Problem with Postgres ALTER TABLE

Question

I have one problem with the ALTER TABLE in postgre. I want to change size of the varchar column. When I try to do this, It says that the view is dependent on that column. I can't drop the view because comething else is dependent on it. Is there any other way than to drop everything and recreate it again?

I just found one option, which is to remove the table joining from the view, when I will not change the returned columns, I can do that. But still, there is more views I'll need to change. Isn't there anything how can I say that it should be deferred and checked with commit?

Answer 1

I have run into this problem and couldn't find any way around it. Unfortunately, as best I can tell, one must drop the views, alter the column type on the underlying table, and then recreate the views. This can happen entirely in a single transaction.

Constraint deferral doesn't apply to this problem. In other words, even SET CONSTRAINTS ALL DEFERRED has no impact on this limitation. To be specific, constraint deferral does not apply to the consistency check that prints ERROR: cannot alter type of a column used by a view or rule when one tries to alter the type of a column underlying a view.

Answer 2

I'm a little late to the party, but years after this question was posted, a brilliant solution was posted via an article referenced below (not mine -- I'm simply a thankful beneficiary of his brilliance).

I just tested this on an object that is referenced (on the first level) in 136 separate views, and each of those views is referenced in other views. The solution ran in mere seconds.

So, read this article and copy and paste the table and two functions listed:

http://mwenus.blogspot.com/2014/04/postgresql-how-to-handle-table-and-view.html

Implementation example:

alter table mdm.global_item_master_swap
alter column prod_id type varchar(128),
alter column prod_nme type varchar(512);

ERROR: cannot alter type of a column used by a view or rule DETAIL: rule _RETURN on view toolbox_reporting."Average_setcost" depends on column "prod_id" ********** Error **********

ERROR: cannot alter type of a column used by a view or rule

And now for the PostgreSQL ninja's magic:

select util.deps_save_and_drop_dependencies('mdm', 'global_item_master_swap');


alter table mdm.global_item_master_swap
alter column prod_id type varchar(128),
alter column prod_nme type varchar(512);


select util.deps_restore_dependencies('mdm', 'global_item_master_swap');

-- EDIT 11/13/2018 --

It appears the link above might be dead. Here is the code for the two procedures:

Table that stores DDL:

CREATE TABLE util.deps_saved_ddl
(
  deps_id serial NOT NULL,
  deps_view_schema character varying(255),
  deps_view_name character varying(255),
  deps_ddl_to_run text,
  CONSTRAINT deps_saved_ddl_pkey PRIMARY KEY (deps_id)
);

Save and Drop:

-- Edit 8/28/2020 -- -- This stopped working with Pg12. The fix is below to change the parameters of p_view_schema and p_view_name from varchar to name:

CREATE OR REPLACE FUNCTION util.deps_save_and_drop_dependencies(
    p_view_schema name, p_view_name name)
    RETURNS void
    LANGUAGE plpgsql
    COST 100
AS $BODY$

declare
  v_curr record;
begin
for v_curr in 
(
  select obj_schema, obj_name, obj_type from
  (
  with recursive recursive_deps(obj_schema, obj_name, obj_type, depth) as 
  (
    select p_view_schema, p_view_name, null::varchar, 0
    union
    select dep_schema::varchar, dep_name::varchar, dep_type::varchar, recursive_deps.depth + 1 from 
    (
      select ref_nsp.nspname ref_schema, ref_cl.relname ref_name, 
      rwr_cl.relkind dep_type,
      rwr_nsp.nspname dep_schema,
      rwr_cl.relname dep_name
      from pg_depend dep
      join pg_class ref_cl on dep.refobjid = ref_cl.oid
      join pg_namespace ref_nsp on ref_cl.relnamespace = ref_nsp.oid
      join pg_rewrite rwr on dep.objid = rwr.oid
      join pg_class rwr_cl on rwr.ev_class = rwr_cl.oid
      join pg_namespace rwr_nsp on rwr_cl.relnamespace = rwr_nsp.oid
      where dep.deptype = 'n'
      and dep.classid = 'pg_rewrite'::regclass
    ) deps
    join recursive_deps on deps.ref_schema = recursive_deps.obj_schema and deps.ref_name = recursive_deps.obj_name
    where (deps.ref_schema != deps.dep_schema or deps.ref_name != deps.dep_name)
  )
  select obj_schema, obj_name, obj_type, depth
  from recursive_deps 
  where depth > 0
  ) t
  group by obj_schema, obj_name, obj_type
  order by max(depth) desc
) loop

