I would like PostegreSQL to relax a bit. Every time I want to change a column used in a view, it seems I have to drop the view, change the field and then recreate the view. Can I waive the extra protection and just tell PostgreSQL to let me change the field and then figure out the adjustment to the view?
Clarification: I understand what a view is. In fact, it's because the view is like a subquery that I wish I could just change the underlying tables and have the view pick up the change.
Let's say I have the following:
CREATE TABLE monkey
(
"name" character varying(50) NOT NULL,
)
CREATE OR REPLACE VIEW monkey_names AS
SELECT name
FROM monkey
I really just want to do the following in a migration script without having to drop and recreate the view.
ALTER TABLE monkey ALTER COLUMN "name" character varying(100) NOT NULL
Before changing the table and column definitions, call the function to save and drop the dependencies, make the definition changes and finally call the function to restore the dependencies. select util. deps_save_and_drop_dependencies('mdm', 'global_item_master_swap'); alter table mdm.
The altered view replaces the existing view, so you cannot modify specific columns in a view. A view is a virtual table based on the result set of a SELECT query or a UNION of such queries. For more details on views, see Defining and Using Views.
ALTER VIEW [ IF EXISTS ] name ALTER [ COLUMN ] column_name SET DEFAULT expression ALTER VIEW [ IF EXISTS ] name ALTER [ COLUMN ] column_name DROP DEFAULT ALTER VIEW [ IF EXISTS ] name OWNER TO { new_owner | CURRENT_ROLE | CURRENT_USER | SESSION_USER } ALTER VIEW [ IF EXISTS ] name RENAME [ COLUMN ] column_name TO ...
First, specify the name of the table to which the column you want to change belongs in the ALTER TABLE clause. Second, give the name of column whose data type will be changed in the ALTER COLUMN clause. Third, provide the new data type for the column after the TYPE keyword.
To avoid the problem altogether use the data type text
or varchar
/ character varying
without a length specifier instead of character varying(n)
. Read about these data types in the manual.
CREATE TABLE monkey(name text NOT NULL)
If you really want to enforce a maximum length, create a CHECK
constraint:
ALTER TABLE monkey
ADD CONSTRAINT monkey_name_len CHECK (length(name) < 101);
You can change or drop that constraint any time without touching depending objects like views and without forcing Postgres to write new rows in the table due to the change of type (which isn't always necessary any more in modern version of Postgres).
As proposed by @Michael, I add some more general information:
A view in PostgreSQL is not just an "alias to subquery". Views are implemented as special tables with a rule ON SELECT TO my_view DO INSTEAD
. (That's why you can alter views with an ALTER TABLE
command.) You can GRANT
privileges to it, add comments or even define column defaults (useful for a rule ON INSERT TO my_view DO INSTEAD...
). Read more in the manual here or here.
If you change underlying objects, you need to change the defining query of any depending view, too. The ALTER VIEW
statement can only change auxiliary attributes of a view. Use CREATE OR REPLACE VIEW
to change the query - it will preserve any additional attributes.
However, if you want to change data types of resulting columns (like in the case at hand), CREATE OR REPLACE VIEW
is not possible. You have to DROP
the old and CREATE
a new view. This will never delete any data of the underlying tables. It will drop any additional attributes of the view, though, which have to be recreated, too.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With