Re: Bug in to_timestamp().
amulsul <sul_amul@yahoo.co.in>
From: amul sul <sul_amul@yahoo.co.in>
To: Artur Zakirov <a.zakirov@postgrespro.ru>, Pavel Stehule <pavel.stehule@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>, "David G. Johnston" <david.g.johnston@gmail.com>, Alex Ignatov <a.ignatov@postgrespro.ru>, Bruce Momjian <bruce@momjian.us>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2016-08-15T14:56:34Z
Lists: pgsql-hackers
On Thursday, 11 August 2016 3:18 PM, Artur Zakirov <a.zakirov@postgrespro.ru> wrote:
>Here is my patch. It is a proof of concept.>Date/Time Formatting>-------------------->There are changes in date/time formatting rules:-> now to_timestamp() and to_date() skip spaces in the input string and >in the formatting string unless FX option is used, as Amul Sul wrote on >first message of this thread. But Ex.2 gives an error now with this >patch (should we fix this too?).
Why not, currently we are skipping whitespace exists at the start of input string but not if in format string.
[Skipped… ]
>Of course this patch can be completely wrong. But it tries to introduce >more formal rules for formatting.>I will be grateful for notes and remarks.
Following are few scenarios where we break existing behaviour:
SELECT TO_TIMESTAMP('2015-12-31 13:43:36', 'YYYY MM DD HH24 MI SS');SELECT TO_TIMESTAMP('2011$03!18 23_38_15', 'YYYY-MM-DD HH24:MI:SS');SELECT TO_TIMESTAMP('2011*03*18 23^38&15', 'YYYY-MM-DD HH24:MI:SS');SELECT TO_TIMESTAMP('2011*03!18 #%23^38$15', 'YYYY-MM-DD$$$HH24:MI:SS');
But current patch behaviour is not that much bad either at least we have errors, but I am not sure about community acceptance.
I would like to divert communities' attention on following case:SELECT TO_TIMESTAMP('2013--10-01', 'YYYY-MM-DD');
Where the hyphen (-) is not skipped. So ultimately -10 is interpreted using MM as negative 10. So the date goes back by that many months (and probably additional days because of -31), and so the final output becomes 2012-01-30. But the fix is not specific to hyphen case. Ideally the fix would have been to handle it in from_char_parse_int(). Here, -10 is converted to int using strtol. May be we could have done it using strtoul(). Is there any intention behind not considering signed integers versus unsigned ones ?
Another is, shouldn’t we have error in following cases? SELECT TO_TIMESTAMP('2016-06-13 99:99:99', 'YYYY-MM-DD HH24:MI:SS'); SELECT TO_TIMESTAMP('2016-02-30 15:43:36', 'YYYY-MM-DD HH24:MI:SS');
Thanks & Regards,Amul Sul
Commits
-
Improve behavior of to_timestamp()/to_date() functions
- cf984672427e 12.0 landed
-
Implement TZH and TZM timestamp format patterns
- 11b623dd0a2c 11.0 cited
-
as attache of this mail is patch (to the main tree) with to_char's
- b866d2e2d794 7.1.1 cited