Re: Concurrent psql patch

Tom Lane wrote:
> Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
>
>> Tom Lane wrote:
>>
>>> If there's more needed here, let's see an official API change,
>>> not a hack.
>>>
>
>
>> Well, I guess the obvious API would be something like:
>> PGAsyncStatusType PQasyncStatus(const PGconn *conn);
>>
>
> That would mean exposing PGAsyncStatusType, which doesn't seem like an
> amazingly good idea either. What is it that the patch actually wants
> to do?
>

The relevant snippet in command.c says:

/* If we have async_delay set then we need to allow up to that many
* milliseconds for any responses to come in on *either* connection.
* This ensures that results are printed in a relatively deterministic
* order for regression tests. Note that we only have to allow for n
* milliseconds total between the two connections so we don't reset the
* timer for the second wait.
*
* XXX this should maybe be changed to a select/poll loop instead
*/
if (pset.async_delay > 0 && pset.c->db)
{
GETTIMEOFDAY(&start);
do {
if (pset.c->db->asyncStatus != PGASYNC_BUSY)
{
break;
}
if (CheckQueryResults())
{
ReadQueryResults();
break;
}
pg_usleep(10000);
GETTIMEOFDAY(&now);
} while (DIFF_MSEC(&now, &start) < pset.async_delay);
}

pset.c = &cset[slot-1];
printf(_("[%d] You are now connected to database \"%s\"\n"), slot,
PQdb(pset.c->db));

if (pset.async_delay > 0 && pset.c->db)
{
do {
if (pset.c->db->asyncStatus != PGASYNC_BUSY)
break;
if (CheckQueryResults()) {
ReadQueryResults();
break;
}
pg_usleep(10000);
GETTIMEOFDAY(&now);
} while (DIFF_MSEC(&now, &start) < pset.async_delay);
}

and in mainloop.c it's used like this:

if (pset.c->db && pset.c->db->asyncStatus != PGASYNC_IDLE &&
CheckQueryResults()) {
ReadQueryResults();
/* If we processed a query cancellation cancel_pressed may
still be
* set and will interrupt the processing of the next command
unless
* we start the main loop over to reset it. */
if (cancel_pressed)
break;
}

cheers

andrew