|From:||Amit Langote <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp>|
|To:||Pg Hackers <pgsql-hackers(at)postgresql(dot)org>|
|Cc:||Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp>|
|Subject:||using expression syntax for partition bounds (was: Re: Boolean partitions syntax)|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
(patch and discussion for PG 12)
On 2018/04/22 1:28, Tom Lane wrote:
> Amit Langote <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> writes:
>> [ v8-0001-Allow-generalized-expression-syntax-for-partition.patch ]
> I find what you did to a_expr here to be pretty horrid.
Thanks for the review.
> I think what you should do is lose the partbound_datum and
> PartitionRangeDatum productions altogether, replacing those with just
> a_expr, as in the attached grammar-only patch. This would result in
> needing to identify MINVALUE and MAXVALUE during parse analysis, since
> the grammar would just treat them as ColId expressions. But since we're
> not intending to ever allow any actual column references in partition
> expressions, I don't see any harm in allowing parse analysis to treat
> ColumnRefs containing those names as meaning the special items.
I have to agree this is better.
> This is a little bit grotty, in that both MINVALUE and "minvalue" would
> be recognized as the keyword, but it's sure a lot less messy than what's
> there now. And IIRC there are some other places where we're a bit
> squishy about the difference between identifiers and keywords, too.
I tried to update the patch to do things that way. I'm going to create a
new entry in the next CF titled "generalized expression syntax for
partition bounds" and add the patch there.
|Next Message||Amit Langote||2018-04-23 02:44:14||Re: using expression syntax for partition bounds (was: Re: Boolean partitions syntax)|
|Previous Message||Andrew Gierth||2018-04-23 02:09:22||Re: Toast issues with OldestXmin going backwards|