Bruce Momjian wrote:
>> On Wed, 24 Jan 2001, Bruce Momjian wrote:
>I smell TODO item. In fact, I now see a TODO item:
>* Unique index on base column not honored on inserts from inherited table
> INSERT INTO inherit_table (unique_index_col) VALUES (dup) should fail
>So it seems the fact the UNIQUE doesn't apply to the new table is just a
>manifestion of the fact that people expect UNIQUE to span the entire
>inheritance tree. I will add the emails to [inherit] and mark it as
Bruce, could you add this text to TODO.detail on the subject of
inherited constraints. I first sent it on Christmas Eve, and I
think most people were too busy holidaying to comment.
Tom Lane wrote:
>Hm. The short-term answer seems to be to modify the queries generated
>by the RI triggers to say "ONLY foo". I am not sure whether we
>understand the semantics involved in allowing a REFERENCES target to be
>taken as an inheritance tree rather than just one table, but certainly
>the current implementation won't handle that correctly.
May I propose these semantics as a basis for future development:
1. An inheritance hierarchy (starting at any point in a tree) should be
equivalent to an updatable view of all the tables at the point of
reference and below. By default, all descendant tables are combined
with the ancestor for all purposes. The keyword ONLY must be used to
alter this behaviour. Only inherited columns of descendant tables are
visible from higher in the tree. Columns may not be dropped in descendants.
If columns are added to ancestors, they must be inserted correctly in
descendants so as to preserve column ordering and inheritance. If
a column is dropped in an ancestor, it is dropped in all descendants.
2. Insertion into a hierarchy means insertion into the table named in
the INSERT statement; updating or deletion affects whichever table(s)
the affected rows are found in. Updating cannot move a row from one
table to another.
3. Inheritance of a table implies inheriting all its constraints unless
ONLY is used or the constraints are subsequently dropped; again, dropping
operates through all descendant tables. A primary key, foreign key or
unique constraint cannot be dropped or modified for a descendant. A
unique index on a column is shared by all tables below the table for
which it is declared. It cannot be dropped for any descendant.
In other words, only NOT NULL and CHECK constraints can be dropped in
In multiple inheritance, a column may inherit multiple unique indices
from its several ancestors. All inherited constraints must be satisfied
together (though check constraints may be dropped).
4. RI to a table implies the inclusion of all its descendants in the
check. Since a referenced column may be uniquely indexed further up
the hierarchy than in the table named, the check must ensure that
the referenced value occurs in the right segment of the hierarchy. RI
to one particular level of the hierarchy, excluding descendants, requires
the use of ONLY in the constraint.
5. Dropping a table implies dropping all its descendants.
6. Changes of permissions on a table propagate to all its descendants.
Permissions on descendants may be looser than those on ancestors; they
may not be more restrictive.
This scheme is a lot more restrictive than C++'s or Eiffel's definition
of inheritance, but it seems to me to make the concept truly useful,
without introducing excessive complexity.
Oliver Elphick Oliver(dot)Elphick(at)lfix(dot)co(dot)uk
Isle of Wight http://www.lfix.co.uk/oliver
PGP: 1024R/32B8FAA1: 97 EA 1D 47 72 3F 28 47 6B 7E 39 CC 56 E4 C1 47
GPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C
"If anyone has material possessions and sees his
brother in need but has no pity on him, how can the
love of God be in him?"
I John 3:17
pgsql-hackers by date
|Next:||From: Peter Eisentraut||Date: 2001-01-24 22:02:09|
|Subject: Permissions on CHECKPOINT|
|Previous:||From: Peter Eisentraut||Date: 2001-01-24 21:36:37|
|Subject: Re: LEFT JOIN in pg_dumpall is a bug |