Re: Manipulating complex types as non-contiguous structures in-memory

[ this is addressing a tangential point ... ]

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> * Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
>> * Although I said above that everything owned by a deserialized object
>> has to live in a single memory context, I do have ideas about relaxing
>> that. The core idea would be to invent a "memory context reset/delete
>> callback" feature in mcxt.c. Then a deserialized object could register
>> such a callback on its own memory context, and use the callback to clean
>> up resources outside its context.

> Being able to register a callback to be used on deletion of the context
> would certainly be very nice and strikes me as pretty independent of the
> rest of this. You've probably thought of this already, but registering
> the callback should probably allow the caller to pass in a pointer to be
> passed back to the callback function when the delete happens, so that
> there's a place for the metadata to be stored about what the callback
> function needs to clean up when it's called.

Yeah, there would likely be use-cases for that outside of deserialized
objects. I could submit a separate patch for that now, but I'm hesitant
to add a mechanism without any use-case in the same patch. But maybe we
could find a caller somewhere in the core aggregate code --- there are
some aggregates that need cleanup callbacks already, IIRC, and maybe we
could change them to use a memory context callback instead of whatever
they're doing now.

regards, tom lane