from sqlglot import exp, parse, parse_one from tests.dialects.test_dialect import Validator class TestTSQL(Validator): dialect = "tsql" def test_tsql(self): self.validate_identity("SELECT CASE WHEN a > 1 THEN b END") self.validate_identity("SELECT * FROM taxi ORDER BY 1 OFFSET 0 ROWS FETCH NEXT 3 ROWS ONLY") self.validate_identity("END") self.validate_identity("@x") self.validate_identity("#x") self.validate_identity("DECLARE @TestVariable AS VARCHAR(100)='Save Our Planet'") self.validate_identity("PRINT @TestVariable") self.validate_identity("SELECT Employee_ID, Department_ID FROM @MyTableVar") self.validate_identity("INSERT INTO @TestTable VALUES (1, 'Value1', 12, 20)") self.validate_identity( "SELECT x FROM @MyTableVar AS m JOIN Employee ON m.EmployeeID = Employee.EmployeeID" ) self.validate_identity('SELECT "x"."y" FROM foo') self.validate_identity("SELECT * FROM #foo") self.validate_identity("SELECT * FROM ##foo") self.validate_identity( "SELECT DISTINCT DepartmentName, PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY BaseRate) OVER (PARTITION BY DepartmentName) AS MedianCont FROM dbo.DimEmployee" ) self.validate_all( "SELECT CAST([a].[b] AS SMALLINT) FROM foo", write={ "tsql": 'SELECT CAST("a"."b" AS SMALLINT) FROM foo', "spark": "SELECT CAST(`a`.`b` AS SHORT) FROM foo", }, ) self.validate_all( "CONVERT(INT, CONVERT(NUMERIC, '444.75'))", write={ "mysql": "CAST(CAST('444.75' AS DECIMAL) AS INT)", "tsql": "CAST(CAST('444.75' AS NUMERIC) AS INTEGER)", }, ) self.validate_all( "STRING_AGG(x, y) WITHIN GROUP (ORDER BY z DESC)", write={ "tsql": "STRING_AGG(x, y) WITHIN GROUP (ORDER BY z DESC)", "mysql": "GROUP_CONCAT(x ORDER BY z DESC SEPARATOR y)", "sqlite": "GROUP_CONCAT(x, y)", "postgres": "STRING_AGG(x, y ORDER BY z DESC NULLS LAST)", }, ) self.validate_all( "STRING_AGG(x, '|') WITHIN GROUP (ORDER BY z ASC)", write={ "tsql": "STRING_AGG(x, '|') WITHIN GROUP (ORDER BY z)", "mysql": "GROUP_CONCAT(x ORDER BY z SEPARATOR '|')", "sqlite": "GROUP_CONCAT(x, '|')", "postgres": "STRING_AGG(x, '|' ORDER BY z NULLS FIRST)", }, ) self.validate_all( "STRING_AGG(x, '|')", write={ "tsql": "STRING_AGG(x, '|')", "mysql": "GROUP_CONCAT(x SEPARATOR '|')", "sqlite": "GROUP_CONCAT(x, '|')", "postgres": "STRING_AGG(x, '|')", }, ) self.validate_all( "SELECT CAST([a].[b] AS SMALLINT) FROM foo", write={ "tsql": 'SELECT CAST("a"."b" AS SMALLINT) FROM foo', "spark": "SELECT CAST(`a`.`b` AS SHORT) FROM foo", }, ) self.validate_all( "HASHBYTES('SHA1', x)", read={ "spark": "SHA(x)", }, write={ "tsql": "HASHBYTES('SHA1', x)", "spark": "SHA(x)", }, ) self.validate_all( "HASHBYTES('SHA2_256', x)", read={ "spark": "SHA2(x, 256)", }, write={ "tsql": "HASHBYTES('SHA2_256', x)", "spark": "SHA2(x, 256)", }, ) self.validate_all( "HASHBYTES('SHA2_512', x)", read={ "spark": "SHA2(x, 512)", }, write={ "tsql": "HASHBYTES('SHA2_512', x)", "spark": "SHA2(x, 512)", }, ) self.validate_all( "HASHBYTES('MD5', 'x')", read={ "spark": "MD5('x')", }, write={ "tsql": "HASHBYTES('MD5', 'x')", "spark": "MD5('x')", }, ) self.validate_identity("HASHBYTES('MD2', 'x')") def