I have a table created like
CREATE TABLE data
(value1 smallint references labels,
value2 smallint references labels,
value3 smallint references labels,
otherdata varchar(32)
);
and a second 'label holding' table created like
CREATE TABLE labels (id serial primary key, name varchar(32));
The rationale behind it is that value1-3 are a very limited set of strings (6 options) and it seems inefficient to enter them directly in the data table as varchar types. On the other hand these do occasionally change, which makes enum types unsuitable.
My question is, how can I execute a single query such that instead of the label IDs I get the relevant labels?
I looked at creating a function for it and stumbled at the point where I needed to pass the label holding table name to the function (there are several such (label holding) tables across the schema). Do I need to create a function per label table to avoid that?
create or replace function translate
(ref_id smallint,reference_table regclass) returns varchar(128) as
$$
begin
select name from reference_table where id = ref_id;
return name;
end;
$$
language plpgsql;
And then do
select
translate(value1, labels) as foo,
translate(value2, labels) as bar
from data;
This however errors out with
ERROR: relation "reference_table" does not exist
All suggestions welcome - at this point a can still alter just about anything...
CREATE TABLE labels
( id smallserial primary key
, name varchar(32) UNIQUE -- <<-- might want this, too
);
CREATE TABLE data
( value1 smallint NOT NULL REFERENCES labels(id) -- <<-- here
, value2 smallint NOT NULL REFERENCES labels(id)
, value3 smallint NOT NULL REFERENCES labels(id)
, otherdata varchar(32)
, PRIMARY KEY (value1,value2,value3) -- <<-- added primary key here
);
-- No need for a function here.
-- For small sizes of the `labels` table, the query below will always
-- result in hash-joins to perform the lookups.
SELECT l1.name AS name1, l2.name AS name2, l3.name AS name3
, d.otherdata AS the_data
FROM data d
JOIN labels l1 ON l1.id = d.value1
JOIN labels l2 ON l2.id = d.value2
JOIN labels l3 ON l3.id = d.value3
;
Note: labels.id -> labels.name is a functional dependency (id is the primary key), but that doesn't mean that you need a function. The query just acts like a function.
You can pass the label table name as string, construct a query as string and execute it:
sql = `select name from ` || reference_table_name || `where id = ` || ref_id;
EXECUTE sql INTO name;
RETURN name;
Related
currently I try to make a history table based on postgresql jsonb, currently as a example I have two table's:
CREATE TABLE data (id BIGSERIAL PRIMARY KEY, price NUMERIC(10,4) NOT NULL, article TEXT NOT NULL, quantity BIGINT NOT NULL, lose BIGINT NOT NULL, username TEXT NOT NULL);
CREATE TABLE data_history (id BIGSERIAL PRIMARY KEY, data JSONB NOT NULL, username TEXT NOT NULL);
The history table act's a simple history (the username there could be avoided).
I populate the data of the history with a trigger:
CREATE OR REPLACE FUNCTION insert_history() RETURNS TRIGGER AS $$
BEGIN
INSERT INTO data_history (data, username) VALUES (row_to_json(NEW.*), NEW.username);
RETURN NEW;
END;
$$ LANGUAGE plpgsql;
Now I try to populate the history back to the data table:
SELECT jsonb_populate_record(NULL::data, data) FROM data_history;
However the result will now be a tuple and not a table:
jsonb_populate_record
-------------------------------------
(1,45.4500,0A45477,100,1,c.schmitt)
(2,5.4500,0A45477,100,1,c.schmitt)
(2 rows)
Is there any way to get the data back as the table data back? I know there is jsonb_populate_recordset, too, however it doesn't accept a query?!
jsonb_populate_record() returns a row-type (or record-type), so if you use it in the SELECT cluase, you'll get a single column, which is a row-type.
To avoid this, use it in the FROM clause instead (with an implicit LATERAL JOIN):
SELECT r.*
FROM data_history,
jsonb_populate_record(NULL::data, data) r
Technically, the statement below could work too
-- DO NOT use, just for illustration
SELECT jsonb_populate_record(NULL::data, data).*
FROM data_history
but it will call jsonb_populate_record() for each column in data (as a result of an engine limitation).
I have a table in postgres:
create table fubar (
name1 text,
name2 text, ...,
key integer);
I want to write a function which returns field values from fubar given the column names:
function getFubarValues(col_name text, key integer) returns text ...
where getFubarValues returns the value of the specified column in the row identified by key. Seems like this should be easy.
I'm at a loss. Can someone help? Thanks.
