On 2021-02-25 19:55, Jiří Činčura wrote:
Could you describe the scenario you're thinking of that you think
could
be problematic.
Well, I was mostly thinking what would happen is suddenly some dates
become colliding in UTC in i.e. unique index and you rebuild the
index. It was fine before, but it's not now.
If you rebuild the index, there will be no such problem, because the
date is stored in UTC, so that will not have changed.
But it could be also something like:
insert <some date> 15:00 in foo/bar (+3)
TZ database change, where "foo/bar" becomes +2
insert <some date> 15:00 in foo/bar (+2)
Same local time, but different UTC, hence no unique violation. Is that
the expected experience? Or when "insert <some date> 14:00 in foo/bar
(+2)" and the unique constraint is gonna complain, although it's
different local time. Again, is that the expected experience?
Yes, that is expected, because the first datetime would now be 14:00
foo/bar (this is one of the big caveats of using named zones instead of
offsets), not 15:00 foo/bar. So the second insert you mention would not
conflict, and the third insert you mention would.
Mark
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
