Dimitri Fontaine <dimitri@2ndQuadrant.fr> writes:
> So I have a Node *parsetree containing some CHECK and DEFAULT raw
> expressions to work with. Those can reference non existing tables,
> either to-be-created or already-dropped.
Why don't you just pass the original query string, instead of writing
a mass of maintenance-requiring new code to reproduce it?
This would require (1) making sure the query string is still available
where needed. I think we are 99% of the way there but maybe not 100%.
(2) being able to identify the substring corresponding to the current
command, when we're processing a multi-command string. The parser could
easily provide that, I think --- we've just never insisted that it do
so before.
regards, tom lane