from unittest import mock from sqlglot import exp, parse_one from tests.dialects.test_dialect import Validator class TestSpark(Validator): dialect = "spark" def test_ddl(self): self.validate_identity("CREATE TABLE foo (col VARCHAR(50))") self.validate_identity("CREATE TABLE foo (col STRUCT)") self.validate_identity("CREATE TABLE foo (col STRING) CLUSTERED BY (col) INTO 10 BUCKETS") self.validate_identity( "CREATE TABLE foo (col STRING) CLUSTERED BY (col) SORTED BY (col) INTO 10 BUCKETS" ) self.validate_all( "CREATE TABLE db.example_table (col_a struct)", write={ "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b TEXT))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b VARCHAR))", "hive": "CREATE TABLE db.example_table (col_a STRUCT)", "spark": "CREATE TABLE db.example_table (col_a STRUCT)", }, ) self.validate_all( "CREATE TABLE db.example_table (col_a struct>)", write={ "bigquery": "CREATE TABLE db.example_table (col_a STRUCT>)", "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b STRUCT(nested_col_a TEXT, nested_col_b TEXT)))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b ROW(nested_col_a VARCHAR, nested_col_b VARCHAR)))", "hive": "CREATE TABLE db.example_table (col_a STRUCT>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT>)", }, ) self.validate_all( "CREATE TABLE db.example_table (col_a array, col_b array>)", write={ "bigquery": "CREATE TABLE db.example_table (col_a ARRAY, col_b ARRAY>)", "duckdb": "CREATE TABLE db.example_table (col_a INT[], col_b INT[][])", "presto": "CREATE TABLE db.example_table (col_a ARRAY(INTEGER), col_b ARRAY(ARRAY(INTEGER)))", "hive": "CREATE TABLE db.example_table (col_a ARRAY, col_b ARRAY>)", "spark": "CREATE TABLE db.example_table (col_a ARRAY, col_b ARRAY>)", "snowflake": "CREATE TABLE db.example_table (col_a ARRAY, col_b ARRAY)", }, ) self.validate_all( "CREATE TABLE x USING ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", write={ "duckdb": "CREATE TABLE x", "presto": "CREATE TABLE x WITH (FORMAT='ICEBERG', PARTITIONED_BY=ARRAY['MONTHS'])", "hive": "CREATE TABLE x STORED AS ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", "spark": "CREATE TABLE x USING ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", }, ) self.validate_all( "CREATE TABLE test STORED AS PARQUET AS SELECT 1", write={ "duckdb": "CREATE TABLE test AS SELECT 1", "presto": "CREATE TABLE test WITH (FORMAT='PARQUET') AS SELECT 1", "hive": "CREATE TABLE test STORED AS PARQUET AS SELECT 1", "spark": "CREATE TABLE test USING PARQUET AS SELECT 1", }, ) self.validate_all( """CREATE TABLE blah (col_a INT) COMMENT "Test comment: blah" PARTITIONED BY (date STRING) STORED AS ICEBERG TBLPROPERTIES('x' = '1')""", write={ "duckdb": """CREATE TABLE blah ( col_a INT )""", # Partition columns should exist in table "presto": """CREATE TABLE blah ( col_a INTEGER, date VARCHAR ) COMMENT='Test comment: blah' WITH ( PARTITIONED_BY=ARRAY['date'], FORMAT='ICEBERG', x='1' )""", "hive": """CREATE TABLE blah ( col_a INT ) COMMENT 'Test comment: blah' PARTITIONED BY ( date STRING ) STORED AS ICEBERG TBLPROPERTIES ( 'x'='1' )""", "spark": """CREATE TABLE blah ( col_a INT ) COMMENT 'Test comment: blah' PARTITIONED BY ( date STRING ) USING ICEBERG TBLPROPERTIES ( 'x'='1' )""", }, pretty=True, ) self.validate_all( "CACHE TABLE testCache OPTIONS ('storageLevel' 'DISK_ONLY') SELECT * FROM testData", write={ "spark": "CACHE TABLE testCache OPTIONS('storageLevel' = 'DISK_ONLY') AS SELECT * FROM testData" }, ) self.validate_all( "ALTER TABLE StudentInfo ADD COLUMNS (LastName STRING, DOB TIMESTAMP)", write={ "spark": "ALTER TABLE StudentInfo ADD COLUMNS (LastName STRING, DOB TIMESTAMP)", }, ) self.validate_all( "ALTER TABLE StudentInfo DROP COLUMNS (LastName, DOB)", write={ "spark": "ALTER TABLE StudentInfo DROP COLUMNS (LastName, DOB)", }, ) self.validate_all( "CREATE TABLE x USING ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", identify=True, write={ "spark": "CREATE TABLE `x` USING ICEBERG PARTITIONED BY (MONTHS(`y`)) LOCATION 's3://z'", }, ) def test_to_date(self): self.validate_all( "TO_DATE(x, 'yyyy-MM-dd')", write={ "duckdb": "CAST(x AS DATE)", "hive": "TO_DATE(x)", "presto": "CAST(CAST(x AS TIMESTAMP) AS DATE)", "spark": "TO_DATE(x)", }, ) self.validate_all( "TO_DATE(x, 'yyyy')", write={ "duckdb": "CAST(STRPTIME(x, '%Y') AS DATE)", "hive": "TO_DATE(x, 'yyyy')", "presto": "CAST(DATE_PARSE(x, '%Y') AS DATE)", "spark": "TO_DATE(x, 'yyyy')", }, ) @mock.patch("sqlglot.generator.logger") def test_hint(self, logger): self.validate_all( "SELECT /*+ COALESCE(3) */ * FROM x", write={ "spark": "SELECT /*+ COALESCE(3) */ * FROM x", "bigquery": "SELECT * FROM x", }, ) self.validate_all( "SELECT /*+ COALESCE(3), REPARTITION(1) */ * FROM x", write={ "spark": "SELECT /*+ COALESCE(3), REPARTITION(1) */ * FROM x", "bigquery": "SELECT * FROM x", }, ) self.validate_all( "SELECT /*+ BROADCAST(table) */ cola FROM table", write={ "spark": "SELECT /*+ BROADCAST(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ BROADCASTJOIN(table) */ cola FROM table", write={ "spark": "SELECT /*+ BROADCASTJOIN(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ MAPJOIN(table) */ cola FROM table", write={ "spark": "SELECT /*+ MAPJOIN(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ MERGE(table) */ cola FROM table", write={ "spark": "SELECT /*+ MERGE(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ SHUFFLEMERGE(table) */ cola FROM table", write={ "spark": "SELECT /*+ SHUFFLEMERGE(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ MERGEJOIN(table) */ cola FROM table", write={ "spark": "SELECT /*+ MERGEJOIN(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ SHUFFLE_HASH(table) */ cola FROM table", write={ "spark": "SELECT /*+ SHUFFLE_HASH(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) self.validate_all( "SELECT /*+ SHUFFLE_REPLICATE_NL(table) */ cola FROM table", write={ "spark": "SELECT /*+ SHUFFLE_REPLICATE_NL(table) */ cola FROM table", "bigquery": "SELECT cola FROM table", }, ) def test_spark(self): expr = parse_one("any_value(col, true)", read="spark") self.assertIsInstance(expr.args.get("ignore_nulls"), exp.Boolean) self.assertEqual(expr.sql(dialect="spark"), "ANY_VALUE(col, TRUE)") self.validate_identity("SELECT * FROM t1 SEMI JOIN t2 ON t1.x = t2.x") self.validate_identity("SELECT TRANSFORM(ARRAY(1, 2, 3), x -> x + 1)") self.validate_identity("SELECT TRANSFORM(ARRAY(1, 2, 3), (x, i) -> x + i)") self.validate_identity("REFRESH table a.b.c") self.validate_identity("INTERVAL -86 days") self.validate_identity("SELECT UNIX_TIMESTAMP()") self.validate_identity("TRIM(' SparkSQL ')") self.validate_identity("TRIM(BOTH 'SL' FROM 'SSparkSQLS')") self.validate_identity("TRIM(LEADING 'SL' FROM 'SSparkSQLS')") self.validate_identity("TRIM(TRAILING 'SL' FROM 'SSparkSQLS')") self.validate_identity("SPLIT(str, pattern, lim)") self.validate_identity( "SELECT REGEXP_REPLACE('100-200', r'([^0-9])', '')", "SELECT REGEXP_REPLACE('100-200', '([^0-9])', '')", ) self.validate_identity( "SELECT REGEXP_REPLACE('100-200', R'([^0-9])', '')", "SELECT REGEXP_REPLACE('100-200', '([^0-9])', '')", ) self.validate_identity( "SELECT STR_TO_MAP('a:1,b:2,c:3')", "SELECT STR_TO_MAP('a:1,b:2,c:3', ',', ':')", ) self.validate_all( "foo.bar", read={ "": "STRUCT_EXTRACT(foo, bar)", }, ) self.validate_all( "MAP(1, 2, 3, 4)", write={ "spark": "MAP(1, 2, 3, 4)", "trino": "MAP(ARRAY[1, 3], ARRAY[2, 4])", }, ) self.validate_all( "MAP()", read={ "spark": "MAP()", "trino": "MAP()", }, write={ "trino": "MAP(ARRAY[], ARRAY[])", }, ) self.validate_all( "SELECT STR_TO_MAP('a:1,b:2,c:3', ',', ':')", read={ "presto": "SELECT SPLIT_TO_MAP('a:1,b:2,c:3', ',', ':')", "spark": "SELECT STR_TO_MAP('a:1,b:2,c:3', ',', ':')", }, write={ "presto": "SELECT SPLIT_TO_MAP('a:1,b:2,c:3', ',', ':')", "spark": "SELECT STR_TO_MAP('a:1,b:2,c:3', ',', ':')", }, ) self.validate_all( "SELECT DATEDIFF(month, CAST('1996-10-30' AS TIMESTAMP), CAST('1997-02-28 10:30:00' AS TIMESTAMP))", read={ "duckdb": "SELECT DATEDIFF('month', CAST('1996-10-30' AS TIMESTAMP), CAST('1997-02-28 10:30:00' AS TIMESTAMP))", }, write={ "spark": "SELECT DATEDIFF(month, TO_DATE(CAST('1996-10-30' AS TIMESTAMP)), TO_DATE(CAST('1997-02-28 10:30:00' AS TIMESTAMP)))", "spark2": "SELECT MONTHS_BETWEEN(TO_DATE(CAST('1997-02-28 10:30:00' AS TIMESTAMP)), TO_DATE(CAST('1996-10-30' AS TIMESTAMP)))", }, ) self.validate_all( "SELECT MONTHS_BETWEEN('1997-02-28 10:30:00', '1996-10-30')", write={ "duckdb": "SELECT DATEDIFF('month', CAST('1996-10-30' AS TIMESTAMP), CAST('1997-02-28 10:30:00' AS TIMESTAMP))", "hive": "SELECT MONTHS_BETWEEN('1997-02-28 10:30:00', '1996-10-30')", "spark": "SELECT