On Tue, Jul 16, 2019 at 10:22 AM Andres Freund <andres@anarazel.de> wrote:
Hi,
On 2019-07-11 17:27:46 -0700, Ashwin Agrawal wrote: > Please find attached the patch to remove IndexBuildCallback's dependency on > HeapTuple, as discussed. Changed to have the argument as ItemPointer > instead of HeapTuple. Other larger refactoring if feasible for > index_build_range_scan API itself can be performed as follow-up changes.
> From f73b0061795f0c320f96ecfed0c0602ae318d73e Mon Sep 17 00:00:00 2001 > From: Ashwin Agrawal <aagrawal@pivotal.io> > Date: Thu, 11 Jul 2019 16:58:50 -0700 > Subject: [PATCH v1] Remove IndexBuildCallback's dependency on HeapTuple. > > With IndexBuildCallback taking input as HeapTuple, all table AMs > irrespective of storing the tuples in HeapTuple form or not, are > forced to construct HeapTuple, to insert the tuple in Index. Since, > only thing required by the index callbacks is TID and not really the > full tuple, modify callback to only take ItemPointer.
Looks good to me. Planning to apply this unless somebody wants to argue against it soon?
Andres, I didn't yet register this for next commitfest. If its going in soon anyways will not do it otherwise let me know and I will add it to the list.
