pointer scope and memory contexts

From: "Tim Keitt" <tkeitt(at)keittlab(dot)org>
To: pgsql-hackers(at)postgresql(dot)org
Subject: pointer scope and memory contexts
Date: 2008-11-06 01:31:57
Message-ID: 6262c54c0811051731s90245a1nea1321652d77c3d@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

I am working on a set returning function and have a question about
switching memory contexts. Basically, what I want to know is whether
memory allocated in one context can be referenced when a different
context is current. The docs give examples like:

if (SRF_IS_FIRSTCALL())
{
funcctx = SRF_FIRSTCALL_INIT();
oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);

/* allocate some memory that needs to span calls */
my_ptr = (a_type *) palloc(size);

MemoryContextSwitchTo(oldcontext);
}

funcctx = SRF_PERCALL_SETUP();

/* can I now dereference here? Is it necessary to manually switch
to multi_call_memory_ctx? */
some_function(my_ptr[0]);
...

In a related question, do I only need to switch to
multi_call_memory_ctx when allocating memory that needs to be
referenced in subsequent calls? Would the following make sense?

if (SRF_IS_FIRSTCALL())
{
funcctx = SRF_FIRSTCALL_INIT();

/* this will be deallocated at the end of the first call, right? */
scratch_ptr = (a_type *) palloc(size);

oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
/* allocate some memory that needs to span calls */
my_ptr = (a_type *) palloc(size);
MemoryContextSwitchTo(oldcontext);

/* can I dereference my_ptr here? */
func_called_once(scratch_ptr, my_ptr);

...

}

/* per call stuff follows... */

In other words, if I have allocations that only are needed during the
first call, can I allocate them outside of the multi_call_memory_ctx
(so that they do not live past the first call) and only allocate the
memory needed for multiple calls within the multi_call_memory_ctx?

One last question: if I call SPI_finish, on the first call, do I need
to switch contexts in the per-call section? (I saw some example code
that suggested one needs to switch contexts back to
multi_call_memory_ctx after SPI_finish.)

Thanks.

THK

--
Timothy H. Keitt
http://www.keittlab.org/

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Gregory Stark 2008-11-06 01:38:26 Re: [WIP] In-place upgrade
Previous Message Bernd Helmle 2008-11-06 01:09:07 Re: Patch for ALTER DATABASE WITH TABLESPACE