On Mon, Dec 20, 2021 at 5:17 AM Amit Langote <amitlangot...@gmail.com> wrote:
> On Mon, Dec 20, 2021 at 6:19 PM Zhihong Yu <z...@yugabyte.com> wrote: > > On Sun, Dec 19, 2021 at 10:20 PM Amit Langote <amitlangot...@gmail.com> > wrote: > >> > >> On Mon, Dec 20, 2021 at 2:00 PM Corey Huinker <corey.huin...@gmail.com> > wrote: > >> > Sorry for the delay. This patch no longer applies, it has some > conflict with d6f96ed94e73052f99a2e545ed17a8b2fdc1fb8a > >> > >> Thanks Corey for the heads up. Rebased with some cosmetic adjustments. > >> > > Hi, > > > > + Assert(partidx < 0 || partidx < partdesc->nparts); > > + partoid = partdesc->oids[partidx]; > > > > If partidx < 0, do we still need to fill out partoid and is_leaf ? It > seems we can return early based on (should call table_close(rel) first): > > > > + /* No partition found. */ > > + if (partidx < 0) > > + return NULL; > > Good catch, thanks. Patch updated. > > Hi, + int lockflags = 0; + TM_Result test; + + lockflags = TUPLE_LOCK_FLAG_LOCK_UPDATE_IN_PROGRESS; The above assignment can be meged with the line where variable lockflags is declared. + GetUserIdAndSecContext(&save_userid, &save_sec_context); save_userid -> saved_userid save_sec_context -> saved_sec_context + * the transaction-snapshot mode. If we didn't push one already, do didn't push -> haven't pushed For ri_PerformCheck(): + bool source_is_pk = true; It seems the value of source_is_pk doesn't change - the value true can be plugged into ri_ExtractValues() calls directly. Cheers
