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?
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.
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;
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
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.
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
do $$
declare gorev_lisans_ihlali_def text;
declare exec_text text;
begin
gorev_lisans_ihlali_def := pg_get_viewdef('public.gorev_lisans_ihlali');
drop view public.gorev_lisans_ihlali;
exec_text := format('create view public.gorev_lisans_ihlali as %s',
gorev_lisans_ihlali_def);
ALTER TABLE public.ara_bakis_duyma
ALTER COLUMN gain TYPE DOUBLE PRECISION;
execute exec_text;
end $$;
Related
All I need is to get a SQL query output as :
ALTER TABLE TABLE_NAME
ADD CONSTRAINT
FOREIGN KEY (COLUMN_NAME)
REFERENCES (PARENT_TABLE_NAME);
I'm running the below DYNAMIC query USING DATA DICTIONARY TABLES,
SELECT DISTINCT
'ALTER TABLE ' || cs.TABLE_NAME ||
'ADD CONSTRAINT' || rc.CONSTRAINT_NAME ||
'FOREIGN KEY' || c.COLUMN_NAME ||
'REFERENCES' || cs.TABLE_NAME ||
' (' || cs.CONSTRAINT_NAME || ') ' ||
' ON UPDATE ' || rc.UPDATE_RULE ||
' ON DELETE ' || rc.DELETE_RULE
FROM INFORMATION_SCHEMA.REFERENTIAL_CONSTRAINTS RC,
INFORMATION_SCHEMA.TABLE_CONSTRAINTS CS,
INFORMATION_SCHEMA.COLUMNS C
WHERE cs.CONSTRAINT_NAME = rc.CONSTRAINT_NAME
AND cs.TABLE_NAME = c.TABLE_NAME
AND UPPER(cs.TABLE_SCHEMA) = 'SSP2_PCAT';
But here even though I'm able to generate the desired output, the concern is its not giving the PARENT_TABLE_NAME here,
rather its giving the same table_name after the ALTER TABLE Keywords.
I hope this is clear as we are using Dynamic SQL here and any help is absolutely appreciated!
Your query is missing a couple of join tables and join conditions. Also, don't forget that a foreign key can be defined on more than one column. Finally, your query is vulnerable to SQL injection via object names.
But it would be much simpler if you used pg_catalog.pg_constraint rather than the `information_schema':
SELECT format('ALTER TABLE %s ADD CONSTRAINT %I %s',
conrelid::regclass,
conname,
pg_get_constraintdef(oid))
FROM pg_catalog.pg_constraint
WHERE contype = 'f'
AND upper(connamespace::regnamespace::text) = 'SSP2_PCAT';
I am trying to write sub-queries so that I search all tables for a column named id and since there are multiple tables with id column, I want to add the condition, so that id = 3119093.
My attempt was:
Select *
from information_schema.tables
where id = '3119093' and id IN (
Select table_name
from information_schema.columns
where column_name = 'id' );
This didn't work so I tried:
Select *
from information_schema.tables
where table_name IN (
Select table_name
from information_schema.columns
where column_name = 'id' and 'id' IN (
Select * from table_name where 'id' = 3119093));
This isn't the right way either. Any help would be appreciated. Thanks!
A harder attempt is:
CREATE OR REPLACE FUNCTION search_columns(
needle text,
haystack_tables name[] default '{}',
haystack_schema name[] default '{public}'
)
RETURNS table(schemaname text, tablename text, columnname text, rowctid text)
AS $$
begin
FOR schemaname,tablename,columnname IN
SELECT c.table_schema,c.table_name,c.column_name
FROM information_schema.columns c
JOIN information_schema.tables t ON
(t.table_name=c.table_name AND t.table_schema=c.table_schema)
WHERE (c.table_name=ANY(haystack_tables) OR haystack_tables='{}')
AND c.table_schema=ANY(haystack_schema)
AND t.table_type='BASE TABLE'
--AND c.column_name = "id"
LOOP
EXECUTE format('SELECT ctid FROM %I.%I WHERE cast(%I as text) like %L',
schemaname,
tablename,
columnname,
needle
) INTO rowctid;
IF rowctid is not null THEN
RETURN NEXT;
END IF;
END LOOP;
END;
$$ language plpgsql;
select * from search_columns('%3119093%'::varchar,'{}'::name[]) ;
The only problem is this code displays the table name and column name. I have to then manually enter
Select * from table_name where id = 3119093
where I got the table name from the code above.
