On 2016/08/19 5:35, Robert Haas wrote:
> On Thu, Aug 18, 2016 at 5:15 AM, Amit Langote
> <[email protected]> wrote:
>> On 2016/07/25 17:18, Amit Langote wrote:
>>> The comment seems to have been copied from ATExecAddColumn, which says:
>>>
>>> /*
>>> * If we are told not to recurse, there had better not be any
>>> - * child tables; else the addition would put them out of step.
>>>
>>> For ATExecValidateConstraint, it should say something like:
>>>
>>> + * child tables; else validating the constraint would put them
>>> + * out of step.
>>>
>>> Attached patch fixes it.
>>
>> I noticed that the ALTER TABLE documentation doesn't mention that VALIDATE
>> CONSTRAINT will fail if ONLY is specified and there are descendant tables.
>> It only mentions that a constraint cannot be renamed unless also renamed
>> in the descendant tables.
>>
>> Attached patch fixes that.
>
> I did some wordsmithing on the two patches you posted to this thread.
> I suggest the attached version. What do you think?
Reads much less ambiguous, so +1.
Except in the doc patch:
s/change the type of a column constraint/change the type of a column/g
I fixed that part in the attached.
Thanks,
Amit
diff --git a/doc/src/sgml/ref/alter_table.sgml b/doc/src/sgml/ref/alter_table.sgml
index 6f51cbc..e48ccf2 100644
--- a/doc/src/sgml/ref/alter_table.sgml
+++ b/doc/src/sgml/ref/alter_table.sgml
@@ -1028,11 +1028,15 @@ ALTER TABLE ALL IN TABLESPACE <replaceable class="PARAMETER">name</replaceable>
<para>
If a table has any descendant tables, it is not permitted to add,
- rename, or change the type of a column, or rename an inherited constraint
- in the parent table without doing
- the same to the descendants. That is, <command>ALTER TABLE ONLY</command>
- will be rejected. This ensures that the descendants always have
- columns matching the parent.
+ rename, or change the type of a column in the parent table without doing
+ same to the descendants. This ensures that the descendants always have
+ columns matching the parent. Similarly, a constraint cannot be renamed
+ in the parent without also renaming it in all descendents, so that
+ constraints also match between the parent and its descendents.
+ Also, because selecting from the parent also selects from its descendents,
+ a constraint on the parent cannot be marked valid unless it is also marked
+ valid for those descendents. In all of these cases, <command>ALTER TABLE
+ ONLY</command> will be rejected.
</para>
<para>
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 86e9814..d312762 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -6908,7 +6908,8 @@ ATExecValidateConstraint(Relation rel, char *constrName, bool recurse,
/*
* If we are told not to recurse, there had better not be any
- * child tables; else the addition would put them out of step.
+ * child tables, because we can't mark the constraint on the
+ * parent valid unless it is valid for all child tables.
*/
if (!recurse)
ereport(ERROR,
--
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
