Re: simplifying foreign key/RI checks

On Tue, Jan 19, 2021 at 12:00 PM Zhihong Yu <zyu(at)yugabyte(dot)com> wrote:
> + if (mapped_partkey_attnums[i] == pk_attnums[j])
> + {
> + partkey_vals[i] = pk_vals[j];
> + partkey_isnull[i] = pk_nulls[j] == 'n' ? true : false;
> + i++;
> + break;
>
> The way counter (i) is incremented is out of my expectation.
> In the rare case, where some i doesn't have corresponding pk_attnums[j], wouldn't there be a dead loop ?
>
> I think the goal of adding the assertion should be not loop infinitely even if the invariant is not satisfied.
>
> I guess a counter other than i would be better for this purpose.

I have done that in v3. Thanks.

--
Amit Langote
EDB: http://www.enterprisedb.com