Re: ALTER INDEX .. SET STATISTICS ... behaviour

On Wed, May 31, 2017 at 7:18 PM, Alexander Korotkov <
a(dot)korotkov(at)postgrespro(dot)ru> wrote:

> On Wed, May 31, 2017 at 6:53 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
>> Alexander Korotkov <a(dot)korotkov(at)postgrespro(dot)ru> writes:
>> > I've discovered that PostgreSQL is able to run following kind of
>> queries in
>> > order to change statistic-gathering target for an indexed expression.
>>
>> > ALTER INDEX index_name ALTER COLUMN expr SET STATISTICS stat_target;
>>
>> > It's been previously discussed in [1].
>>
>> > I think this should be fixed not just in docs. This is why I've started
>> > thread in pgsql-hackers. For me usage of internal column names "expr",
>> > "expr1", "expr2" etc. looks weird. And I think we should replace it
>> with a
>> > better syntax. What do you think about these options?
>>
>> > ALTER INDEX index_name ALTER EXPRESSION 0 SET STATISTICS stat_target; --
>> > Refer expression by its number
>> > ALTER INDEX index_name ALTER EXPRESSION (x + y) SET STATISTICS
>> stat_target;
>> > -- Refer expression by its definition
>>
>> Don't like either of those particularly, but what about just targeting
>> a column by column number, independently of whether it's an expression?
>>
>> ALTER INDEX myindex ALTER COLUMN 4 SET STATISTICS 1000;
>>
>
> I completely agree with that.
> If no objections, I will write a patch for that.
>

Please, find it in attached patch.

I decided to forbid referencing columns by numbers for tables, because
those numbers could contain gaps. Also, I forbid altering statistics
target for non-expression index columns, because it has no effect.

------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company