test_types(self): self.validate_identity("CAST(x AS XML)") self.validate_identity("CAST(x AS UNIQUEIDENTIFIER)") self.validate_identity("CAST(x AS MONEY)") self.validate_identity("CAST(x AS SMALLMONEY)") self.validate_identity("CAST(x AS ROWVERSION)") self.validate_identity("CAST(x AS IMAGE)") self.validate_identity("CAST(x AS SQL_VARIANT)") self.validate_identity("CAST(x AS BIT)") self.validate_all( "CAST(x AS DATETIME2)", read={ "": "CAST(x AS DATETIME)", }, write={ "mysql": "CAST(x AS DATETIME)", "tsql": "CAST(x AS DATETIME2)", }, ) self.validate_all( "CAST(x AS DATETIME2(6))", write={ "hive": "CAST(x AS TIMESTAMP)", }, ) def test_udf(self): self.validate_identity( "CREATE PROCEDURE foo @a INTEGER, @b INTEGER AS SELECT @a = SUM(bla) FROM baz AS bar" ) self.validate_identity( "CREATE PROC foo @ID INTEGER, @AGE INTEGER AS SELECT DB_NAME(@ID) AS ThatDB" ) self.validate_identity("CREATE PROC foo AS SELECT BAR() AS baz") self.validate_identity("CREATE PROCEDURE foo AS SELECT BAR() AS baz") self.validate_identity("CREATE FUNCTION foo(@bar INTEGER) RETURNS TABLE AS RETURN SELECT 1") self.validate_identity("CREATE FUNCTION dbo.ISOweek(@DATE DATETIME2) RETURNS INTEGER") # The following two cases don't necessarily correspond to valid TSQL, but they are used to verify # that the syntax RETURNS @return_variable TABLE ... is parsed correctly. # # See also "Transact-SQL Multi-Statement Table-Valued Function Syntax" # https://learn.microsoft.com/en-us/sql/t-sql/statements/create-function-transact-sql?view=sql-server-ver16 self.validate_identity( "CREATE FUNCTION foo(@bar INTEGER) RETURNS @foo TABLE (x INTEGER, y NUMERIC) AS RETURN SELECT 1" ) self.validate_identity( "CREATE FUNCTION foo() RETURNS @contacts TABLE (first_name VARCHAR(50), phone VARCHAR(25)) AS SELECT @fname, @phone" ) self.validate_all( """ CREATE FUNCTION udfProductInYear ( @model_year INT ) RETURNS TABLE AS RETURN SELECT product_name, model_year, list_price FROM production.products WHERE model_year = @model_year """, write={ "tsql": """CREATE FUNCTION udfProductInYear( @model_year INTEGER ) RETURNS TABLE AS RETURN SELECT product_name, model_year, list_price FROM production.products WHERE model_year = @model_year""", }, pretty=True, ) sql = """ CREATE procedure [TRANSF].[SP_Merge_Sales_Real] @Loadid INTEGER ,@NumberOfRows INTEGER AS BEGIN SET XACT_ABORT ON; DECLARE @DWH_DateCreated DATETIME = CONVERT(DATETIME, getdate(), 104); DECLARE @DWH_DateModified DATETIME = CONVERT(DATETIME, getdate(), 104); DECLARE @DWH_IdUserCreated INTEGER = SUSER_ID (SYSTEM_USER); DECLARE @DWH_IdUserModified INTEGER = SUSER_ID (SYSTEM_USER); DECLARE @SalesAmountBefore float; SELECT @SalesAmountBefore=SUM(SalesAmount) FROM TRANSF.