On Thu, Aug 20, 2026 at 12:11 PM Henson Choi <assam258@gmail.com> wrote:
> From your message of May 13:
>
> > Does this structure work for you, or would you prefer a different
> > approach?
>
> It works. 0001 baseline capture, then 0002 with the fix, is a good
> idea.
+1
> That is what I had asked for. I have now reviewed v3 and I think it
> should go in as it stands. CI is green, so I am marking this Ready for
> Committer.
Doesn't that just contradict what you said above? I've set it back to
waiting on author.
While looking, I've found another behavior change that should be
called out -- data type width calculations can change query plans:
createdb -T template0 -E EUC_KR --locale=C euc_kr
psql -d euc_kr -X <<'EOF'
create table t(c char(10), v varchar(10));
explain select c from t;
explain select v from t;
EOF
master:
QUERY PLAN
-----------------------------------------------------
Seq Scan on t (cost=0.00..18.50 rows=850 width=34)
(1 row)
QUERY PLAN
-----------------------------------------------------
Seq Scan on t (cost=0.00..18.50 rows=850 width=33)
(1 row)
v3:
QUERY PLAN
------------------------------------------------------
Seq Scan on t (cost=0.00..20.70 rows=1070 width=24)
(1 row)
QUERY PLAN
------------------------------------------------------
Seq Scan on t (cost=0.00..20.70 rows=1070 width=24)
(1 row)
--
John Naylor
Amazon Web Services
