Add PostgreSQL `CREATE/ALTER TEXT SEARCH` DDL Parsing by LucaCappelletti94 · Pull Request #2250 · apache/datafusion-sqlparser-rs

LucaCappelletti94 · 2026-02-27T07:37:43Z

Adds parser and AST support for PostgreSQL text search DDL statements:

CREATE TEXT SEARCH DICTIONARY ...
CREATE TEXT SEARCH CONFIGURATION ...
CREATE TEXT SEARCH TEMPLATE ...
CREATE TEXT SEARCH PARSER ...
ALTER TEXT SEARCH DICTIONARY ...
ALTER TEXT SEARCH CONFIGURATION ...
ALTER TEXT SEARCH TEMPLATE ...
ALTER TEXT SEARCH PARSER ...

The work also adds focused regression coverage for previously failing cases.

iffyio · 2026-03-04T13:53:58Z

+            if or_replace || or_alter || temporary || global.is_some() || transient || persistent {
+                return Err(ParserError::ParserError(
+                    "CREATE TEXT SEARCH does not support CREATE modifiers".to_string(),
+                ));
+            }


Suggested change

if or_replace || or_alter || temporary || global.is_some() || transient || persistent {

return Err(ParserError::ParserError(

"CREATE TEXT SEARCH does not support CREATE modifiers".to_string(),

));

}

I think we can skip this validation and leave up to downstream crates if desired

Well this is not semantics, it would be constructing a syntax tree that is not currently valid syntax in any of the supported dialects. I have already encountered this genre of issue a few times and I am a bit confused as to which direction should be preferable.

You would prefer going for the permissive route?

Yeah we can have the parser be permissive in such cases

iffyio · 2026-03-04T13:54:56Z

            && self.parse_one_of_keywords(&[Keyword::PERSISTENT]).is_some();
        let create_view_params = self.parse_create_view_params()?;
-        if self.parse_keyword(Keyword::TABLE) {
+        if self.parse_keywords(&[Keyword::TEXT, Keyword::SEARCH]) {


instead of parse, can we 'peek' TEXT, SEARCH? so that parse_create_text_search() can be somewhat standalone

iffyio · 2026-03-04T13:56:05Z

+        let object_type = self.parse_text_search_object_type()?;
+        let name = self.parse_object_name(false)?;
+        self.expect_token(&Token::LParen)?;
+        let options = self.parse_comma_separated(Parser::parse_text_search_option)?;


Suggested change

let options = self.parse_comma_separated(Parser::parse_text_search_option)?;

let options = self.parse_comma_separated(Parser::parse_sql_option)?;

I think we can reuse this parser given it matches the representation

iffyio · 2026-03-04T13:58:31Z

+    ];
+
+    for sql in sql_cases {
+        if let Err(err) = pg().parse_sql_statements(sql) {


can the tests use verified_statement, also pg_and_generic() where possible?

iffyio · 2026-03-04T13:59:42Z

+        "CREATE TEXT SEARCH DICTIONARY alt_ts_dict1 (template=simple)",
+        "CREATE TEXT SEARCH DICTIONARY alt_ts_dict2 (template=simple)",


could we do a pass through the test cases to remove duplicate/superfluous cases? (for example the first two highlighted here seem identical and expected to cover the same codepaths)

These are all distinct tests cases from the postgres regression corpus, I just included all of the associated ones. That being said, some really seem identical to each other. I will trim them down.

Introduce AST structures for CREATE/ALTER TEXT SEARCH object types\n(dictionary, configuration, template, parser), including display\nimplementations, statement variants, From conversions, and span wiring.

Add parser support for CREATE/ALTER TEXT SEARCH DICTIONARY,\nCONFIGURATION, TEMPLATE, and PARSER forms, including operation-specific\nALTER clauses and strict CREATE option parsing.\n\nRegister text-search object names as parser keywords and reject unsupported\nCREATE modifiers for text-search objects.

Add regression coverage for the provided CREATE/ALTER TEXT SEARCH\nstatements and guardrails for rejected forms (quoted object type,\nmissing key/value option syntax, and unsupported CREATE modifiers).

…tarpaulin

iffyio · 2026-04-24T08:16:27Z

    }
 }

+/// PostgreSQL text search object kind.


Suggested change

/// PostgreSQL text search object kind.

/// Text search object kind.

Can we instead add a [PostgreSQL] link to the reference docs. We can avoid mentioning a specific dialect as part of the main description (it goes stale quickly/immediately if there are other dialects that support the feature)

iffyio · 2026-04-24T08:16:56Z

+    }
+}
+
+/// PostgreSQL `CREATE TEXT SEARCH ...` statement.


Suggested change

/// PostgreSQL `CREATE TEXT SEARCH ...` statement.

/// `CREATE TEXT SEARCH ...` statement.

same comment regarding doc link

iffyio · 2026-04-24T08:17:41Z

+    }
+}
+
+/// Operation for PostgreSQL `ALTER TEXT SEARCH ...`.


Suggested change

/// Operation for PostgreSQL `ALTER TEXT SEARCH ...`.

