Re: Concern about memory management with SRFs

Tom Lane wrote:
> I think we should document that any memory that is allocated in the
> first call for use in subsequent calls must come from the memory context
> saved in FuncCallContext (and let's choose a more meaningful name than
> fmctx, please). This would mean adding code like
>
> oldcontext = MemoryContextSwitchTo(funcctx->fmctx);
>
> ...
>
> MemoryContextSwitchTo(oldcontext);
>
> around the setup code that follows SRF_FIRSTCALL_INIT. Then it would be
> safe for nodeFunctionscan.c to do a reset before each function call.

That sounds like a good plan.

But can/should we wrap those calls in either existing or new macros? Or
is it better to have the function author keenly aware of the memory
management details? I tend to think the former is better.

Maybe SRF_FIRSTCALL_INIT()(init_MultiFuncCall()) should:
- save CurrentMemoryContext to funcctx->per_call_memory_ctx
(new member of the struct)
- save fcinfo->flinfo->fn_mcxt to funcctx->multi_call_memory_ctx
(nee funcctx->fmctx)
- leave fcinfo->flinfo->fn_mcxt as the current memory context when it
exits

Then SRF_PERCALL_SETUP() (per_MultiFuncCall()) can change back to
funcctx->per_call_memory_ctx.

Would this work?

Joe