MONTHS_BETWEEN('1997-02-28 10:30:00', '1996-10-30')", }, ) self.validate_all( "SELECT MONTHS_BETWEEN('1997-02-28 10:30:00', '1996-10-30', FALSE)", write={ "duckdb": "SELECT DATEDIFF('month', CAST('1996-10-30' AS TIMESTAMP), CAST('1997-02-28 10:30:00' AS TIMESTAMP))", "hive": "SELECT MONTHS_BETWEEN('1997-02-28 10:30:00', '1996-10-30')", "spark": "SELECT MONTHS_BETWEEN('1997-02-28 10:30:00', '1996-10-30', FALSE)", }, ) self.validate_all( "SELECT TO_TIMESTAMP('2016-12-31 00:12:00')", write={ "": "SELECT CAST('2016-12-31 00:12:00' AS TIMESTAMP)", "duckdb": "SELECT CAST('2016-12-31 00:12:00' AS TIMESTAMP)", "spark": "SELECT CAST('2016-12-31 00:12:00' AS TIMESTAMP)", }, ) self.validate_all( "SELECT TO_TIMESTAMP('2016-12-31', 'yyyy-MM-dd')", read={ "duckdb": "SELECT STRPTIME('2016-12-31', '%Y-%m-%d')", }, write={ "": "SELECT STR_TO_TIME('2016-12-31', '%Y-%m-%d')", "duckdb": "SELECT STRPTIME('2016-12-31', '%Y-%m-%d')", "spark": "SELECT TO_TIMESTAMP('2016-12-31', 'yyyy-MM-dd')", }, ) self.validate_all( "SELECT RLIKE('John Doe', 'John.*')", write={ "bigquery": "SELECT REGEXP_CONTAINS('John Doe', 'John.*')", "hive": "SELECT 'John Doe' RLIKE 'John.*'", "postgres": "SELECT 'John Doe' ~ 'John.*'", "snowflake": "SELECT REGEXP_LIKE('John Doe', 'John.*')", "spark": "SELECT 'John Doe' RLIKE 'John.*'", }, ) self.validate_all( "UNHEX(MD5(x))", write={ "bigquery": "FROM_HEX(TO_HEX(MD5(x)))", "spark": "UNHEX(MD5(x))", }, ) self.validate_all( "SELECT * FROM ((VALUES 1))", write={"spark": "SELECT * FROM (VALUES (1))"} ) self.validate_all( "SELECT CAST(STRUCT('fooo') AS STRUCT)", write={"spark": "SELECT CAST(STRUCT('fooo') AS STRUCT)"}, ) self.validate_all( "SELECT CAST(123456 AS VARCHAR(3))", write={ "": "SELECT TRY_CAST(123456 AS TEXT)", "spark": "SELECT CAST(123456 AS STRING)", }, ) self.validate_all( "SELECT piv.Q1 FROM (SELECT * FROM produce PIVOT(SUM(sales) FOR quarter IN ('Q1', 'Q2'))) AS piv", read={ "snowflake": "SELECT piv.Q1 FROM produce PIVOT(SUM(sales) FOR quarter IN ('Q1', 'Q2')) piv", }, ) self.validate_all( "SELECT piv.Q1 FROM (SELECT * FROM (SELECT * FROM produce) PIVOT(SUM(sales) FOR quarter IN ('Q1', 'Q2'))) AS piv", read={ "snowflake": "SELECT piv.Q1 FROM (SELECT * FROM produce) PIVOT(SUM(sales) FOR quarter IN ('Q1', 'Q2')) piv", }, ) self.validate_all( "SELECT * FROM produce PIVOT(SUM(produce.sales) FOR quarter IN ('Q1', 'Q2'))", read={ "snowflake": "SELECT * FROM produce PIVOT (SUM(produce.sales) FOR produce.quarter IN ('Q1', 'Q2'))", }, ) self.validate_all( "SELECT * FROM produce AS p PIVOT(SUM(p.sales) AS sales FOR quarter IN ('Q1' AS Q1, 'Q2' AS Q1))", read={ "bigquery": "SELECT * FROM produce AS p PIVOT(SUM(p.sales) AS sales FOR p.quarter IN ('Q1' AS Q1, 'Q2' AS Q1))", }, ) self.validate_all( "SELECT DATEDIFF(MONTH, '2020-01-01', '2020-03-05')", write={ "databricks": "SELECT DATEDIFF(MONTH, TO_DATE('2020-01-01'), TO_DATE('2020-03-05'))", "hive": "SELECT