  insert into util.deps_saved_ddl(deps_view_schema, deps_view_name, deps_ddl_to_run)
  select p_view_schema, p_view_name, 'COMMENT ON ' ||
  case
  when c.relkind = 'v' then 'VIEW'
  when c.relkind = 'm' then 'MATERIALIZED VIEW'
  else ''
  end
  || ' ' || n.nspname || '.' || c.relname || ' IS ''' || replace(d.description, '''', '''''') || ''';'
  from pg_class c
  join pg_namespace n on n.oid = c.relnamespace
  join pg_description d on d.objoid = c.oid and d.objsubid = 0
  where n.nspname = v_curr.obj_schema and c.relname = v_curr.obj_name and d.description is not null;

  insert into util.deps_saved_ddl(deps_view_schema, deps_view_name, deps_ddl_to_run)
  select p_view_schema, p_view_name, 'COMMENT ON COLUMN ' || n.nspname || '.' || c.relname || '.' || a.attname || ' IS ''' || replace(d.description, '''', '''''') || ''';'
  from pg_class c
  join pg_attribute a on c.oid = a.attrelid
  join pg_namespace n on n.oid = c.relnamespace
  join pg_description d on d.objoid = c.oid and d.objsubid = a.attnum
  where n.nspname = v_curr.obj_schema and c.relname = v_curr.obj_name and d.description is not null;
  
  insert into util.deps_saved_ddl(deps_view_schema, deps_view_name, deps_ddl_to_run)
  select p_view_schema, p_view_name, 'GRANT ' || privilege_type || ' ON ' || table_schema || '.' || table_name || ' TO ' || grantee
  from information_schema.role_table_grants
  where table_schema = v_curr.obj_schema and table_name = v_curr.obj_name;
  
  if v_curr.obj_type = 'v' then
    insert into util.deps_saved_ddl(deps_view_schema, deps_view_name, deps_ddl_to_run)
    select p_view_schema, p_view_name, 'CREATE VIEW ' || v_curr.obj_schema || '.' || v_curr.obj_name || ' AS ' || view_definition
    from information_schema.views
    where table_schema = v_curr.obj_schema and table_name = v_curr.obj_name;
  elsif v_curr.obj_type = 'm' then
    insert into util.deps_saved_ddl(deps_view_schema, deps_view_name, deps_ddl_to_run)
    select p_view_schema, p_view_name, 'CREATE MATERIALIZED VIEW ' || v_curr.obj_schema || '.' || v_curr.obj_name || ' AS ' || definition
    from pg_matviews
    where schemaname = v_curr.obj_schema and matviewname = v_curr.obj_name;
  end if;
  
  execute 'DROP ' ||
  case 
    when v_curr.obj_type = 'v' then 'VIEW'
    when v_curr.obj_type = 'm' then 'MATERIALIZED VIEW'
  end
  || ' ' || v_curr.obj_schema || '.' || v_curr.obj_name;
  
end loop;
end;
$BODY$

Restore:

CREATE OR REPLACE FUNCTION util.deps_restore_dependencies(
    p_view_schema character varying,
    p_view_name character varying)
  RETURNS void AS
$BODY$
declare
  v_curr record;
begin
for v_curr in 
(
  select deps_ddl_to_run 
  from util.deps_saved_ddl
  where deps_view_schema = p_view_schema and deps_view_name = p_view_name
  order by deps_id desc
) loop
  execute v_curr.deps_ddl_to_run;
end loop;
delete from util.deps_saved_ddl
where deps_view_schema = p_view_schema and deps_view_name = p_view_name;
end;
$BODY$
  LANGUAGE plpgsql VOLATILE
  COST 100;

Answer 3

If you don't need to change the type of the field, but just the size of it, this approach should work:

Starting with these tables:

CREATE TABLE foo (id integer primary key, names varchar(10));
CREATE VIEW voo AS (SELECT id, names FROM foo);

\\d foo and \\d voo both show the length as 10:

id     | integer               | not null
names  | character varying(10) | 

Now change the lengths to 20 in the pg_attribute table:

UPDATE pg_attribute SET atttypmod = 20+4
WHERE attrelid IN ('foo'::regclass, 'voo'::regclass)
AND attname = 'names';

(note: the 20+4 is some crazy postgresql legacy thing, the +4 is compulsory.)

