* Fixed SELECT cases to actually read data; * Using hash instead of rand() to make tests more reproducible; * Set number of rows to 1000000 for both INSERT and SELECT queries; * Using {type} substitution for fill_query.