On Tue, Oct 21, 2025 at 3:56 PM Chao Li <li.evan.chao@gmail.com> wrote:
> > On Oct 21, 2025, at 15:30, jian he <jian.universality@gmail.com> wrote:
> >
> > +-- So, you can change a column's type as long as any dependent generated
> > +-- column already has a set expression defined:
> > +ALTER TABLE gtest27
> > + ALTER COLUMN a TYPE float8,
> > + ALTER COLUMN x SET EXPRESSION AS ((a::int + b) * 2);
> > +\d gtest27
> > + Table "generated_stored_tests.gtest27"
> > + Column | Type | Collation | Nullable |
> > Default
> > +--------+------------------+-----------+----------+---------------------------------------------------
> > + a | double precision | | |
> > + b | bigint | | |
> > + x | bigint | | | generated always
> > as ((a::integer + b) * 2) stored
> > +
> >
> > the above output seems wrong?
> > one way to quickly test it is create table gtest27 again
>
> I don’t get what is wrong? Actually “make check” passed from my side:
>
Sorry for the noise, I misunderstood something.
I think I understand what your patch does now.
I’ll polish my patch and submit it in a new thread.