MONTHS_BETWEEN(TO_DATE('2020-03-05'), TO_DATE('2020-01-01'))", "presto": "SELECT DATE_DIFF('MONTH', CAST(CAST('2020-01-01' AS TIMESTAMP) AS DATE), CAST(CAST('2020-03-05' AS TIMESTAMP) AS DATE))", "spark": "SELECT DATEDIFF(MONTH, TO_DATE('2020-01-01'), TO_DATE('2020-03-05'))", "spark2": "SELECT MONTHS_BETWEEN(TO_DATE('2020-03-05'), TO_DATE('2020-01-01'))", "trino": "SELECT DATE_DIFF('MONTH', CAST(CAST('2020-01-01' AS TIMESTAMP) AS DATE), CAST(CAST('2020-03-05' AS TIMESTAMP) AS DATE))", }, ) for data_type in ("BOOLEAN", "DATE", "DOUBLE", "FLOAT", "INT", "TIMESTAMP"): self.validate_all( f"{data_type}(x)", write={ "": f"CAST(x AS {data_type})", "spark": f"CAST(x AS {data_type})", }, ) self.validate_all( "STRING(x)", write={ "": "CAST(x AS TEXT)", "spark": "CAST(x AS STRING)", }, ) self.validate_all( "CAST(x AS TIMESTAMP)", read={"trino": "CAST(x AS TIMESTAMP(6) WITH TIME ZONE)"} ) self.validate_all( "SELECT DATE_ADD(my_date_column, 1)", write={ "spark": "SELECT DATE_ADD(my_date_column, 1)", "bigquery": "SELECT DATE_ADD(my_date_column, INTERVAL 1 DAY)", }, ) self.validate_all( "AGGREGATE(my_arr, 0, (acc, x) -> acc + x, s -> s * 2)", write={ "trino": "REDUCE(my_arr, 0, (acc, x) -> acc + x, s -> s * 2)", "duckdb": "REDUCE(my_arr, 0, (acc, x) -> acc + x, s -> s * 2)", "hive": "REDUCE(my_arr, 0, (acc, x) -> acc + x, s -> s * 2)", "presto": "REDUCE(my_arr, 0, (acc, x) -> acc + x, s -> s * 2)", "spark": "AGGREGATE(my_arr, 0, (acc, x) -> acc + x, s -> s * 2)", }, ) self.validate_all( "TRIM('SL', 'SSparkSQLS')", write={"spark": "TRIM('SL' FROM 'SSparkSQLS')"} ) self.validate_all( "ARRAY_SORT(x, (left, right) -> -1)", write={ "duckdb": "ARRAY_SORT(x)", "presto": "ARRAY_SORT(x, (left, right) -> -1)", "hive": "SORT_ARRAY(x)", "spark": "ARRAY_SORT(x, (left, right) -> -1)", }, ) self.validate_all( "ARRAY(0, 1, 2)", write={ "bigquery": "[0, 1, 2]", "duckdb": "[0, 1, 2]", "presto": "ARRAY[0, 1, 2]", "hive": "ARRAY(0, 1, 2)", "spark": "ARRAY(0, 1, 2)", }, ) self.validate_all( "SELECT fname, lname, age FROM person ORDER BY age DESC NULLS FIRST, fname ASC NULLS LAST, lname", write={ "clickhouse": "SELECT fname, lname, age FROM person ORDER BY age DESC NULLS FIRST, fname ASC, lname NULLS FIRST", "duckdb": "SELECT fname, lname, age FROM person ORDER BY age DESC NULLS FIRST, fname ASC, lname NULLS FIRST", "postgres": "SELECT fname, lname, age FROM person ORDER BY age DESC, fname ASC, lname NULLS FIRST", "presto": "SELECT fname, lname, age FROM person ORDER BY age DESC NULLS FIRST, fname ASC, lname NULLS FIRST", "hive": "SELECT fname, lname, age FROM person ORDER BY age DESC NULLS FIRST, fname ASC NULLS LAST, lname", "spark": "SELECT fname, lname, age FROM person ORDER BY age DESC NULLS FIRST, fname ASC NULLS LAST, lname", "snowflake": "SELECT fname, lname, age FROM person ORDER BY age DESC, fname ASC, lname NULLS