[Pre_Merge_Sales_Real] S; END """ expected_sqls = [ 'CREATE PROCEDURE "TRANSF"."SP_Merge_Sales_Real" @Loadid INTEGER, @NumberOfRows INTEGER AS BEGIN SET XACT_ABORT ON', "DECLARE @DWH_DateCreated DATETIME = CONVERT(DATETIME, getdate(), 104)", "DECLARE @DWH_DateModified DATETIME = CONVERT(DATETIME, getdate(), 104)", "DECLARE @DWH_IdUserCreated INTEGER = SUSER_ID (SYSTEM_USER)", "DECLARE @DWH_IdUserModified INTEGER = SUSER_ID (SYSTEM_USER)", "DECLARE @SalesAmountBefore float", 'SELECT @SalesAmountBefore = SUM(SalesAmount) FROM TRANSF."Pre_Merge_Sales_Real" AS S', "END", ] for expr, expected_sql in zip(parse(sql, read="tsql"), expected_sqls): self.assertEqual(expr.sql(dialect="tsql"), expected_sql) def test_charindex(self): self.validate_all( "CHARINDEX(x, y, 9)", write={ "spark": "LOCATE(x, y, 9)", }, ) self.validate_all( "CHARINDEX(x, y)", write={ "spark": "LOCATE(x, y)", }, ) self.validate_all( "CHARINDEX('sub', 'testsubstring', 3)", write={ "spark": "LOCATE('sub', 'testsubstring', 3)", }, ) self.validate_all( "CHARINDEX('sub', 'testsubstring')", write={ "spark": "LOCATE('sub', 'testsubstring')", }, ) def test_len(self): self.validate_all("LEN(x)", write={"spark": "LENGTH(x)"}) def test_replicate(self): self.validate_all("REPLICATE('x', 2)", write={"spark": "REPEAT('x', 2)"}) def test_isnull(self): self.validate_all("ISNULL(x, y)", write={"spark": "COALESCE(x, y)"}) def test_jsonvalue(self): self.validate_all( "JSON_VALUE(r.JSON, '$.Attr_INT')", write={"spark": "GET_JSON_OBJECT(r.JSON, '$.Attr_INT')"}, ) def test_datefromparts(self): self.validate_all( "SELECT DATEFROMPARTS('2020', 10, 01)", write={"spark": "SELECT MAKE_DATE('2020', 10, 01)"}, ) def test_datename(self): self.validate_all( "SELECT DATENAME(mm,'01-01-1970')", write={"spark": "SELECT DATE_FORMAT('01-01-1970', 'MMMM')"}, ) self.validate_all( "SELECT DATENAME(dw,'01-01-1970')", write={"spark": "SELECT DATE_FORMAT('01-01-1970', 'EEEE')"}, ) def test_datepart(self): self.validate_all( "SELECT DATEPART(month,'01-01-1970')", write={"spark": "SELECT DATE_FORMAT('01-01-1970', 'MM')"}, ) def test_convert_date_format(self): self.validate_all( "CONVERT(NVARCHAR(200), x)", write={ "spark": "CAST(x AS VARCHAR(200))", }, ) self.validate_all( "CONVERT(NVARCHAR, x)", write={ "spark": "CAST(x AS VARCHAR(30))", }, ) self.validate_all( "CONVERT(NVARCHAR(MAX), x)", write={ "spark": "CAST(x AS STRING)", }, ) self.validate_all( "CONVERT(VARCHAR(200), x)", write={ "spark": "CAST(x AS VARCHAR(200))", }, ) self.validate_all( "CONVERT(VARCHAR, x)", write={ "spark": "CAST(x AS VARCHAR(30))", }, ) self.validate_all( "CONVERT(VARCHAR(MAX), x)", write={ "spark": "CAST(x AS STRING)", }, ) self.validate_all( "CONVERT(CHAR(40), x)", write={ "spark": "CAST(x AS CHAR(40))", }, ) self.validate_all( "CONVERT(CHAR, x)", write={ "spark": "CAST(x AS CHAR(30))", }, ) self.validate_all( "CONVERT(NCHAR(40), x)", write={ "spark": "CAST(x AS CHAR(40))", }, ) self.validate_all( "CONVERT(NCHAR, x)", write={ "spark": "CAST(x AS CHAR(30))", }, ) self.validate_all( "CONVERT(VARCHAR, x, 121)", write={ "spark": "CAST(DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS') AS VARCHAR(30))", }, ) self.validate_all( "CONVERT(VARCHAR(40), x, 121)", write={ "spark": "CAST(DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS') AS VARCHAR(40))", }, ) self.validate_all( "CONVERT(VARCHAR(MAX), x, 121)", write={ "spark": "DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS')", }, ) self.validate_all( "CONVERT(NVARCHAR, x, 