Now \\d foo shows:

id     | integer               | not null
names  | character varying(20) | 

Bonus: that was waaay faster than doing:

ALTER TABLE foo ALTER COLUMN names TYPE varchar(20);

Technically you can change the size of the table column without changing the size of the view column, but no guarantees on what side effects that will have; it's probably best to change them both at once.

source and fuller explanation: http://sniptools.com/databases/resize-a-column-in-a-postgresql-table-without-changing-data

Answer 4

I ran into this problem today and found a work around to avoid dropping and recreating the VIEW . I cannot just drop my VIEW because it is a master VIEW that has many dependent VIEWs built on top of it. Short of having a rebuild script to DROP CASCADE and then recreate ALL of my VIEWs this is a work around.

I change my master VIEW to use a dummy value for the offending column, altered the column in the table, and switched my VIEW back to the column. Using a setup like this:

CREATE TABLE base_table
(
  base_table_id integer,
  base_table_field1 numeric(10,4)
);

CREATE OR REPLACE VIEW master_view AS 
  SELECT
    base_table_id AS id,
    (base_table_field1 * .01)::numeric AS field1
  FROM base_table;

CREATE OR REPLACE VIEW dependent_view AS 
  SELECT
    id AS dependent_id,
    field1 AS dependent_field1
  FROM master_view;

Trying to alter base_table_field1 type like this:

ALTER TABLE base_table ALTER COLUMN base_table_field1 TYPE numeric(10,6);

Will give you this error:

ERROR:  cannot alter type of a column used by a view or rule
DETAIL:  rule _RETURN on view master_view depends on column "base_table_field1"

If you change master_view to use a dummy value for the column like this:

CREATE OR REPLACE VIEW master_view AS 
  SELECT
    base_table_id AS id,
    0.9999 AS field1
  FROM base_table;

Then run your alter:

ALTER TABLE base_table ALTER COLUMN base_table_field1 TYPE numeric(10,6);

And switch your view back:

CREATE OR REPLACE VIEW master_view AS 
  SELECT
    base_table_id AS id,
    (base_table_field1 * .01)::numeric AS field1
  FROM base_table;

It all depends on if your master_view has an explicit type that does not change. Since my VIEW uses '(base_table_field1 * .01)::numeric AS field1' it works, but 'base_table_field1 AS field1' would not because the column type changes. This approach might help in some cases like mine.

Answer 5

I wanted to comment on the second answer but cannot since I'm too new to stackoverflow, so here my comment: To those interested in the original article mentioned in that answer, the blogspot entry is not available any more but the wayback machine has it still stored: https://web.archive.org/web/20180323155900/http://mwenus.blogspot.com/2014/04/postgresql-how-to-handle-table-and-view.html Here is the article itself in case archive.org should be turned off at some future point in time: 2014-04-22 PostgreSQL: How to handle table and view dependencies PostgreSQL is very restrictive when it comes to modyfing existing objects. Very often when you try to ALTER TABLE or REPLACE VIEW it tells you that you cannot do it, because there's another object (typically a view or materialized view), which depends on the one you want to modify. It seems that the only solution is to DROP dependent objects, make desired changes to the target object and then recreate dropped objects.

It is tedious and cumbersome, because those dependent objects can have further dependencies, which also may have other dependencies and so on. I created utility functions which can help in such situations.

The usage is very simple - you just have to call: select deps_save_and_drop_dependencies(p_schema_name, p_object_name); You have to pass two arguments: the name of the schema and the name of the object in that schema. This object can be a table, a view or a materialized view. The function will drop all views and materialized views dependent on p_schema_name.p_object_name and save DDL which restores them in a helper table.

When you want to restore those dropped objects (for example when you are done modyfing p_schema_name.p_object_name), you just need to make another simple call: select deps_restore_dependencies(p_schema_name, p_object_name); and the dropped objects will be recreated.

These functions take care about: dependencies hierarchy proper order of dropping and creating views/materialized views across hierarchy restoring comments and grants on views/materialized views Click here for a working sqlfiddle example or check this gist for a complete source code.

Autor: Mateusz Wenus o 19:32

Answer 6

do $$            
  declare gorev_lisans_ihlali_def text;
  declare exec_text text;
begin          
  gorev_lisans_ihlali_def := pg_get_viewdef('public.gorev_lisans_ihlali');
  drop view public.gorev_lisans_ihlali;

    
  exec_text := format('create view public.gorev_lisans_ihlali as %s', 
     gorev_lisans_ihlali_def);
      ALTER TABLE public.ara_bakis_duyma
        ALTER COLUMN gain TYPE   DOUBLE PRECISION;
  execute exec_text;
end $$;