Did an additional test on Supabase Docs RLS guide Advanced policies

Using method 5 for the comments policy and method 2 for the posts policy had significant impact.
See https://github.com/GaryAustin1/RLS-Perfomance for methods

Tested with 100K comments, 100k posts and both 10k and 1K posts by the test user.

Table	Example Policy	(Method) Policy	Posts by User
comments	204ms/773ms(JS)	(5) 43ms/43ms(JS)	10000
comments	190ms	(5) 29ms	1000
posts	173ms/841ms(JS)	(2) 12ms/38ms(JS)	10000

comments table policy
Example: exists (select 1 from posts where posts.id = comments.post_id)
Method 5: post_id in (select id from posts where creator_id = (select auth.uid()))

posts table policy
Example: select auth.uid() = posts.creator_id
Method 2: (select auth.uid()) = posts.creator_id

Notes: At some point method 5 does not make much sense as the list it builds up can get very large. The method clearly makes sense for most cases, especially if the join table is smaller than the main table, or the row being filtered on is a small subset. But I tested 50K Posts by the user and the optimizer still chose that method, but time was up to 80ms. At 90K posts by the user the optimizer picked another method, but did not break. Note this article https://www.dbi-services.com/blog/what-is-the-maximum-in-list-size-in-postgresql/ had the list up to 2M for the in and then at 3M ran out of memory. So at a minimum probably need a note to think about the size of the in list when it is some % more of the main table rows.

Some other "bad" examples from that page...

create policy "Team members can update team details if they belong to the team."
  on teams
  for update using (
    auth.uid() in (
      select user_id from members
      where team_id = id
    )
  );

This works because it is on an update, but probably better to show it as:

create policy "Team members can update team details if they belong to the team."
  on teams
  for update using (
    id in (
      select team_id from members
      where user_id = auth.uid()
    )
  );

As that would make a huge difference if that policy is used for select also.