Question

I am trying to create one view (or view in each schema to be used without modifications) on table which exists in multiple schemas with the same name

create schema company_1;
create schema company_2;
...
CREATE TABLE company_1.orders
(
  id serial NOT NULL,
  amount real,
  paid real,
  CONSTRAINT orders_pkey PRIMARY KEY (id )
)
WITH (
  OIDS=FALSE
);

CREATE TABLE company_2.orders
(
  id serial NOT NULL,
  amount real,
  paid real,
  CONSTRAINT orders_pkey PRIMARY KEY (id )
)
WITH (
  OIDS=FALSE
);

....

What is correct way of creating view on table orders without specifying schema for every view or specifying current schema?

What I need and failed to get is either

CREATE OR REPLACE VIEW
public.full_orders AS
SELECT id, amount FROM orders;

or

CREATE OR REPLACE VIEW
company_1.full_orders AS
-- company_2.full_orders AS
-- company_n.full_orders AS
SELECT id, amount FROM current_schema.orders;

Using postgresql 9.2.2

EDIT: The way I went:

CREATE VIEW company_1.full_orders AS
SELECT id, amount FROM company_1.orders;

On schema copy discussed here I butaly do this

FOR src_table IN 
   SELECT table_name 
   FROM information_schema.TABLES 
   WHERE table_schema = source_schema AND table_type = 'VIEW'
LOOP
    SELECT view_definition 
    FROM information_schema.views 
    WHERE table_name = src_table AND table_schema = source_schema INTO q;
    trg_table := target_schema||'.'||src_table;
    EXECUTE 'CREATE VIEW ' || trg_table || ' AS '||replace(q, source_schema, target_schema);
END LOOP;

Still looking for better solution…

Answer 1

It’s not possible to do with with a straightforward view. The view records the underlying table’s identity at creation time, so it is not affected by schema settings done later on.

You could do it using a set-returning function using dynamic SQL, and then wrap that into a view. But I don’t think that’s a good solution.

I would just create quasi-duplicates for the view, as you have been doing, and enhance my deployment script to keep them all up to date.