121)", write={ "spark": "CAST(DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS') AS VARCHAR(30))", }, ) self.validate_all( "CONVERT(NVARCHAR(40), x, 121)", write={ "spark": "CAST(DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS') AS VARCHAR(40))", }, ) self.validate_all( "CONVERT(NVARCHAR(MAX), x, 121)", write={ "spark": "DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS')", }, ) self.validate_all( "CONVERT(DATE, x, 121)", write={ "spark": "TO_DATE(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS')", }, ) self.validate_all( "CONVERT(DATETIME, x, 121)", write={ "spark": "TO_TIMESTAMP(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS')", }, ) self.validate_all( "CONVERT(DATETIME2, x, 121)", write={ "spark": "TO_TIMESTAMP(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS')", }, ) self.validate_all( "CONVERT(INT, x)", write={ "spark": "CAST(x AS INT)", }, ) self.validate_all( "CONVERT(INT, x, 121)", write={ "spark": "CAST(x AS INT)", }, ) self.validate_all( "TRY_CONVERT(NVARCHAR, x, 121)", write={ "spark": "CAST(DATE_FORMAT(x, 'yyyy-MM-dd HH:mm:ss.SSSSSS') AS VARCHAR(30))", }, ) self.validate_all( "TRY_CONVERT(INT, x)", write={ "spark": "CAST(x AS INT)", }, ) self.validate_all( "TRY_CAST(x AS INT)", write={ "spark": "CAST(x AS INT)", }, ) self.validate_all( "CAST(x AS INT)", write={ "spark": "CAST(x AS INT)", }, ) self.validate_all( "SELECT CONVERT(VARCHAR(10), testdb.dbo.test.x, 120) y FROM testdb.dbo.test", write={ "mysql": "SELECT CAST(DATE_FORMAT(testdb.dbo.test.x, '%Y-%m-%d %T') AS VARCHAR(10)) AS y FROM testdb.dbo.test", "spark": "SELECT CAST(DATE_FORMAT(testdb.dbo.test.x, 'yyyy-MM-dd HH:mm:ss') AS VARCHAR(10)) AS y FROM testdb.dbo.test", }, ) self.validate_all( "SELECT CONVERT(VARCHAR(10), y.x) z FROM testdb.dbo.test y", write={ "mysql": "SELECT CAST(y.x AS VARCHAR(10)) AS z FROM testdb.dbo.test AS y", "spark": "SELECT CAST(y.x AS VARCHAR(10)) AS z FROM testdb.dbo.test AS y", }, ) self.validate_all( "SELECT CAST((SELECT x FROM y) AS VARCHAR) AS test", write={ "spark": "SELECT CAST((SELECT x FROM y) AS STRING) AS test", }, ) def test_add_date(self): self.validate_identity("SELECT DATEADD(year, 1, '2017/08/25')") self.validate_all( "SELECT DATEADD(year, 1, '2017/08/25')", write={"spark": "SELECT ADD_MONTHS('2017/08/25', 12)"}, ) self.validate_all( "SELECT DATEADD(qq, 1, '2017/08/25')", write={"spark": "SELECT ADD_MONTHS('2017/08/25', 3)"}, ) self.validate_all( "SELECT DATEADD(wk, 1, '2017/08/25')", write={ "spark": "SELECT DATE_ADD('2017/08/25', 7)", "databricks": "SELECT DATEADD(week, 1, '2017/08/25')", }, ) def test_date_diff(self): self.validate_identity("SELECT DATEDIFF(year, '2020/01/01', '2021/01/01')") self.validate_all( "SELECT DATEDIFF(year, '2020/01/01', '2021/01/01')", write={ "tsql": "SELECT DATEDIFF(year, '2020/01/01', '2021/01/01')", "spark": "SELECT MONTHS_BETWEEN('2021/01/01', '2020/01/01') / 12", }, ) self.validate_all( "SELECT DATEDIFF(mm, 'start','end')", write={ "spark": "SELECT MONTHS_BETWEEN('end', 