> b. Extract the values of expressions contained within expression indexes where practical, rather than recomputing the expression. > c. The NOT NULL and IS NUL

> b. Extract the values of expressions contained within expression indexes where practical, rather than recomputing the expression. > c. The NOT NULL and IS NULL operators (and their equivalents) avoid loading the content of large strings and BLOB values from disk. > d. Avoid materializing a view on which a full scan is performed exactly once. Use and discard the rows of the view as they are computed. > e. Allow flattening of a subquery that is the right-hand operand of a LEFT JOIN in an aggregate query. Gulp. When writing SQL I tend to assume these kinds of things have always been done... I should probably favor my own logical optimizations when I can. (Of course, nothing beats measuring/profiling if you really need to optimize.)