Thread: Example of Asynchronous Query Processing in C?

Greetings,

I was reading through the docs for libpq and in the Asynchronous Query
Processing section there is mention of an example program that demonstrates
the use of select(2) to wait for the backend.  I could not find the
example, does anyone have one?  I understand what needs to be done, however
it is always nice to have some sort of reference program to follow.

What I am hoping to accomplish is to have my frontend accept a query, pass
it to the backend, and build the output based on the query.  I would also
like the user to be able to limit the number of records (tuples) returned
and cancel the query if necessary; hence the use of the async functions and
select(2).

I know psql does this and I'm about to go poking around in its source, but
the Asynchronous Query Processing doc mentions the psql uses the PQexec
function in conjunction with PQrequestCancel, so I'm not sure if it will
give me the example I'm looking for.

If someone has a function that does what I'm looking for, hey that would be
swell ;), or just a simple example of async processing would be nice too.

Thanks,
Matthew

Matthew Hagerty <matthew@venux.net> writes:
> I was reading through the docs for libpq and in the Asynchronous Query
> Processing section there is mention of an example program that demonstrates
> the use of select(2) to wait for the backend.  I could not find the
> example, does anyone have one?

Hmm, it does promise one doesn't it?  What I was actually intending to
do was just modify the second sample program to wait for a NOTIFY with
select() rather than sleep().  But I forgot :-(

As I was explaining to someone else recently, a reasonable sample
program for async query processing is hard to come up with.  The only
reason you'd do it is because your app has something else it needs to
do concurrently, which is tough to demonstrate in a simple program.

The actual use of select() is not hard --- you can use pqWait() in
interfaces/libpq/fe-misc.c as an example.  The issue that's hard to
show in a small program is what sort of control structure all this
is likely to be embedded in.

> What I am hoping to accomplish is to have my frontend accept a query, pass
> it to the backend, and build the output based on the query.  I would also
> like the user to be able to limit the number of records (tuples) returned
> and cancel the query if necessary; hence the use of the async functions and
> select(2).

Limiting the tuple count will be a feature in 6.5 but is not really
possible without backend support.  As for cancel, if you can issue the
cancel from a signal handler like psql does, you don't really need the
async query interface ...

            regards, tom lane