'start')", "tsql": "SELECT DATEDIFF(month, 'start', 'end')", "databricks": "SELECT DATEDIFF(month, 'start', 'end')", }, ) self.validate_all( "SELECT DATEDIFF(quarter, 'start', 'end')", write={ "spark": "SELECT MONTHS_BETWEEN('end', 'start') / 3", "databricks": "SELECT DATEDIFF(quarter, 'start', 'end')", }, ) def test_iif(self): self.validate_identity("SELECT IIF(cond, 'True', 'False')") self.validate_all( "SELECT IIF(cond, 'True', 'False');", write={ "spark": "SELECT IF(cond, 'True', 'False')", }, ) def test_lateral_subquery(self): self.validate_all( "SELECT x.a, x.b, t.v, t.y FROM x CROSS APPLY (SELECT v, y FROM t) t(v, y)", write={ "spark": "SELECT x.a, x.b, t.v, t.y FROM x JOIN LATERAL (SELECT v, y FROM t) AS t(v, y)", }, ) self.validate_all( "SELECT x.a, x.b, t.v, t.y FROM x OUTER APPLY (SELECT v, y FROM t) t(v, y)", write={ "spark": "SELECT x.a, x.b, t.v, t.y FROM x LEFT JOIN LATERAL (SELECT v, y FROM t) AS t(v, y)", }, ) self.validate_all( "SELECT x.a, x.b, t.v, t.y, s.v, s.y FROM x OUTER APPLY (SELECT v, y FROM t) t(v, y) OUTER APPLY (SELECT v, y FROM t) s(v, y) LEFT JOIN z ON z.id = s.id", write={ "spark": "SELECT x.a, x.b, t.v, t.y, s.v, s.y FROM x LEFT JOIN LATERAL (SELECT v, y FROM t) AS t(v, y) LEFT JOIN LATERAL (SELECT v, y FROM t) AS s(v, y) LEFT JOIN z ON z.id = s.id", }, ) def test_lateral_table_valued_function(self): self.validate_all( "SELECT t.x, y.z FROM x CROSS APPLY tvfTest(t.x)y(z)", write={ "spark": "SELECT t.x, y.z FROM x JOIN LATERAL TVFTEST(t.x) AS y(z)", }, ) self.validate_all( "SELECT t.x, y.z FROM x OUTER APPLY tvfTest(t.x)y(z)", write={ "spark": "SELECT t.x, y.z FROM x LEFT JOIN LATERAL TVFTEST(t.x) AS y(z)", }, ) self.validate_all( "SELECT t.x, y.z FROM x OUTER APPLY a.b.tvfTest(t.x)y(z)", write={ "spark": "SELECT t.x, y.z FROM x LEFT JOIN LATERAL a.b.TVFTEST(t.x) AS y(z)", }, ) def test_top(self): self.validate_all( "SELECT TOP 3 * FROM A", write={ "spark": "SELECT * FROM A LIMIT 3", }, ) self.validate_all( "SELECT TOP (3) * FROM A", write={ "spark": "SELECT * FROM A LIMIT 3", }, ) def test_format(self): self.validate_identity("SELECT FORMAT('01-01-1991', 'd.mm.yyyy')") self.validate_identity("SELECT FORMAT(12345, '###.###.###')") self.validate_identity("SELECT FORMAT(1234567, 'f')") self.validate_all( "SELECT FORMAT(1000000.01,'###,###.###')", write={"spark": "SELECT FORMAT_NUMBER(1000000.01, '###,###.###')"}, ) self.validate_all( "SELECT FORMAT(1234567, 'f')", write={"spark": "SELECT FORMAT_NUMBER(1234567, 'f')"} ) self.validate_all( "SELECT FORMAT('01-01-1991', 'dd.mm.yyyy')", write={"spark": "SELECT DATE_FORMAT('01-01-1991', 'dd.mm.yyyy')"}, ) self.validate_all( "SELECT FORMAT(date_col, 'dd.mm.yyyy')", write={"spark": "SELECT DATE_FORMAT(date_col, 'dd.mm.yyyy')"}, ) self.validate_all( "SELECT FORMAT(date_col, 'm')", write={"spark": "SELECT DATE_FORMAT(date_col, 'MMMM d')"}, ) self.validate_all( "SELECT FORMAT(num_col, 'c')", write={"spark": "SELECT FORMAT_NUMBER(num_col, 'c')"} ) def test_string(self): self.validate_all( "SELECT N'test'", write={"spark": "SELECT 'test'"}, ) self.validate_all( "SELECT