I want to automatically implement returning rows from a table but I don't know how to get the table name automatically.
I took the time to make it work for you.
For starters, some information on what is going on inside the code.
Explanation
function takes two input arguments: column name and column value
it requires a created type that it will be returning a set of
first loop identifies tables that have a column name specified as the input argument
then it forms a query which aggregates all rows that match the input condition inside every table taken from step 3 with comparison based on ILIKE - as per your example
function goes into the second loop only if there is at least one row in currently visited table that matches specified condition (then the array is not null)
second loop unnests the array of rows that match the condition and for every element it puts it in the function output with RETURN NEXT rec clause
Notes
Searching with LIKE is inefficient - I suggest adding another input argument "column type" and restrict it in the lookup by adding a join to pg_catalog.pg_type table.
The second loop is there so that if more than 1 row is found for a particular table, then every row gets returned.
If you are looking for something else, like you need key-value pairs, not just the values, then you need to extend the function. You could for example build json format from rows.
Now, to the code.
Test case
CREATE TABLE tbl1 (col1 int, id int); -- does contain values
CREATE TABLE tbl2 (col1 int, col2 int); -- doesn't contain column "id"
CREATE TABLE tbl3 (id int, col5 int); -- doesn't contain values
INSERT INTO tbl1 (col1, id)
VALUES (1, 5), (1, 33), (1, 25);
Table stores data:
postgres=# select * From tbl1;
col1 | id
------+----
1 | 5
1 | 33
1 | 25
(3 rows)
Creating type
CREATE TYPE sometype AS ( schemaname text, tablename text, colname text, entirerow text );
Function code
CREATE OR REPLACE FUNCTION search_tables_for_column (
v_column_name text
, v_column_value text
)
RETURNS SETOF sometype
LANGUAGE plpgsql
STABLE
AS
$$
DECLARE
rec sometype%rowtype;
v_row_array text[];
rec2 record;
arr_el text;
BEGIN
FOR rec IN
SELECT
nam.nspname AS schemaname
, cls.relname AS tablename
, att.attname AS colname
, null::text AS entirerow
FROM
pg_attribute att
JOIN pg_class cls ON att.attrelid = cls.oid
JOIN pg_namespace nam ON cls.relnamespace = nam.oid
WHERE
cls.relkind = 'r'
AND att.attname = v_column_name
LOOP
EXECUTE format('SELECT ARRAY_AGG(row(tablename.*)::text) FROM %I.%I AS tablename WHERE %I::text ILIKE %s',
rec.schemaname, rec.tablename, rec.colname, quote_literal(concat('%',v_column_value,'%'))) INTO v_row_array;
IF v_row_array is not null THEN
FOR rec2 IN
SELECT unnest(v_row_array) AS one_row
LOOP
rec.entirerow := rec2.one_row;
RETURN NEXT rec;
END LOOP;
END IF;
END LOOP;
END
$$;
Exemplary call & output
postgres=# select * from search_tables_for_column('id','5');
schemaname | tablename | colname | entirerow
------------+-----------+---------+-----------
public | tbl1 | id | (1,5)
public | tbl1 | id | (1,25)
(2 rows)
This is a simple example of what I need, for any given table, I need to get all the instances of the primary keys, this is a little example, but I need a generic way to do it.
create table foo
(
a numeric
,b text
,c numeric
constraint pk_foo primary key (a,b)
)
insert into foo(a,b,c) values (1,'a',1),(2,'b',2),(3,'c',3);
select <the magical thing>
result
a|b
1 |1|a|
2 |2|b|
3 |3|c|
.. ...
