I need to create a function in PostgreSQL for the following :
Query multiple tables based on a business logic (all result sets return the same type of data)
Compile all result sets into one table and return that table
Is it possible to accomplish this without using the temp tables in PostgreSQL?
I currently do this in Microsoft SQL server using Table Variables, below is a sample function:
create FUNCTION test(#search_in nvarchar(500))
RETURNS #data_table TABLE
(
item_id int,
item_type nvarchar(1),
first_name nvarchar(100),
last_name nvarchar(100))
) AS
BEGIN
-- from first table
if charindex('search_in_authors', #search_in) > 0
insert into #data_table
select item_id, 'a', first_name, last_name
from authors
where first_name = 'james'
-- from second table
if charindex('search_in_editors', #search_in) > 0
insert into #data_table
select item_id, 'e', first_name, last_name
from editors
where first_name = 'james'
-- from third table
if charindex('search_in_publishers', #search_in) > 0
insert into #data_table
select item_id, 'p', first_name, last_name
from publishes
where first_name = 'james'
-- there could be more like these based on the business logic...
(...)
-- finally return the records compiled in #data_table
RETURN
END
Sample calls to the function:
select * from dbo.test('search_in_authors')
select * from dbo.test('search_in_authors, search_in_editors')
select * from dbo.test('search_in_authors, search_in_editors,search_in_publishers ')
Are there any options in PostgreSQL to achieve this other than using a temp table ?
Thanks,
San
You can use RETURN QUERY to add the result of various queries to the output.
CREATE OR REPLACE FUNCTION public.testf()
RETURNS TABLE(id INTEGER, name text)
STABLE
AS $$
DECLARE
BEGIN
RETURN QUERY select 1 as id, 'abc' as name;
RETURN QUERY select 2 as id, 'def' as name;
RETURN QUERY select 3 as id, 'xyz' as name;
-- Final return as set is now complete.
return;
END;
$$ LANGUAGE plpgsql;
select * from public.testf();
id | name
----+------
1 | abc
2 | def
3 | xyz
(3 rows)
Related
I know I didn't make it clear in the question title. Let me explain.
Say I have a Table SOURCE_TABLE with 1 column that looks like this:
Filter
------------------|
Name='John'
Surname = 'Smith'
Age = '25'
And I want to use this table as a filter. Like below:
SELECT * FROM TARGET_TABLE WHERE (SELECT FILTER FROM SOURCE_TABLE)
I heard maybe evaluate function could help me but to be honest I couldn't understand how.
Do you know any method that I can use a column as my filter source like above?
Edit1:
DECLARE
my_filter VARCHAR2(100);
my_query VARCHAR2(500);
BEGIN
my_query := 'SELECT FILTER FROM SOURCE_TABLE WHERE ROWNUM=1';
EXECUTE IMMEDIATE my_query INTO my_filter;
EXECUTE IMMEDIATE 'SELECT * FROM TARGET_TABLE WHERE '|| my_filter;
END;
#Sujitmohanty30 I have come up with the above after learning EXECUTE IMMEDIATE. But there is a problem I stumble upon. This needs to be dynamic regarding to the end result and i want to see the result of the select query at the end.
Let's say we have these 2 tables:
create table TARGET_TABLE(name varchar2(30), surname varchar2(30));
insert into TARGET_TABLE(name,surname) values ('John', 'Doe');
insert into TARGET_TABLE(name,surname) values ('Ann', 'Smith');
insert into TARGET_TABLE(name,surname) values ('Steven', 'Feuerstein');
insert into TARGET_TABLE(name,surname) values ('Steven', 'King');
commit;
create table SOURCE_TABLE(filter) as
select q'[name='John']' from dual union all
select q'[name='Ann']' from dual union all
select q'[name='Steven' and surname='King']' from dual union all
select q'[surname='Feuerstein']' from dual;
Then you can use xmltable and dbms_xmlgen.getXMLtype to get such rows dynamically:
https://dbfiddle.uk/?rdbms=oracle_18&fiddle=72abdf18b149cf30882cb4e1736c9c33
select *
from SOURCE_TABLE
, xmltable(
'/ROWSET/ROW'
passing dbms_xmlgen.getXMLtype('select * from TARGET_TABLE where '|| SOURCE_TABLE.filter)
columns
name varchar2(30) path 'NAME',
surname varchar2(30) path 'SURNAME'
) x;
Results:
FILTER NAME SURNAME
-------------------------------- ------------------------------ ------------------------------
name='John' John Doe
name='Ann' Ann Smith
name='Steven' and surname='King' Steven King
surname='Feuerstein' Steven Feuerstein
I am working through an intro SQL textbook and am confused by the following problem, where we are given the table and values:
CREATE TABLE LineageTable (
parent INT,
id INT,
genus_name VARCHAR(30),
PRIMARY KEY (id)
);
INSERT INTO LineageTable VALUES
(3, 1, 'FamilyA'),
(2, 4, 'FamilyB'),
(7, 2, 'FamilyC');
And I want to write a function that will return a text string representing the path from the a given name to the desired root
My Attempt:
CREATE FUNCTION LineageTable (input VARCHAR(50))
RETURNS TABLE (input VARCHAR(50))
AS $$
BEGIN
RETURN QUERY
SELECT input
FROM LineageTable1
INNER JOIN LineageTable ON LineageTable.parent = LineageTable.id
WHERE LineageTable1.genus_name = LineageTable1.genus_name;
END $$
However, I am confused as how to iterate through this table multiple times to string the path together properly. Any ideas? Thanks all!
