On 23.06.2016 20:40, Tom Lane wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Thu, Jun 23, 2016 at 1:12 PM, David G. Johnston
>> <david.g.johnston@gmail.com> wrote:
>>> My understanding is that is not going to change for 9.6.
>> That's exactly what is under discussion here.
> I would definitely agree with David on that point. Making to_timestamp
> noticeably better on this score seems like a nontrivial project, and
> post-beta is not the time for that sort of thing, even if we had full
> consensus on what to do. I'd suggest somebody work on a patch and put
> it up for review in the next cycle.
>
> Now, if you were to narrowly define the problem as "whether to skip
> non-spaces for a space in the format", maybe that could be fixed
> post-beta, but I think that's a wrongheaded approach. to_timestamp's
> issues with input that doesn't match the format are far wider than that.
> IMO we should try to resolve the whole problem with one coherent change,
> not make incremental incompatible changes at the margins.
>
> At the very least I'd want to see a thought-through proposal that
> addresses all three of these interrelated points:
>
> * what should a space in the format match
> * what should a non-space, non-format-code character in the format match
> * how should we handle fields that are not exactly the width suggested
> by the format
>
> regards, tom lane
>
>
Totally agree that we need more discussion about error handling in this
function!
Also this behavior is observed in to_date() and to_number() function:
postgres=# SELECT
TO_DATE('2!0!1!6----!0!6-/-/-/-/-/-/-1!/-/-/-/-/-/-/-3!', 'YYYY-MM-DD'); to_date
------------ 0002-01-01
(1 row)
postgres=# postgres=# select to_number('1$#@!!,2,%,%4,5,@%5@4..8-',
'999G999D9S'); to_number
----------- 12
(1 row)
On the our side we have some discussions about to write a patch that
will change this incorrect behavior. So stay tuned.
--
Alex Ignatov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company