feat(starrocks)!: stop eliminating semi/anti joins, QUALIFY, and FULL OUTER JOIN [CLAUDE] (#7524)

* feat(starrocks): support semi/anti joins, eliminate DISTINCT ON & GENERATE_DATE_ARRAY [CLAUDE]

StarRocks supports LEFT SEMI JOIN and LEFT ANTI JOIN natively, so
remove ANTI and SEMI from TABLE_ALIAS_TOKENS (MySQL added them back
since it lacks semi/anti join syntax).

Add Select preprocessors to eliminate DISTINCT ON (not supported) and
rewrite UNNEST(GENERATE_DATE_ARRAY(...)) to recursive CTEs (StarRocks
only supports numeric inputs for array_generate/generate_series).

Made-with: Cursor

* Update tests/dialects/test_starrocks.py

* Update tests/dialects/test_starrocks.py

---------

Co-authored-by: Jo <46752250+georgesittas@users.noreply.github.com>

Author: dwoldemariam-klav

sqlglot/generators/starrocks.py

@@ -100,6 +100,14 @@ class StarRocksGenerator(MySQLGenerator):
         exp.JSONExtract: arrow_json_extract_sql,
         exp.Property: property_sql,
         exp.RegexpLike: rename_func("REGEXP"),
+        # Inherited from MySQL, minus operations StarRocks supports natively
+        # (QUALIFY, FULL OUTER JOIN, SEMI/ANTI JOIN)
+        exp.Select: transforms.preprocess(
+            [
+                transforms.eliminate_distinct_on,
+                transforms.unnest_generate_date_array_using_recursive_cte,
+            ]
+        ),
         exp.SchemaCommentProperty: lambda self, e: self.naked_property(e),
         exp.SqlSecurityProperty: lambda self, e: f"SECURITY {self.sql(e.this)}",
         exp.StDistance: st_distance_sphere,

sqlglot/parsers/starrocks.py

@@ -5,9 +5,14 @@
 from sqlglot.dialects.dialect import build_date_delta_with_interval, build_timestamp_trunc
 from sqlglot.helper import seq_get
 from sqlglot.parsers.mysql import MySQLParser
+from sqlglot.tokens import TokenType
 
 
 class StarRocksParser(MySQLParser):
+    # StarRocks supports LEFT SEMI JOIN and LEFT ANTI JOIN natively
+    # https://docs.starrocks.io/docs/sql-reference/sql-statements/table_bucket_part_index/SELECT/SELECT_JOIN/
+    TABLE_ALIAS_TOKENS = MySQLParser.TABLE_ALIAS_TOKENS - {TokenType.ANTI, TokenType.SEMI}
+
     FUNCTIONS = {
         **MySQLParser.FUNCTIONS,
         "ADDDATE": build_date_delta_with_interval(exp.DateAdd, default_unit="DAY"),

tests/dialects/test_dialect.py

@@ -3641,6 +3641,7 @@ def test_generate_date_array(self):
                 "databricks": "SELECT * FROM EXPLODE(SEQUENCE(CAST('2020-01-01' AS DATE), CAST('2020-02-01' AS DATE), INTERVAL '1' WEEK))",
                 "duckdb": "SELECT * FROM UNNEST(CAST(GENERATE_SERIES(CAST('2020-01-01' AS DATE), CAST('2020-02-01' AS DATE), INTERVAL '1' WEEK) AS DATE[]))",
                 "mysql": "WITH RECURSIVE _generated_dates(date_value) AS (SELECT CAST('2020-01-01' AS DATE) AS date_value UNION ALL SELECT CAST(DATE_ADD(date_value, INTERVAL 1 WEEK) AS DATE) FROM _generated_dates WHERE CAST(DATE_ADD(date_value, INTERVAL 1 WEEK) AS DATE) <= CAST('2020-02-01' AS DATE)) SELECT * FROM (SELECT date_value FROM _generated_dates) AS _generated_dates",
+                "starrocks": "WITH RECURSIVE _generated_dates(date_value) AS (SELECT CAST('2020-01-01' AS DATE) AS date_value UNION ALL SELECT CAST(DATE_ADD(date_value, INTERVAL 1 WEEK) AS DATE) FROM _generated_dates WHERE CAST(DATE_ADD(date_value, INTERVAL 1 WEEK) AS DATE) <= CAST('2020-02-01' AS DATE)) SELECT * FROM (SELECT date_value FROM _generated_dates) AS _generated_dates",
                 "postgres": "SELECT * FROM (SELECT CAST(value AS DATE) FROM GENERATE_SERIES(CAST('2020-01-01' AS DATE), CAST('2020-02-01' AS DATE), INTERVAL '1 WEEK') AS _t(value)) AS _unnested_generate_series",
                 "presto": "SELECT * FROM UNNEST(SEQUENCE(CAST('2020-01-01' AS DATE), CAST('2020-02-01' AS DATE), (1 * INTERVAL '7' DAY)))",
                 "redshift": "WITH RECURSIVE _generated_dates(date_value) AS (SELECT CAST('2020-01-01' AS DATE) AS date_value UNION ALL SELECT CAST(DATEADD(WEEK, 1, date_value) AS DATE) FROM _generated_dates WHERE CAST(DATEADD(WEEK, 1, date_value) AS DATE) <= CAST('2020-02-01' AS DATE)) SELECT * FROM (SELECT date_value FROM _generated_dates) AS _generated_dates",

