewing other open PRs, which is where the project is most short-handed right now. Since you seem interested in PostgreSQL, running the fuzz_postgres_accepts harness would help a lot too. It feeds generated SQL to PostgreSQL's own grammar through pg_query and reports every statement PostgreSQL accepts that PostgreSqlDialect rejects. You can start it from the fuzz directory with

cargo +nightly fuzz run fuzz_postgres_accepts fuzz_seeds -- -max_total_time=600

and docs/fuzzing.md has the details. Minimized reproductions of what it finds, as issues or small fix PRs, are worth much more to us at this point than new syntax.

cc @alamb @iffyio