Klin's answer is a good (i.e. safe) approach to the question as posed, but it can be simplified:
PostgreSQL's -> operator allows expressions. For example:
CREATE TABLE test (
id SERIAL,
js JSON NOT NULL,
k TEXT NOT NULL
);
INSERT INTO test (js,k) VALUES ('{"abc":"def","ghi":"jkl"}','abc');
SELECT js->k AS value FROM test;
Produces
value
-------
"def"
So we can combine that with row_to_json:
CREATE TABLE test (
id SERIAL,
a TEXT,
b TEXT,
k TEXT NOT NULL
);
INSERT INTO test (a,b,k) VALUES
('foo','bar','a'),
('zip','zag','b');
SELECT row_to_json(test)->k AS value FROM test;
Produces:
value
-------
"foo"
"zag"
Here I'm getting the key from the table itself but of course you could get it from any source / expression. It's just a value. Also note that the result returned is a JSON value type (it doesn't know if it's text, numeric, or boolean). If you want it to be text, just cast it: (row_to_json(test)->k)::TEXT
Now that the question itself is answered, here's why you shouldn't do this, and what you should do instead!
Never trust any data. Even if it already lives inside your database, you shouldn't trust it. The method I've posted here is safe against SQL injection attacks, but an attacker could still set k to 'id' and see a column which was not intended to be visible to them.
A much better approach is to structure your data with this type of query in mind. Postgres has some excellent datatypes for this; HSTORE and JSON/JSONB. Merge your dynamic columns into a single column with one of those types (I'd suggest HSTORE for its simplicity and generally being more complete).
This has several advantages: your schema is well-defined and does not need to change if you add more dynamic columns, you do not need to perform expensive re-casting (i.e. row_to_json), and you are able to take advantage of indexes on your columns (thanks to PostgreSQL's functional indexes).
The equivalent to the code I wrote above would be:
CREATE EXTENSION HSTORE; -- necessary if you're not already using HSTORE
CREATE TABLE test (
id SERIAL,
cols HSTORE NOT NULL,
k TEXT NOT NULL
);
INSERT INTO test (cols,k) VALUES
('a=>"foo",b=>"bar"','a'),
('a=>"zip",b=>"zag"','b');
SELECT cols->k AS value FROM test;
Or, for automatic escaping of your values when inserting, you can use one of:
INSERT INTO test (cols,k) VALUES
(hstore( 'a', 'foo' ) || hstore( 'b', 'bar' ), 'a'),
(hstore( ARRAY['a','b'], ARRAY['zip','zag'] ), 'b');
See http://www.postgresql.org/docs/9.1/static/hstore.html for more details.
You can use dynamic SQL to select a column by name:
create or replace function get_fubar_values (col_name text, row_key integer)
returns setof text language plpgsql as $$begin
return query execute 'select ' || quote_ident(col_name) ||
' from fubar where key = $1' using row_key;
end$$;
This is my table :
CREATE TABLE [dbo].[TestTable]
(
[Name1] varchar(50) COLLATE French_CI_AS NOT NULL,
[Name2] varchar(255) COLLATE French_CI_AS NULL,
CONSTRAINT [TestTable_uniqueName1] UNIQUE ([Name1]),
CONSTRAINT [TestTable_uniqueName1Name2] UNIQUE ([Name1], [Name2])
)
ALTER TABLE [dbo].[TestTable]
ADD CONSTRAINT [TestTable_uniqueName1]
UNIQUE NONCLUSTERED ([Name1])
ALTER TABLE [dbo].[TestTable]
ADD CONSTRAINT [TestTable_uniqueName1Name2]
UNIQUE NONCLUSTERED ([Name1], [Name2])
GO
ALTER INDEX [TestTable_uniqueName1]
ON [dbo].[TestTable]
DISABLE
GO
My idea is to enable/disable one or other unique contraint depending on the customer application. With this way, I can catch the thrown exception in my c# code, and display a specific error message to the GUI.
Now, my problem is to alter the collation of columns Name1 & Name2, I need to make them case sensitive (French_CS_AS). To alter these fields, I have to drop the two constraints and recreate it. According to the explained schema, I cannot create an enabled constraint and then disable it, because by some customers, I have duplicate keys for one or other constraint.
For my update script, my idea number 1 was
Save the name of enabled constraints in a temp table
Drop the constraints
Alter columns
Create DISABLED unique constraints
Enable specific constraints according to the saved values in points 1.
My problem is in point 4., I don't find how to create a disabled unique constraint with an ALTER TABLE statement. Is it possible to create it directly in the sys.indexes table ?
My idea number 2 was
Rename TestTable to TestTableCopy
Recreate TestTable with the new fields collation, and otherwise the same schema (indexes, FK, triggers, ...)
