Justin Clift wrote:
> Bruce Momjian wrote:
> > The issue here is whether dropping a column should automatically drop a
> > multi-column index of which that column is a member.
> >
> > The example shown below is particularly good because the dropped field
> > is second in the index, meaning that the index is useful for lookups on
> > field1 alone, so dropping field2 removes a useful index on field1. I
> > don't think it is defensible to allow DROP COLUMN to remove the index.
> > Instead, I think we have to refuse the DROP COLUMN and require the user
> > to drop the index and recreate it just on field1 if desired. I don't
> > think CASCASE enters into this because of the effect on field1.
> >
> > Comments?
>
> Would it be possible/practical to have PostgreSQL recreate the
> multi-column index, but without the dropped column?
Yes, we could, but creating an index takes time, so is probably
something we would want the admin to do manually.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073
