On 2 January 2012 18:07, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Jean-Yves F. Barbier" <12ukwn@gmail.com> writes:
>> I'm trying to insert as:
>> INSERT INTO tst1m VALUES ( default,
>> 'źŠÿǚŦÍÏÔJŽććżûŘäpèăîÍŢŃDµŴŻĊĜÿG' COLLATE "hu_HU.utf8",
>> 'úĽčŏṻŗĿÁťħĵŇğŚáÛĨæħŽīhĴÝUĶďeźİīŇč' COLLATE "hu_HU.utf8" );
>
>> and get: ERROR: unsupported node type: 325
>
> Hmm, that's a bug, which I will go fix --- thanks for the report!
> However, the reason that nobody noticed the bug before now is that it's
> entirely useless to attach a COLLATE clause to an INSERT value.
> COLLATE is only interesting when attached to an argument of an operator
> or function that pays attention to collation, such as the "<" operator.
> Here, your values are going to go directly into the table without any
> comparisons or anything else happening to them, and once they're in
> the table it'll be the table column's declared collation that controls
> how they're compared.
For reference, Tom committed the fix for this in both master and 9.1
branch: http://archives.postgresql.org/pgsql-committers/2012-01/msg00019.php
--
Thom