Обсуждение: Bug #620: create index ... (date_part('year', d)) fails

Ernst Molitor (molitor@uni-bonn.de) reports a bug with a severity of 3
The lower the number the more severe it is.

Short Description
create index ... (date_part('year', d)) fails

Long Description
IMHO, creation of an index on a part of a date field should work, 
but it fails due to the definition of "create index ..." in gram.y. 

Maybe I'm wrong, but I feel that for large tables which will have to be analyzed per year, an index of this kind could
bean advantage. 
 


Sample Code
-- the table
create table test (d date);

-- partial indices:
create index d_year on test  (date_part('year', d));

--  fails with: ERROR:  parser: parse error at or near "'"

create index d_year on test (extract(year from d));

--  fails with: ERROR:  parser: parse error at or near "("

No file was uploaded with this report

sszabo@bigpanda.com

On Fri, 15 Mar 2002 pgsql-bugs@postgresql.org wrote:

> Ernst Molitor (molitor@uni-bonn.de) reports a bug with a severity of 3
> The lower the number the more severe it is.
>
> Short Description
> create index ... (date_part('year', d)) fails
>
> Long Description
> IMHO, creation of an index on a part of a date field should work,
> but it fails due to the definition of "create index ..." in gram.y.
>
> Maybe I'm wrong, but I feel that for large tables which will have to
> be analyzed per year, an index of this kind could be an advantage.

You need to make a function that is marked iscachable that returns
date_part('year', <argument>) and use that.  IIRC, functional indices
require that all of the arguments be column references.