Fix phantom fields for boolean literals

[crisptrutski]

Pretty straight forward. Also adds a failing test with override for CTEs.

[piranha]

I guess I can approve seems this doesn't seem a regression, do we need to track that as a separate issue?

[piranha]

I think the result here is incorrect, because x is from both t1 and t2, since cte is not referenced anywhere:

testmetabase=# with cte as (select id from core_user limit 1) select id, name from report_card limit 2;
 id |              name              
----+--------------------------------
  2 | Orders with People
  1 | Aerodynamic Copper Knife trend
(2 rows)

[crisptrutski]

Whoops, I meant for cte to be in scope.

mysql > with cte as (select id from core_user limit 1)
        select id name from report_card join cte limit 2;

ERROR 1052 (23000): Column 'id' in field list is ambiguous

Since Macaw is garbage-in-garbage-out for invalid queries, I think the correct conclusion is that id come from core_user only.

As for the actual query I wrote here, I guess you're right that in the short term returning both qualifications makes the most sense. One could imagine it doing some "dead scope analysis" however and excluding both the table and fields for the unused CTE.

[crisptrutski]

I guess I can approve seems this doesn't seem a regression, do we need to track that as a separate issue?

I see all the CTE and subselect defects being nested under metabase/metabase#42586

Hmm, going to start adding ticket references to each :override for good measure.

[crisptrutski]

Using links instead of just #1212312 to disambiguate whether it's tracked in metabase or macaw.