BUG #16386: drop contraint in inherited table is missing in pg_dump backup

On Fri, 24 Apr 2020 at 07:09, PG Bug reporting form <noreply@postgresql.org>
wrote:

create table b (aaa int primary key,bb date );
create table A (id int primary key) inherits (B);
alter table a alter column aaa drop not null;

after i made pg_dump and pg_restore the not null contraint isexist on table
a;

It does not make sense to exclude a not null constraint on an inherited

table because column "aaa" can be null in table "a" but a SELECT in table
"b" will return a NULL for a primary key (ugh). It is just one of the ways
to shoot yourself in the foot. If you check CREATE TABLE synopsis, there
isn't NO INHERIT for not null constraints. Maybe it is worth adding a note
in CREATE TABLE.

--
Euler Taveira http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

On 2020-Apr-24, Euler Taveira wrote:

On Fri, 24 Apr 2020 at 07:09, PG Bug reporting form <noreply@postgresql.org>
wrote:

create table b (aaa int primary key,bb date );
create table A (id int primary key) inherits (B);
alter table a alter column aaa drop not null;

after i made pg_dump and pg_restore the not null contraint isexist on table
a;

It does not make sense to exclude a not null constraint on an inherited
table because column "aaa" can be null in table "a" but a SELECT in table
"b" will return a NULL for a primary key (ugh). It is just one of the ways
to shoot yourself in the foot. If you check CREATE TABLE synopsis, there
isn't NO INHERIT for not null constraints. Maybe it is worth adding a note
in CREATE TABLE.

I agree with your analysis, but in that case we should make the DROP NOT
NULL throw an error instead of proceeding.

--
�lvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Alvaro Herrera <alvherre@2ndquadrant.com> writes:

On 2020-Apr-24, Euler Taveira wrote:

On Fri, 24 Apr 2020 at 07:09, PG Bug reporting form <noreply@postgresql.org>
wrote:

create table b (aaa int primary key,bb date );
create table A (id int primary key) inherits (B);
alter table a alter column aaa drop not null;

It does not make sense to exclude a not null constraint on an inherited
table because column "aaa" can be null in table "a" but a SELECT in table
"b" will return a NULL for a primary key (ugh). It is just one of the ways
to shoot yourself in the foot. If you check CREATE TABLE synopsis, there
isn't NO INHERIT for not null constraints. Maybe it is worth adding a note
in CREATE TABLE.

I agree with your analysis, but in that case we should make the DROP NOT
NULL throw an error instead of proceeding.

Indeed, but we lack the catalog infrastructure to do that reasonably.
If we ever get around to creating pg_constraint entries for NOT NULL
constraints, it'd be simple to do, since those would carry info about
whether or not the constraint is inherited. (Weren't you working on
that awhile ago?)

regards, tom lane