Re: Is postorder tree traversal possible with recursive CTE's?

On 06/19/2018 02:05 PM, Alban Hertroys wrote:
> On the more theoretical front: The question remains whether it is possible to calculate fields in post-order tree traversal. I think that would be a semantically proper way to express this type of problem and it wouldn't need the kinds of pre/post-processing that after-the-fact aggregation (like in above solution) requires. So, leaner, and probably faster.
> That implies that the SQL committee thought of the possibility in the first place though, which I'm beginning to doubt...

If this interests you, you might enjoy this StackOverflow question:

https://stackoverflow.com/questions/35956486/generate-nested-json-with-couting-in-postgresql

Briefly, how do you construct a nested JSON structure from a recursive
CTE? The only answers at that link rely on plpgsql, but of course that
is cheating. :-) I took a stab at it a couple years ago but couldn't
figure it out, and it seemed like post-order processing was exactly the
missing piece.

If anyone has any ideas I'd be intrigued to hear them!

--
Paul ~{:-)
pj(at)illuminatedcomputing(dot)com