I think Dan's idea is quite good. Remember the important tenets of the query dispatcher:
- we need to support multiple DBs. Possibly more than just Oracle and PG in the future. This means giving up some coding convenience. Such is life: give up short-term convenience for long-term maintainability.
- we should make it possible for DBAs to optimize queries without digging into Tcl code.
- we should cleanly separate Tcl from SQL so that the query dispatcher can be ported to other languages (for those Java and C# enthusiasts) without any dependence on the Tcl language.
Dan's proposal significantly improves the maintainability of the code with minimal programmer hassle, and without breaking the above tenets. I'd love to hear more comments on this particular proposal, as I am seriously considering implementing it in some form.
