I am writing the library that implements the syntax I described. I will show working code soon (or right now upon request). I am sorely tempted to overload the comma operator for "select", "from" and "in". I know the usual caveats (low precedence, default behavior exists for UDTs). The alternatives for open lists are : use the preprocessor to crank out overloads like in the MPL ; build lists one at a time [i.e. select(t1.id)(t2.name)(t3.age).from(t1)(t2)(t3).where(t1.id.in(1)(2))]...and overload the comma operator on well-targeted types [e.g. template<typename T> operator ,(expression<T>, expression<T>)]. And perhaps make the comma operator available but in a namespace (gives choice but usually alternative syntaxes are a bad idea). Opinions ? J-L