| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Jelte Fennema <me(at)jeltef(dot)nl> |
| Cc: | PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: Support prepared statement invalidation when result types change |
| Date: | 2024-07-24 15:39:56 |
| Message-ID: | 1238027.1721835596@sss.pgh.pa.us |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Jelte Fennema <me(at)jeltef(dot)nl> writes:
> The cached plan for a prepared statements can get invalidated when DDL
> changes the tables used in the query, or when search_path changes.
> ...
> However, we would throw an error if the the result of the query is of a
> different type than it was before:
> ERROR: cached plan must not change result type
Yes, this is intentional.
> This patch starts to allow a prepared statement to continue to work even
> when the result type changes.
What this is is a wire protocol break. What if the client has
previously done a Describe Statement on the prepared statement?
We have no mechanism for notifying it that that information is
now falsified. The error is thrown to prevent us from getting
into a situation where we'd need to do that.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2024-07-24 15:43:43 | Re: Slow catchup of 2PC (twophase) transactions on replica in LR |
| Previous Message | Peter Eisentraut | 2024-07-24 15:39:00 | Re: Detect buffer underflow in get_th() |