/// Operation for `ALTER TEXT SEARCH ...`.

same comment regarding doc link

iffyio · 2026-04-24T08:17:55Z

+    }
+}
+
+/// Option assignment used by `ALTER TEXT SEARCH DICTIONARY ... ( ... )`.


Can we add a link to the docs for this struct?

iffyio · 2026-04-24T08:18:19Z

+    }
+}
+
+/// PostgreSQL `ALTER TEXT SEARCH ...` statement.


Suggested change

/// PostgreSQL `ALTER TEXT SEARCH ...` statement.

/// `ALTER TEXT SEARCH ...` statement.

same comment regarding doc link

iffyio · 2026-04-24T08:20:13Z

+            Keyword::CONFIGURATION => Ok(TextSearchObjectType::Configuration),
+            Keyword::TEMPLATE => Ok(TextSearchObjectType::Template),
+            Keyword::PARSER => Ok(TextSearchObjectType::Parser),
+            // unreachable because expect_one_of_keywords used above


Suggested change

// unreachable because expect_one_of_keywords used above

iffyio · 2026-04-24T08:24:13Z

+        Ok(AlterTextSearchDictionaryOption { key, value })
+    }
+
+    /// Parse a PostgreSQL `ALTER TEXT SEARCH ...` statement.


Suggested change

/// Parse a PostgreSQL `ALTER TEXT SEARCH ...` statement.

/// Parse a `ALTER TEXT SEARCH ...` statement.

iffyio · 2026-04-24T08:26:39Z

 }

+#[test]
+fn parse_create_and_alter_text_search() {


can we split this into parse_create_text_search and parse_alter_text_search functions? since they're different syntax

iffyio · 2026-04-24T08:27:18Z

+    );
+
+    // CREATE with quoted option key
+    pg_and_generic().verified_stmt("CREATE TEXT SEARCH TEMPLATE t (\"Init\" = init_function)");


for the text strings, we can use a raw string literal to avoid manually escaping quotes

iffyio · 2026-04-24T08:30:32Z

+                if self.parse_keywords(&[Keyword::RENAME, Keyword::TO]) {
+                    AlterTextSearchOperation::RenameTo {
+                        new_name: self.parse_identifier()?,
+                    }
+                } else if self.parse_keywords(&[Keyword::OWNER, Keyword::TO]) {
+                    AlterTextSearchOperation::OwnerTo(self.parse_owner()?)
+                } else if self.parse_keywords(&[Keyword::SET, Keyword::SCHEMA]) {
+                    AlterTextSearchOperation::SetSchema {
+                        schema_name: self.parse_object_name(false)?,


it looks like these clauses are duplicated for each object type? if so it doesn't seem like we need to make the clauses per object type, we can parse which ever ones show up and if a downstream crate needs to validate which clauses are valid for which object type they can explicitly do so.

LucaCappelletti94 marked this pull request as ready for review February 27, 2026 07:42

iffyio reviewed Mar 4, 2026

View reviewed changes

LucaCappelletti94 force-pushed the postgres-regression-5 branch from 2abd77c to dabed7d Compare March 29, 2026 21:06

LucaCappelletti94 requested a review from iffyio March 29, 2026 21:39

LucaCappelletti94 force-pushed the postgres-regression-5 branch from 9ce9e9a to dabed7d Compare April 10, 2026 12:41

LucaCappelletti94 added 4 commits April 10, 2026 14:51

ast: add PostgreSQL text search DDL statement nodes

e7b0545

Introduce AST structures for CREATE/ALTER TEXT SEARCH object types\n(dictionary, configuration, template, parser), including display\nimplementations, statement variants, From conversions, and span wiring.

tests: add PostgreSQL text search DDL regression coverage

143e0ef

Add regression coverage for the provided CREATE/ALTER TEXT SEARCH\nstatements and guardrails for rejected forms (quoted object type,\nmissing key/value option syntax, and unsupported CREATE modifiers).

Addressed reviewer concerns and extended test coveraged checked with …

182faf5

…tarpaulin

LucaCappelletti94 force-pushed the postgres-regression-5 branch from dabed7d to 182faf5 Compare April 10, 2026 12:54

iffyio reviewed Apr 24, 2026

View reviewed changes

	let options = self.parse_comma_separated(Parser::parse_text_search_option)?;
	let options = self.parse_comma_separated(Parser::parse_sql_option)?;

		"CREATE TEXT SEARCH DICTIONARY alt_ts_dict1 (template=simple)",
		"CREATE TEXT SEARCH DICTIONARY alt_ts_dict2 (template=simple)",

	/// PostgreSQL text search object kind.
	/// Text search object kind.

	/// PostgreSQL `CREATE TEXT SEARCH ...` statement.
	/// `CREATE TEXT SEARCH ...` statement.

	/// Operation for PostgreSQL `ALTER TEXT SEARCH ...`.
	/// Operation for `ALTER TEXT SEARCH ...`.

	/// PostgreSQL `ALTER TEXT SEARCH ...` statement.
	/// `ALTER TEXT SEARCH ...` statement.

	/// Parse a PostgreSQL `ALTER TEXT SEARCH ...` statement.
	/// Parse a `ALTER TEXT SEARCH ...` statement.

Conversation

LucaCappelletti94 commented Feb 27, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants