Eric Blake <ebl...@redhat.com> writes: > On 05/24/2013 07:08 AM, Luiz Capitulino wrote: >> This sounds like premature optimization to me, but I wonder if instead >> of cluttering commands with arguments to do the filtering we could add >> some standard way of doing this in the QAPI. > > Maybe we could make QAPI support generic filtering for all query-* > commands. But there's still a cost with making all query-* commands > malloc the space for an entire list only to then have a QAPI layer on > top of it do filtering before handing back the single element over the > wire, compared to passing the filtering down to the query-* > implementation to do the filtering in place.
I'd expect the time spent on malloc to be dwarved several times over by I/O latency. If we worry about malloc impacting our latency, then we should not be using QAPI! It's really, really malloc-happy. In QMP, we generally don't. >> It was you who suggested a filter command? > > No, Stefan suggested it on v1: > https://lists.gnu.org/archive/html/qemu-devel/2013-04/msg03102.html So far, I'm with Luiz here.