Re: display hot standby state in psql prompt
Jim Jones <jim.jones@uni-muenster.de>
From: Jim Jones <jim.jones@uni-muenster.de>
To: Nathan Bossart <nathandbossart@gmail.com>
Cc: Fujii Masao <masao.fujii@gmail.com>,
Srinath Reddy Sadipiralla <srinath2133@gmail.com>,
Greg Sabino Mullane <htamfids@gmail.com>,
PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>,
Chao Li <li.evan.chao@gmail.com>
Date: 2025-10-31T21:00:51Z
Lists: pgsql-hackers
Commits
Same data as JSON:
GET /api/v1/messages/:b64id/commits
the thread's linked commits as JSON, with link sources.
API reference →
-
psql: Add %i prompt escape to indicate hot standby status.
- dddbbc253b92 19 (unreleased) landed
-
Mark search_path as GUC_REPORT
- 28a1121fd912 18.0 cited
Attachments
- v7-0001-Add-i-prompt-escape-to-indicate-server-read-only-.patch (text/x-patch) patch v7-0001
On 30/10/2025 11:16, Jim Jones wrote:
> if (!hs || !ro)
> strlcpy(buf, _("unknown"), sizeof(buf));
> else if (strcmp(hs, "on") == 0 || strcmp(ro, "on") == 0)
> strlcpy(buf, _("read-only"), sizeof(buf));
> else
> {
> const char *tr = NULL;
> PGresult *res;
>
> res = PQexec(pset.db, "SHOW transaction_read_only");
> if (PQresultStatus(res) == PGRES_TUPLES_OK &&
> PQntuples(res) == 1)
> tr = PQgetvalue(res, 0, 0);
>
> if (!tr)
> strlcpy(buf, _("unknown"), sizeof(buf));
> else if (strcmp(tr, "on") == 0)
> strlcpy(buf, _("read-only"), sizeof(buf));
> else
> strlcpy(buf, _("read/write"), sizeof(buf));
>
> PQclear(res);
> }
While reviewing another patch, I had another idea to further minimise
the overhead of checking transaction_read_only, namely, to check it only
when within a transaction block:
if (!hs || !ro)
strlcpy(buf, _("unknown"), sizeof(buf));
else if (strcmp(hs, "on") == 0 || strcmp(ro, "on") == 0)
strlcpy(buf, _("read-only"), sizeof(buf));
else
{
PGTransactionStatusType tstatus = PQtransactionStatus(pset.db);
/*
* Check transaction_read_only only when in a transaction
* block. When idle (not in a transaction), the value of
* transaction_read_only is the same as
* default_transaction_read_only, which we already checked
* above. Avoiding the query improves performance,
* especially for prompt redisplays.
*/
if (tstatus == PQTRANS_IDLE)
strlcpy(buf, _("read/write"), sizeof(buf));
else
{
/* In a transaction block, need to check transaction_read_only */
const char *tr = NULL;
PGresult *res;
res = PQexec(pset.db, "SHOW transaction_read_only");
if (PQresultStatus(res) == PGRES_TUPLES_OK && PQntuples(res) == 1)
tr = PQgetvalue(res, 0, 0);
if (!tr)
strlcpy(buf, _("unknown"), sizeof(buf));
else if (strcmp(tr, "on") == 0)
strlcpy(buf, _("read-only"), sizeof(buf));
else
strlcpy(buf, _("read/write"), sizeof(buf));
PQclear(res);
}
}
The idea is to skip the test if tstatus == PQTRANS_IDLE, which indicates
that it is not in a transaction block, making the check for
transaction_read_only unnecessary.
Thoughts on this approach?
v7 attached.
Best, Jim