Disable specifical unique contraints in TestTable
Migrate data from TestTableCopy to TestTable
Drop TestTableCopy
In this way, my fear is to loose some links with other tables/dependencies, beceause it is a central table in my database.
Is there any other way to achieve my goal?
If necessary, I can use unique indexes instead of unique constraints.
It looks like it is impossible to create a unique index on a column that already has duplicate values.
So, rather than having a disabled unique index either:
not have an index at all (which is the same as having a disabled index from the query processor point of view),
or create a non-unique index.
For those instanses where your client has unique data create unique index. For those instanses where your client has non-unique data create non-unique index.
CREATE PROCEDURE [dbo].[spUsers_AddUsers]
#Name1 varchar(50) ,
#Name2 varchar(50) ,
#Unique bit
AS
declare #err int
begin tran
if #Unique = 1 begin
if not exists (SELECT * FROM Users WHERE Name1 = #Name1 and Name2 = #Name2)
begin
INSERT INTO Users (Name1,Name2)
VALUES (#Name1,#Name2)
set #err = ##ERROR
end else
begin
UPDATE Users
set Name1 = #Name1,
Name2 = #Name2
where Name1 = #Name1 and Name2 = #Name2
set #err = ##ERROR
end
end else begin
if not exists ( SELECT * FROM Users WHERE Name1 = #Name1 )
begin
INSERT INTO Users (Name1,Name2)
VALUES (#Name1,#Name2)
set #err = ##ERROR
end else
begin
UPDATE Users
set Name1 = #Name1,
Name2 = #Name2
where Name1 = #Name1
set #err = ##ERROR
end
if #err = 0 commit tran
else rollback tran
So first you check if you need an unique Name1 and Name2 or just Name1. Then if you do you an insert/update based on what constrain you have.
CREATE TABLE test ( id int PRIMARY KEY , name );
CREATE TABLE test1 ( id integer[] REFERENCES test , rollid int );
ERROR: foreign key constraint "test3_id_fkey" cannot be implemented
DETAIL: Key columns "id" and "id" are of incompatible types: integer[] and integer.
after that I try to another way also
CREATE TABLE test1 ( id integer[] , rollid int);
ALTER TABLE test1 ADD CONSTRAINT foreignkeyarray FOREIGN KEY (id) REFERENCES test;
ERROR: foreign key constraint "fkarray" cannot be implemented
DETAIL: Key columns "id" and "id" are of incompatible types: integer[] and integer.
so I try create a foreign key array means it say error. please tell me anyone?
postgresql version is 9.1.
What you're trying to do simply can't be done. At all. No ifs, no buts.
Create a new table, test1_test, containing two fields, test1_id, test_id. Put the foreign keys as needed on that one, and make test1's id an integer.
Using arrays with foreign element keys is usually a sign of incorrect design. You need to do separate table with one to many relationship.
But technically it is possible. Example of checking array values without triggers. One reusable function with paramethers and dynamic sql. Tested on PostgreSQL 10.5
create schema if not exists test;
CREATE OR REPLACE FUNCTION test.check_foreign_key_array(data anyarray, ref_schema text, ref_table text, ref_column text)
RETURNS BOOL
RETURNS NULL ON NULL INPUT
LANGUAGE plpgsql
AS
$body$
DECLARE
fake_id text;
sql text default format($$
select id::text
from unnest($1) as x(id)
where id is not null
and id not in (select %3$I
from %1$I.%2$I
where %3$I = any($1))
limit 1;
$$, ref_schema, ref_table, ref_column);
BEGIN
EXECUTE sql USING data INTO fake_id;
IF (fake_id IS NOT NULL) THEN
RAISE NOTICE 'Array element value % does not exist in column %.%.%', fake_id, ref_schema, ref_table, ref_column;
RETURN false;
END IF;
RETURN true;
END
$body$;
drop table if exists test.t1, test.t2;
create table test.t1 (
id integer generated by default as identity primary key
);
create table test.t2 (
id integer generated by default as identity primary key,
t1_ids integer[] not null check (test.check_foreign_key_array(t1_ids, 'test', 't1', 'id'))
);
insert into test.t1 (id) values (default), (default), (default); --ok
insert into test.t2 (id, t1_ids) values (default, array[1,2,3]); --ok
insert into test.t2 (id, t1_ids) values (default, array[1,2,3,555]); --error
If you are able to put there just values from test.id, then you can try this:
CREATE OR REPLACE FUNCTION test_trigger() RETURNS trigger
LANGUAGE plpgsql AS $BODY$
DECLARE
val integer;
BEGIN
SELECT id INTO val
FROM (
SELECT UNNEST(id) AS id
FROM test1
) AS q
WHERE id = OLD.id;
IF val IS NULL THEN RETURN OLD;
ELSE
RAISE 'Integrity Constraint Violation: ID "%" in Test1', val USING ERRCODE = '23000';
RETURN NULL;
END IF;
END; $BODY$;
-- DROP TRIGGER test_delete_trigger ON test;
CREATE TRIGGER test_delete_trigger BEFORE DELETE OR UPDATE OF id ON test
FOR EACH ROW EXECUTE PROCEDURE test_trigger();
I currently have a parent table:
CREATE TABLE members (
member_id SERIAL NOT NULL, UNIQUE, PRIMARY KEY
first_name varchar(20)
last_name varchar(20)
address address (composite type)
contact_numbers varchar(11)[3]
date_joined date
type varchar(5)
);
and two related tables:
CREATE TABLE basic_member (
activities varchar[3])
INHERITS (members)
);
CREATE TABLE full_member (
activities varchar[])
INHERITS (members)
);
If the type is full the details are entered to the full_member table or if type is basic into the basic_member table. What I want is that if I run an update and change the type to basic or full the tuple goes into the corresponding table.