tests/dialects/test_duckdb.py

@@ -233,7 +233,7 @@ def test_duckdb(self):
                     "snowflake": f"SELECT * FROM t1 WHERE {exists}(SELECT 1 FROM t2 WHERE t1.x = t2.x)",
                     "spark": f"SELECT * FROM t1 {join_type} JOIN t2 ON t1.x = t2.x",
                     "sqlite": f"SELECT * FROM t1 WHERE {exists}(SELECT 1 FROM t2 WHERE t1.x = t2.x)",
-                    "starrocks": f"SELECT * FROM t1 WHERE {exists}(SELECT 1 FROM t2 WHERE t1.x = t2.x)",
+                    "starrocks": f"SELECT * FROM t1 {join_type} JOIN t2 ON t1.x = t2.x",
                     "teradata": f"SELECT * FROM t1 WHERE {exists}(SELECT 1 FROM t2 WHERE t1.x = t2.x)",
                     "trino": f"SELECT * FROM t1 WHERE {exists}(SELECT 1 FROM t2 WHERE t1.x = t2.x)",
                     "tsql": f"SELECT * FROM t1 WHERE {exists}(SELECT 1 FROM t2 WHERE t1.x = t2.x)",

tests/dialects/test_starrocks.py

@@ -27,6 +27,21 @@ def test_starrocks(self):
 
         self.validate_identity("CURRENT_VERSION()")
 
+        self.validate_identity("SELECT t1.id FROM t1 LEFT ANTI JOIN t2 ON t1.id = t2.id")
+        self.validate_identity("SELECT t1.id FROM t1 LEFT SEMI JOIN t2 ON t1.id = t2.id")
+
+    def test_distinct_on(self):
+        self.validate_identity(
+            "SELECT DISTINCT ON (a) a, b FROM x ORDER BY c DESC",
+            "SELECT a, b FROM (SELECT a AS a, b AS b, ROW_NUMBER() OVER (PARTITION BY a ORDER BY c DESC) AS _row_number FROM x) AS _t WHERE _row_number = 1",
+        )
+
+    def test_generate_date_array(self):
+        self.validate_identity(
+            "SELECT * FROM UNNEST(GENERATE_DATE_ARRAY(DATE '2020-01-01', DATE '2020-02-01', INTERVAL 1 WEEK)) AS _q(date_week)",
+            "WITH RECURSIVE _generated_dates(date_week) AS (SELECT CAST('2020-01-01' AS DATE) AS date_week UNION ALL SELECT CAST(DATE_ADD(date_week, INTERVAL 1 WEEK) AS DATE) FROM _generated_dates WHERE CAST(DATE_ADD(date_week, INTERVAL 1 WEEK) AS DATE) <= CAST('2020-02-01' AS DATE)) SELECT * FROM (SELECT date_week FROM _generated_dates) AS _generated_dates",
+        )
+
     def test_ddl(self):
         self.validate_identity("INSERT OVERWRITE my_table SELECT * FROM other_table")
         self.validate_identity("CREATE TABLE t (c INT) COMMENT 'c'")