FIRST", }, ) self.validate_all( "SELECT APPROX_COUNT_DISTINCT(a) FROM foo", write={ "duckdb": "SELECT APPROX_COUNT_DISTINCT(a) FROM foo", "presto": "SELECT APPROX_DISTINCT(a) FROM foo", "hive": "SELECT APPROX_COUNT_DISTINCT(a) FROM foo", "spark": "SELECT APPROX_COUNT_DISTINCT(a) FROM foo", }, ) self.validate_all( "MONTH('2021-03-01')", write={ "duckdb": "MONTH(CAST('2021-03-01' AS DATE))", "presto": "MONTH(CAST(CAST('2021-03-01' AS TIMESTAMP) AS DATE))", "hive": "MONTH(TO_DATE('2021-03-01'))", "spark": "MONTH(TO_DATE('2021-03-01'))", }, ) self.validate_all( "YEAR('2021-03-01')", write={ "duckdb": "YEAR(CAST('2021-03-01' AS DATE))", "presto": "YEAR(CAST(CAST('2021-03-01' AS TIMESTAMP) AS DATE))", "hive": "YEAR(TO_DATE('2021-03-01'))", "spark": "YEAR(TO_DATE('2021-03-01'))", }, ) self.validate_all( "'\u6bdb'", write={ "duckdb": "'毛'", "presto": "'毛'", "hive": "'毛'", "spark": "'毛'", }, ) self.validate_all( "SELECT LEFT(x, 2), RIGHT(x, 2)", write={ "duckdb": "SELECT LEFT(x, 2), RIGHT(x, 2)", "presto": "SELECT SUBSTRING(x, 1, 2), SUBSTRING(x, LENGTH(x) - (2 - 1))", "hive": "SELECT SUBSTRING(x, 1, 2), SUBSTRING(x, LENGTH(x) - (2 - 1))", "spark": "SELECT LEFT(x, 2), RIGHT(x, 2)", }, ) self.validate_all( "MAP_FROM_ARRAYS(ARRAY(1), c)", write={ "duckdb": "MAP([1], c)", "presto": "MAP(ARRAY[1], c)", "hive": "MAP(ARRAY(1), c)", "spark": "MAP_FROM_ARRAYS(ARRAY(1), c)", "snowflake": "OBJECT_CONSTRUCT([1], c)", }, ) self.validate_all( "SELECT ARRAY_SORT(x)", write={ "duckdb": "SELECT ARRAY_SORT(x)", "presto": "SELECT ARRAY_SORT(x)", "hive": "SELECT SORT_ARRAY(x)", "spark": "SELECT ARRAY_SORT(x)", }, ) def test_iif(self): self.validate_all( "SELECT IIF(cond, 'True', 'False')", write={"spark": "SELECT IF(cond, 'True', 'False')"}, ) def test_bool_or(self): self.validate_all( "SELECT a, LOGICAL_OR(b) FROM table GROUP BY a", write={"spark": "SELECT a, BOOL_OR(b) FROM table GROUP BY a"}, ) def test_current_user(self): self.validate_all( "CURRENT_USER", write={"spark": "CURRENT_USER()"}, ) self.validate_all( "CURRENT_USER()", write={"spark": "CURRENT_USER()"}, ) def test_transform_query(self): self.validate_identity("SELECT TRANSFORM(x) USING 'x' AS (x INT) FROM t") self.validate_identity( "SELECT TRANSFORM(zip_code, name, age) USING 'cat' AS (a, b, c) FROM person WHERE zip_code > 94511" ) self.validate_identity( "SELECT TRANSFORM(zip_code, name, age) USING 'cat' AS (a STRING, b STRING, c STRING) FROM person WHERE zip_code > 94511" ) self.validate_identity( "SELECT TRANSFORM(name, age) ROW FORMAT DELIMITED FIELDS TERMINATED BY ',' LINES TERMINATED BY '\n' NULL DEFINED AS 'NULL' USING 'cat' AS (name_age STRING) ROW FORMAT DELIMITED FIELDS TERMINATED BY '@' LINES TERMINATED BY '\n' NULL DEFINED AS 'NULL' FROM person" ) self.validate_identity( "SELECT TRANSFORM(zip_code, name, age) ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe' WITH SERDEPROPERTIES ('field.delim'='\t') USING 'cat' AS (a STRING, b STRING, c STRING) ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe' WITH SERDEPROPERTIES ('field.delim'='\t') FROM person WHERE zip_code > 94511" ) self.validate_identity( "SELECT TRANSFORM(zip_code, name, age) USING 'cat' FROM person WHERE zip_code > 94500" ) def test_insert_cte(self): self.validate_all( "INSERT OVERWRITE TABLE table WITH cte AS (SELECT cola FROM other_table) SELECT cola FROM cte", write={ "spark": "WITH cte AS (SELECT cola FROM other_table) INSERT OVERWRITE TABLE table SELECT cola FROM cte", "spark2": "WITH cte AS (SELECT cola FROM other_table) INSERT OVERWRITE TABLE table SELECT cola FROM cte", "databricks": "WITH cte AS (SELECT cola FROM other_table) INSERT OVERWRITE TABLE table SELECT cola FROM cte", }, ) def test_explode_to_unnest(self): self.validate_all( "SELECT EXPLODE(x) FROM tbl", write={ "bigquery": "SELECT IF(pos = pos_2, col, NULL) AS col FROM tbl, UNNEST(GENERATE_ARRAY(0, GREATEST(ARRAY_LENGTH(x)) - 1)) AS pos CROSS JOIN UNNEST(x) AS col WITH OFFSET AS pos_2 WHERE pos = pos_2 OR (pos > (ARRAY_LENGTH(x) - 1) AND pos_2 = (ARRAY_LENGTH(x) - 1))", "presto": "SELECT IF(pos = pos_2, col) AS col FROM tbl, UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(x)))) AS _u(pos) CROSS JOIN UNNEST(x) WITH ORDINALITY AS _u_2(col, pos_2) WHERE pos = pos_2 OR (pos > CARDINALITY(x) AND pos_2 = CARDINALITY(x))", "spark": "SELECT EXPLODE(x) FROM tbl", }, ) self.validate_all( "SELECT EXPLODE(col) FROM _u", write={ "bigquery": "SELECT IF(pos = pos_2, col_2, NULL) AS col_2 FROM _u, UNNEST(GENERATE_ARRAY(0, GREATEST(ARRAY_LENGTH(col)) - 1)) AS pos CROSS JOIN UNNEST(col) AS col_2 WITH OFFSET AS pos_2 WHERE pos = pos_2 OR (pos > (ARRAY_LENGTH(col) - 1) AND pos_2 = (ARRAY_LENGTH(col) - 1))", "presto": "SELECT IF(pos = pos_2, col_2) AS col_2 FROM _u, UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(col)))) AS _u_2(pos) CROSS JOIN UNNEST(col) WITH ORDINALITY AS _u_3(col_2, pos_2) WHERE pos = pos_2 OR (pos > CARDINALITY(col) AND pos_2 = CARDINALITY(col))", "spark": "SELECT EXPLODE(col) FROM _u", }, ) self.validate_all( "SELECT EXPLODE(col) AS exploded FROM schema.tbl", write={ "presto": "SELECT IF(pos = pos_2, exploded) AS exploded FROM schema.tbl, UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(col)))) AS _u(pos) CROSS JOIN UNNEST(col) WITH ORDINALITY AS _u_2(exploded, pos_2) WHERE pos = pos_2 OR (pos > CARDINALITY(col) AND pos_2 = CARDINALITY(col))", }, ) self.validate_all( "SELECT EXPLODE(ARRAY(1, 2))", write={ "bigquery": "SELECT IF(pos = pos_2, col, NULL) AS col FROM UNNEST(GENERATE_ARRAY(0, GREATEST(ARRAY_LENGTH([1, 2])) - 1)) AS pos CROSS JOIN UNNEST([1, 2]) AS col WITH OFFSET AS pos_2 WHERE pos = pos_2 OR (pos > (ARRAY_LENGTH([1, 2]) - 1) AND pos_2 = (ARRAY_LENGTH([1, 2]) - 1))", "presto": "SELECT IF(pos = pos_2, col) AS col FROM UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(ARRAY[1, 2])))) AS _u(pos) CROSS JOIN UNNEST(ARRAY[1, 2]) WITH ORDINALITY AS _u_2(col, pos_2) WHERE pos = pos_2 OR (pos > CARDINALITY(ARRAY[1, 2]) AND pos_2 = CARDINALITY(ARRAY[1, 2]))", }, ) self.validate_all( "SELECT POSEXPLODE(ARRAY(2, 3)) AS x", write={ "bigquery": "SELECT IF(pos = pos_2, x, NULL) AS x, IF(pos = pos_2, pos_2, NULL) AS pos_2 FROM UNNEST(GENERATE_ARRAY(0, GREATEST(ARRAY_LENGTH([2, 3])) - 1)) AS pos CROSS JOIN UNNEST([2, 3]) AS x WITH OFFSET AS pos_2 WHERE pos = pos_2 OR (pos > (ARRAY_LENGTH([2, 3]) - 1) AND pos_2 = (ARRAY_LENGTH([2, 3]) - 1))", "presto": "SELECT IF(pos = pos_2, x) AS x, IF(pos = pos_2, pos_2) AS pos_2 FROM UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(ARRAY[2, 3])))) AS _u(pos) CROSS JOIN UNNEST(ARRAY[2, 3]) WITH ORDINALITY AS _u_2(x, pos_2) WHERE pos = pos_2 OR (pos > CARDINALITY(ARRAY[2, 3]) AND pos_2 = CARDINALITY(ARRAY[2, 3]))", }, ) self.validate_all( "SELECT POSEXPLODE(x) AS (a, b)", write={ "presto": "SELECT IF(pos = a, b) AS b, IF(pos = a, a) AS a FROM UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(x)))) AS _u(pos) CROSS JOIN UNNEST(x) WITH ORDINALITY AS _u_2(b, a) WHERE pos = a OR (pos > CARDINALITY(x) AND a = CARDINALITY(x))", }, ) self.validate_all( "SELECT POSEXPLODE(ARRAY(2, 3)), EXPLODE(ARRAY(4, 5, 6)) FROM tbl", write={ "bigquery": "SELECT IF(pos = pos_2, col, NULL) AS col, IF(pos = pos_2, pos_2, NULL) AS pos_2, IF(pos = pos_3, col_2, NULL) AS col_2 FROM tbl, UNNEST(GENERATE_ARRAY(0, GREATEST(ARRAY_LENGTH([2, 3]), ARRAY_LENGTH([4, 5, 6])) - 1)) AS pos CROSS JOIN UNNEST([2, 3]) AS col WITH OFFSET AS pos_2 CROSS JOIN UNNEST([4, 5, 6]) AS col_2 WITH OFFSET AS pos_3 WHERE (pos = pos_2 OR (pos > (ARRAY_LENGTH([2, 3]) - 1) AND pos_2 = (ARRAY_LENGTH([2, 3]) - 1))) AND (pos = pos_3 OR (pos > (ARRAY_LENGTH([4, 5, 6]) - 1) AND pos_3 = (ARRAY_LENGTH([4, 5, 6]) - 1)))", "presto": "SELECT IF(pos = pos_2, col) AS col, IF(pos = pos_2, pos_2) AS pos_2, IF(pos = pos_3, col_2) AS col_2 FROM tbl, UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(ARRAY[2, 3]), CARDINALITY(ARRAY[4, 5, 6])))) AS _u(pos) CROSS JOIN UNNEST(ARRAY[2, 3]) WITH ORDINALITY AS _u_2(col, pos_2) CROSS JOIN UNNEST(ARRAY[4, 5, 6]) WITH ORDINALITY AS _u_3(col_2, pos_3) WHERE (pos = pos_2 OR (pos > CARDINALITY(ARRAY[2, 3]) AND pos_2 = CARDINALITY(ARRAY[2, 3]))) AND (pos = pos_3 OR (pos > CARDINALITY(ARRAY[4, 5, 6]) AND pos_3 = CARDINALITY(ARRAY[4, 5, 6])))", }, ) self.validate_all( "SELECT col, pos, POSEXPLODE(ARRAY(2, 3)) FROM _u", write={ "presto": "SELECT col, pos, IF(pos_2 = pos_3, col_2) AS col_2, IF(pos_2 = pos_3, pos_3) AS pos_3 FROM _u, UNNEST(SEQUENCE(1, GREATEST(CARDINALITY(ARRAY[2, 3])))) AS _u_2(pos_2) CROSS JOIN UNNEST(ARRAY[2, 3]) WITH ORDINALITY AS _u_3(col_2, pos_3) WHERE pos_2 = pos_3 OR (pos_2 > CARDINALITY(ARRAY[2, 3]) AND pos_3 = CARDINALITY(ARRAY[2, 3]))", }, )