I was wondering if I could do this with a rule like:
CREATE RULE tuple_swap_full
AS ON UPDATE TO full_member
WHERE new.type = 'basic'
INSERT INTO basic_member VALUES (old.member_id, old.first_name, old.last_name,
old.address, old.contact_numbers, old.date_joined, new.type, old.activities);
... then delete the record from the full_member
Just wondering if my rule is anywhere near or if there is a better way.
You don't need
member_id SERIAL NOT NULL, UNIQUE, PRIMARY KEY
A PRIMARY KEY implies UNIQUE NOT NULL automatically:
member_id SERIAL PRIMARY KEY
I wouldn't use hard coded max length of varchar(20). Just use text and add a check constraint if you really must enforce a maximum length. Easier to change around.
Syntax for INHERITS is mangled. The key word goes outside the parens around columns.
CREATE TABLE full_member (
activities text[]
) INHERITS (members);
Table names are inconsistent (members <-> member). I use the singular form everywhere in my test case.
Finally, I would not use a RULE for the task. A trigger AFTER UPDATE seems preferable.
Consider the following
Test case:
Tables:
CREATE SCHEMA x; -- I put everything in a test schema named "x".
-- DROP TABLE x.members CASCADE;
CREATE TABLE x.member (
member_id SERIAL PRIMARY KEY
,first_name text
-- more columns ...
,type text);
CREATE TABLE x.basic_member (
activities text[3]
) INHERITS (x.member);
CREATE TABLE x.full_member (
activities text[]
) INHERITS (x.member);
Trigger function:
Data-modifying CTEs (WITH x AS ( DELETE ..) are the best tool for the purpose. Requires PostgreSQL 9.1 or later.
For older versions, first INSERT then DELETE.
CREATE OR REPLACE FUNCTION x.trg_move_member()
RETURNS trigger AS
$BODY$
BEGIN
CASE NEW.type
WHEN 'basic' THEN
WITH x AS (
DELETE FROM x.member
WHERE member_id = NEW.member_id
RETURNING *
)
INSERT INTO x.basic_member (member_id, first_name, type) -- more columns
SELECT member_id, first_name, type -- more columns
FROM x;
WHEN 'full' THEN
WITH x AS (
DELETE FROM x.member
WHERE member_id = NEW.member_id
RETURNING *
)
INSERT INTO x.full_member (member_id, first_name, type) -- more columns
SELECT member_id, first_name, type -- more columns
FROM x;
END CASE;
RETURN NULL;
END;
$BODY$
LANGUAGE plpgsql VOLATILE;
Trigger:
Note that it is an AFTER trigger and has a WHEN condition.
WHEN condition requires PostgreSQL 9.0 or later. For earlier versions, you can just leave it away, the CASE statement in the trigger itself takes care of it.
CREATE TRIGGER up_aft
AFTER UPDATE
ON x.member
FOR EACH ROW
WHEN (NEW.type IN ('basic ','full')) -- OLD.type cannot be IN ('basic ','full')
EXECUTE PROCEDURE x.trg_move_member();
Test:
INSERT INTO x.member (first_name, type) VALUES ('peter', NULL);
UPDATE x.member SET type = 'full' WHERE first_name = 'peter';
SELECT * FROM ONLY x.member;
SELECT * FROM x.basic_member;
SELECT * FROM x.full_member;