Re: support for MERGE - Mailing list pgsql-hackers

From Zhihong Yu
Subject Re: support for MERGE
Date
Msg-id CALNJ-vQeAvD+O+Xfqk6ed-AeocOj5f4TG07ooLi0=rACyhmP3A@mail.gmail.com
Whole thread Raw
In response to Re: support for MERGE  (Alvaro Herrera <alvherre@alvh.no-ip.org>)
List pgsql-hackers


On Sat, Nov 13, 2021 at 7:41 AM Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:
On 2021-Nov-12, Zhihong Yu wrote:

> +lmerge_matched:
> ...
> +   foreach(l, resultRelInfo->ri_matchedMergeAction)
>
> I suggest expanding the foreach macro into the form of for loop where the
> loop condition has extra boolean variable merge_matched.
> Initial value for merge_matched can be true.
> Inside the loop, we can adjust merge_matched's value to control whether the
> for loop continues.
> This would avoid using goto label.

Heh, have you seen the definition of foreach() lately?  I do *not* want
to expand that anywhere.  Anyway, even if we had the old foreach()
(before commit 1cff1b95ab6d), ISTM that the goto makes the whole thing
much clearer.  For example, where would you do the
table_tuple_fetch_row_version call that currently lies after the goto
label but before the foreach loop starts?

The table_tuple_fetch_row_version() can be called before the next iteration starts (followed by the adjustment of loop control variables).

Since you feel using goto is clearer, it is Okay to keep the current formulation.

Cheers
 

pgsql-hackers by date:

Previous
From: "Euler Taveira"
Date:
Subject: Re: Allow users to choose what happens when recovery target is not reached
Next
From: Tom Lane
Date:
Subject: Re: ALTER TABLE DETACH PARTITION violates serializability