> Bruce Momjian wrote:
> >
> > Vadim, would you please review this code, and let me know if it is
> > correct. I am unsure about the calls to ExecStoreTuple(), ExecQual()
> > (is proper context used?), and placement of ReleaseBuffer(). Do I need
> ^^^^^^^^^^^^^^^^^^^^^^
> Yes.
> execScan.c:ExecScan() sets econtext->ecxt_scantuple = slot <--
> returned by Seq/Index scan -- before check Scan qual and
> so econtext->ecxt_scantuple already points to the slot where
> you store new tuple. It seems that no one change this pointer...
>
> But did you try to use multi-index scan in inner plan of
> Nestloop? It seems that you should add this to ExecIndexReScan():
No. I just don't understand the behavior of the various join types on
index use.
>
> if (exprCtxt != NULL)
> node->scan.scanstate->cstate.cs_ExprContext->ecxt_outertuple =
> exprCtxt->ecxt_outertuple;
Added.
>
> Also, ExecIndexReScan() evaluates run-time keys for _current_
> index scan only - this must be changed and indexstate->iss_IndexPtr
> must be setted to 0.
I have added this to the code.
> > to free the tuple buffer if I decide it doesn't meet my ExecQual test?
>
> Yes, you have to free buffer.
Done.
Vadim, I am going to send you the current nodeIndexscan.c for your
review, to make sure I made the proper changes.
--
Bruce Momjian | 830 Blythe Avenue
maillist@candle.pha.pa.us | Drexel Hill, Pennsylvania 19026
+ If your life is a hard drive, | (610) 353-9879(w)
+ Christ can be your backup. | (610) 853-3000(h)