I need to control if the instances of the primary keys are changed by the user, but I don't want to repeat code in too many tables! I need a generic way to do it, I will put <the magical thing>
in a function to put it on a trigger before update and blah blah blah...
In PostgreSQL you must always provide a resulting type for a query. However, you can obtain the code of the query you need, and then execute the query from the client:
create or replace function get_key_only_sql(regclass) returns string as $$
select 'select '|| (
select string_agg(quote_ident(att.attname), ', ' order by col)
from pg_index i
join lateral unnest(indkey) col on (true)
join pg_attribute att on (att.attrelid = i.indrelid and att.attnum = col)
where i.indrelid = $1 and i.indisprimary
group by i.indexrelid
limit 1) || ' from '||$1::text
end;
$$ language sql;
Here's some client pseudocode using the function above:
sql = pgexecscalar("select get_key_only_sql('mytable'::regclass)");
rs = pgopen(sql);
I have a table with a column storing table OID of table where records come from. If this field must be filled (NOT NULL) but can have a default value if not provided. I would like to store the current table OID when inserting.
CREATE TABLE t AS(
Source regclass NOT NULL DEFAULT current_table_name()::regclass
);
Is there any function (current_table_name) in postgreSQL table to perform this task?
There is no such function, but you can achieve the goal in three steps:
Create the table without DEFAULT.
Find the OID of the new table.
ALTER the table and set the DEFAULT.
It is better to use the table OID than to cast the table name to regclass, because with the latter INSERTs will suddenly start failing after the table is renamed.
Here is a DO block that would achieve that:
DO $$DECLARE
reloid oid;
BEGIN
CREATE TABLE t (source regclass NOT NULL);
SELECT t.oid INTO reloid
FROM pg_class t
JOIN pg_namespace n ON t.relnamespace = n.oid
WHERE t.relname = 't' AND n.nspname = current_schema;
EXECUTE 'ALTER TABLE t ALTER source SET DEFAULT ' || reloid;
END;$$;
After the excellent answer by Alexandre GUIDET, I attempted to run the following query:
create table egg (id (SELECT
pg_catalog.format_type(a.atttypid, a.atttypmod) as Datatype
FROM
pg_catalog.pg_attribute a
WHERE
a.attnum > 0
AND NOT a.attisdropped
AND a.attrelid = (
SELECT c.oid
FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE c.relname ~ '^(TABLENAME)$'
AND pg_catalog.pg_table_is_visible(c.oid)
)
and a.attname = 'COLUMNNAME'));
PostgreSQL, however, complains about incorrect syntax. Specifically it says that I cannot write: create table egg (id (SELECT.
Are there any workarounds? Can't I convert the result of a query to text and reuse it as a query?
There is a much simpler way to do that.
SELECT pg_typeof(col)::text FROM tbl LIMIT 1
Only precondition is that the template table holds at least one row. See the manual on pg_typeof()
As Milen wrote, you need to EXECUTE dynamic DDL statements like this.
A much simpler DO statement:
DO $$BEGIN
EXECUTE 'CREATE TABLE egg (id '
|| (SELECT pg_typeof(col)::text FROM tbl LIMIT 1) || ')';
END$$;
Or, if you are not sure the template table has any rows:
DO $$BEGIN
EXECUTE (
SELECT format('CREATE TABLE egg (id %s)'
, format_type(atttypid, atttypmod))
FROM pg_catalog.pg_attribute
WHERE attrelid = 'tbl'::regclass -- name of template table
AND attname = 'col' -- name of template column
AND attnum > 0 AND NOT attisdropped
);
END$$;
These conditions seem redundant, since you look for a specific column any
format() requires Postgres 9.1+.
Related:
How to check if a table exists in a given schema
You can either convert that query to a function or (if you have Postgres 9.0) to an anonymous code block:
DO $$DECLARE the_type text;
BEGIN
SELECT ... AS datatype INTO the_type FROM <the rest of your query>;
EXECUTE 'create table egg ( id ' || the_type || <the rest of your create table statement>;
END$$;
You can either have a table a definition or a query, but not both. Maybe your thinking of the select into command.