Re: possible/feasible to specify field and value in error msg?

Hello

2013/7/4 Willy-Bas Loos <willybas(at)gmail(dot)com>:
> On Wed, Jul 3, 2013 at 5:18 PM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
>>
>> On Wed, Jul 3, 2013 at 11:14:18AM -0400, Bruce Momjian wrote:
>> > We will add optional error details in Postgres 9.3:
>> >
>> > http://momjian.us/main/blogs/pgblog/2013.html#April_11_2013
>
>
>>
>> I just tested this and it doesn't show the offending column name;
>> sorry:
>>
>> test=> CREATE TABLE test(x smallint);
>> CREATE TABLE
>> test=> \set VERBOSITY verbose
>> test=> INSERT INTO test VALUES (10000000);
>> ERROR: 22003: smallint out of range
>> LOCATION: i4toi2, int.c:349
>>
>
>
> It's great to see that you people care about "userland", judging by the
> effort that you describe in your article.
> In fact you're already doing the thing that i asked about, i see that even
> the offending tuple is printed (which is new).
> And of course it's not necessary to mention the column name when you mention
> the constraint name.
> (BTW: your remark about NOT NULL constraints is not necessary, that error
> message is very clear:"ERROR: null value in column "balance" violates
> not-null constraint" )
>
> This is not a constraint going off, and in this case, none of that applies.
> But it seems probable to me that some day it will, seeing as you already
> implemented it for constraints.

this functionality will be enhanced in future - but it hardly depends
on current constraint and checks implementation - for some kind of
errors we are not able to join a exception with related column -
typically it is domain errors, probably we can to fill DATATYPE field
in this case.

Regards

Pavel Stehule

>
> Thanks,
>
> Willy-Bas Loos
>
> --
> "Quality comes from focus and clarity of purpose" -- Mark Shuttleworth