n'test'", write={"spark": "SELECT 'test'"}, ) self.validate_all( "SELECT '''test'''", write={"spark": r"SELECT '\'test\''"}, ) def test_eomonth(self): self.validate_all( "EOMONTH(GETDATE())", write={"spark": "LAST_DAY(CURRENT_TIMESTAMP())"}, ) self.validate_all( "EOMONTH(GETDATE(), -1)", write={"spark": "LAST_DAY(ADD_MONTHS(CURRENT_TIMESTAMP(), -1))"}, ) def test_identifier_prefixes(self): expr = parse_one("#x", read="tsql") self.assertIsInstance(expr, exp.Column) self.assertIsInstance(expr.this, exp.Identifier) self.assertEqual(expr.sql("tsql"), "#x") expr = parse_one("@x", read="tsql") self.assertIsInstance(expr, exp.Parameter) self.assertIsInstance(expr.this, exp.Var) self.assertEqual(expr.sql("tsql"), "@x") table = parse_one("select * from @x", read="tsql").args["from"].expressions[0] self.assertIsInstance(table, exp.Table) self.assertIsInstance(table.this, exp.Parameter) self.assertIsInstance(table.this.this, exp.Var) def test_system_time(self): self.validate_all( "SELECT [x] FROM [a].[b] FOR SYSTEM_TIME AS OF 'foo'", write={ "tsql": """SELECT "x" FROM "a"."b" FOR SYSTEM_TIME AS OF 'foo'""", }, ) self.validate_all( "SELECT [x] FROM [a].[b] FOR SYSTEM_TIME AS OF 'foo' AS alias", write={ "tsql": """SELECT "x" FROM "a"."b" FOR SYSTEM_TIME AS OF 'foo' AS alias""", }, ) self.validate_all( "SELECT [x] FROM [a].[b] FOR SYSTEM_TIME FROM c TO d", write={ "tsql": """SELECT "x" FROM "a"."b" FOR SYSTEM_TIME FROM c TO d""", }, ) self.validate_all( "SELECT [x] FROM [a].[b] FOR SYSTEM_TIME BETWEEN c AND d", write={ "tsql": """SELECT "x" FROM "a"."b" FOR SYSTEM_TIME BETWEEN c AND d""", }, ) self.validate_all( "SELECT [x] FROM [a].[b] FOR SYSTEM_TIME CONTAINED IN (c, d)", write={ "tsql": """SELECT "x" FROM "a"."b" FOR SYSTEM_TIME CONTAINED IN (c, d)""", }, ) self.validate_all( "SELECT [x] FROM [a].[b] FOR SYSTEM_TIME ALL AS alias", write={ "tsql": """SELECT "x" FROM "a"."b" FOR SYSTEM_TIME ALL AS alias""", }, ) def test_current_user(self): self.validate_all( "SUSER_NAME()", write={"spark": "CURRENT_USER()"}, ) self.validate_all( "SUSER_SNAME()", write={"spark": "CURRENT_USER()"}, ) self.validate_all( "SYSTEM_USER()", write={"spark": "CURRENT_USER()"}, ) self.validate_all( "SYSTEM_USER", write={"spark": "CURRENT_USER()"}, ) def test_hints(self): self.validate_all( "SELECT x FROM a INNER HASH JOIN b ON b.id = a.id", write={"spark": "SELECT x FROM a INNER JOIN b ON b.id = a.id"}, ) self.validate_all( "SELECT x FROM a INNER LOOP JOIN b ON b.id = a.id", write={"spark": "SELECT x FROM a INNER JOIN b ON b.id = a.id"}, ) self.validate_all( "SELECT x FROM a INNER REMOTE JOIN b ON b.id = a.id", write={"spark": "SELECT x FROM a INNER JOIN b ON b.id = a.id"}, ) self.validate_all( "SELECT x FROM a INNER MERGE JOIN b ON b.id = a.id", write={"spark": "SELECT x FROM a INNER JOIN b ON b.id = a.id"}, ) self.validate_all( "SELECT x FROM a WITH (NOLOCK)", write={ "spark": "SELECT x FROM a", "tsql": "SELECT x FROM a WITH (NOLOCK)", "": "SELECT x FROM a WITH (NOLOCK)", }, ) self.validate_identity("SELECT x FROM a INNER LOOP JOIN b ON b.id = a.id")