>
>
> Again, apologies if this is a duplication from the past but I can't it in
> pgsql-questions :
>
> -------------------------------
>
> In the reference section, it states there are problems with dumping views
> and rules.
> A pg_dumpall/pg_dump stores the view as a table with a rule.
> Therefore, when loaded back in, the view is now a table and not loaded into
> pg_view.
>
> To change this, do I create a simple script to remove the 'CREATE TABLE' and
> transform the 'CREATE RULE' into a create view statement.
I don't see the point here. pg_views itself is a view, which
uses a special support function to display the views
definition from the ON SELECT rule that is defined on the
table. There is no difference! A view is a class with an
INSTEAD rule on event SELECT in Postgres. You might not have
noticed, that after creating the rule, the table appears in
pg_views.
In fact, the CREATE TABLE, CREATE RULE is nearly what the
backend internally does when you issue CREATE VIEW. Only that
you don't see it. Again, in Postgres views are implemented as
tables with a rewrite rule.
The reason this is splitted in the dump is, that rules on
other event's than SELECT would affect the reloaded result if
you dump the data as INSERT statements instead of COPY data.
pg_dump just don't make any difference between rule events.
The problems with dumping views and rules is, that pg_dump
uses the mentioned special support function to parse back the
internal (querytree) representation of the rule action into
something the parser accepts to generate the same querytree
again. This backparsing isn't tested very much. So you might
be able to define views/rules, that pg_dump will fail to
output the correct statement that could reinstall the rule.
Up to now, the rules create command isn't stored (I plan
change this in the future). So the try to parse back the
querytree is up to now the only chance pg_dump has. As long
as the views pg_views and pg_rules work and don't produce an
error, pg_dump will output a reloadable dump and you aren't
hit so far by that problem.
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#======================================== [EMAIL PROTECTED] (Jan Wieck) #
