On Wed, Dec 2, 2009 at 2:59 PM, Caleb Welton <cwelton@greenplum.com> wrote:
> New patch attached:
>
> 1. Does not add a new error message (though the pg_atoi's error message is a
> little goofy looking).
> 2. Handles int2 overflow cases.
> 3. oidvectorin does NOT suffer from the same problems as int2vectorin,
> someone already fixed it.
>
> As for the use-case I'm not completely sure... I'm not an end-user, I'm just
> responding to a bug report.
>
> My stance here is that returning an error (even a bad error) on trying to
> convert data in is better
> doing something "wrong" with bogus input. In the first case a user
> scratches their head, maybe
> files a bug report, you tell them the correct syntax and they go on. In the
> second case they input
> a bunch of data and then start complaining about "data corruption", "loss of
> data", etc. and the
> support case is 100x worse.
>
> The amount of code we are talking about here is less than 5 lines of code...
I have scrutinized the latest version of this patch and I feel that it
is a modest improvement on the status quo and that there is really no
downside. Absent strong objections, I will commit it later this week.
...Robert