Automated query fuzzing

[systay]

We need better ways of comparing the results produced by Vitess and vanilla MySQL. It would be great if we could produce random queries and be able to run them against both mysql and vitess and check that the results are the same.

Prioritised list of query patterns to test:

• Inner joins
• Grouping
• LIMIT
• Aggregations
• ORDER BY
• Outer joins
• Random expressions
• OFFSET
• WHERE predicates
• HAVING
• DERIVED tables
• Column aliases
• Prevent LIMIT with no ORDER BY
• Possibility to ignore all problems except invalid results and panics
• Derived table columns without alias
• Complex aggregation arguments (SUM(a + b))
• Complex expressions containing aggregations (SUM(a)+SUM(b))
• Minimizing error
• SUBQUERIES (EXISTS, IN, SCALAR)
• Distinct aggregation arguments (count(distinct x)
• UNION
• Tuple comparisons
• Tuple subqueries
• Complicate schema

[arvind-murty]

Current random query production: PR #13260

[arvind-murty]

Left join seems to work fine, but right join fails a bunch

[arvind-murty]

Current random query production: PR #13260

Now PR #13460