maropu / fuzz-testing-for-spark Goto Github PK

SQLSmith use grammar.cc and expr.cc to generates SQL fuzz. Among generated ones, I found a set of SQL statements below that Spark currently does not support;

• OFFSET clauses (SPARK-28330)
• TEXT type name
• main default database
• LATERAL subqueries
• 1 for a true literal and 0 for a false literal
• TABLESAMPLE <method_name> (sample_size)
• UPDATE/MERGE/DELETE statements
• JOIN syntax

postgres=# SELECT * FROM t1
postgres=#     INNER JOIN t2 
postgres=#         LEFT JOIN t3
postgres=#         ON t2.v2 = t3.v3
postgres=#     ON t1.v1 = t2.v2;
 v1 | v2 | v3 
----+----+----
(0 rows)

These issues above have been addressed in lib/sqlsmith-native/sqlsmith-1.2.1.patch.