Hacker News Reader

Comment by grzm | Hacker News Reader

Would you elaborate on how to select data in the custom order efficiently? What you're describing can't be easily indexed, if I understand you correctly. The strategy outlined in the submission ensures minimal table rewrites on updates (insertion, reordering, deletion) as well as efficient retrieval.

View on HN

  with recursive todo_list_sorted as (
  	select
  		*
  	from
  		todo_list tl1 where prev_id is null
  	union all
  	select
  		tl1.*	
  	from
  		todo_list tl1
  	join
  		todo_list_sorted tl2 on tl1.prev_id = tl2.id
  )
  select * from todo_list_sorted

Yes, there are ways to do it. The question is how to do so efficiently. Recursive queries don't benefit from indexes as efficiently as, for example, the solution outlined in the submission.

Well I think this achieves minimal rewrites, but you do lose a bit on indexing. You can't retrieve by numeric position index for example. But reproducing the order is just a matter of starting from the head (where PrecededById is null) or from any point and then repeatedly selecting the entry whose PrecededById = the current ID.

That's more than a bit on indexing. Following a linked list through repeated selections is not nearly as efficient as an index scan.

Yeah that's true. Oh well, fractions it is then!

Reply on news.ycombinator.com