Letting user manually name the multirange (after a few automatic attempts) seems reasonable.
Cheers On Wed, Dec 16, 2020 at 3:34 PM Alexander Korotkov <aekorot...@gmail.com> wrote: > On Thu, Dec 17, 2020 at 1:03 AM Alexander Korotkov <aekorot...@gmail.com> > wrote: > > On Thu, Dec 17, 2020 at 12:54 AM Zhihong Yu <z...@yugabyte.com> wrote: > > > + * The idea is to prepend underscores as needed until we make a > name that > > > + * doesn't collide with anything ... > > > > > > I wonder if other characters (e.g. [a-z0-9]) can be used so that name > without collision can be found without calling truncate_identifier(). > > > > Probably. But multiranges just shares naming logic already existing > > in arrays. If we're going to change this, I think we should change > > this for arrays too. And this change shouldn't be part of multirange > > patch. > > I gave this another thought. Now we have facility to name multirange > types manually. I think we should give up with underscore naming > completely. If both replacing "range" with "mutlirange" in the > typename and appending "_multirange" to the type name failed (very > unlikely), then let user manually name the multirange. Any thoughts? > > ------ > Regards, > Alexander Korotkov >
