Re: tablecmds.c/MergeAttributes() cleanup

On 2023-Jun-28, Peter Eisentraut wrote:

> The MergeAttributes() and related code in and around tablecmds.c is huge and
> ancient, with many things bolted on over time, and difficult to deal with.
> I took some time to make careful incremental updates and refactorings to
> make the code easier to follow, more compact, and more modern in appearance.
> I also found several pieces of obsolete code along the way. This resulted
> in the attached long patch series. Each patch tries to make a single change
> and can be considered incrementally. At the end, the code is shorter,
> better factored, and I hope easier to understand. There shouldn't be any
> change in behavior.

I request to leave this alone for now. I have enough things to juggle
with in the NOTNULLs patch; this patchset looks like it will cause me
messy merge conflicts. 0004 for instance looks problematic, as does
0007 and 0016.

FWIW for the most part that patch is working and I intend to re-submit
shortly, but the relevant pg_upgrade code is really brittle, so it's
taken me much more than I expected to get it in good shape for all
cases.

Thanks

--
Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/