On Postgres you can use a RECURSIVE query:
WITH RECURSIVE Rec as
(
SELECT id, parent_id, Name
FROM Hierarchy
WHERE Name = 'Sirenia'
UNION ALL
SELECT Hierarchy.id, Hierarchy.parent_id, Hierarchy.Name
FROM Hierarchy
INNER JOIN Rec
ON Hierarchy.id = Rec.parent_Id
)
SELECT string_agg(Name, '->') path
FROM Rec;
| path |
|:---------------------------------:|
| Sirenia->Paenungulata->Afrotheria |
Rextester here
I have a postgres database that has several tables (a few hundreds). A subset Foo of the tables in the database have the same schema.
Ideally, I would like to create a stored procedure which can run a query against a single table, or against all tables in subset Foo.
Pseudocode:
CREATE TABLE tbl_a (id INTEGER, name VARCHAR(32), weight double, age INTEGER);
CREATE TABLE tbl_b (id INTEGER, name VARCHAR(32), weight double, age INTEGER);
CREATE TABLE tbl_c (id INTEGER, name VARCHAR(32), weight double, age INTEGER);
CREATE TABLE tbl_d (id INTEGER, name VARCHAR(32), weight double, age INTEGER);
CREATE TYPE person_info AS (id INTEGER, name VARCHAR(32), weight double, age INTEGER);
CREATE FUNCTION generic_func(ARRAY one_or_more_table_names)
RETURNS person_info
-- Run query on table or all specified tables
AS $$ $$
LANGUAGE SQL;
How could I implement this requirement in Postgresql 9.x ?
You should have a look at table inheritance in PostgreSQL, they allow exactly what you speak about.
For example, you could create a table parent_tbl:
CREATE TABLE parent_tbl (id INTEGER, name VARCHAR(32), weight numeric, age INTEGER);
Then link your tables to this parent table:
ALTER TABLE tbl_a INHERIT parent_tbl;
ALTER TABLE tbl_b INHERIT parent_tbl;
ALTER TABLE tbl_c INHERIT parent_tbl;
ALTER TABLE tbl_d INHERIT parent_tbl;
Then a SELECT query over parent_tbl will query all of tbl_x tables, while a query on tbl_x will query only this particular table.
INSERT INTO tbl_a VALUES (1, 'coucou', 42, 42);
SELECT * FROM tbl_a;
id | name | weight | age
----+--------+--------+-----
1 | coucou | 42 | 42
(1 row)
SELECT * FROM parent_tbl;
id | name | weight | age
----+--------+--------+-----
1 | coucou | 42 | 42
(1 row)
SELECT * FROM tbl_b;
id | name | weight | age
----+--------+--------+-----
(0 rows)
It is also possible to filter data from given children tables. For example, if you are interested in data coming from tables tbl_a and tbl_b, you can do
select id, name, weight, age
from parent_tbl
left join pg_class on oid = parent_tbl.tableoid
where relname in ('tbl_a', 'tbl_b');
EDIT : I put numeric for weight instead of double as this type is not supported on my server.
To create select query dynamically using items(table name) in an array you can use following select statement
SELECT string_agg(q, ' union all ')
FROM (
SELECT 'select * from ' || unnest(array ['tble_a','tble_b']) AS q
) t
Result:
string_agg
---------------------------------------------------
select * from tble_a union all select * from tble_b
You can create the function that returns table with columns
id INTEGER
,name VARCHAR(32)
,weight numeric
,age INTEGER
P.S: I am avoiding TYPE person_info
function:
CREATE
OR REPLACE FUNCTION generic_func (tbl varchar [])
RETURNS TABLE ( -- To store the output
id INTEGER
,name VARCHAR(32)
,weight numeric
,age INTEGER
) AS $BODY$
DECLARE qry text;
BEGIN
SELECT string_agg(q, ' union all ') --To create select query dynamically
INTO qry
FROM (
SELECT 'select * from ' || unnest(tbl) AS q
) t;
RAISE NOTICE 'qry %',qry; --optional
RETURN query --Executes the query to the defined table
EXECUTE qry;
END;$BODY$
LANGUAGE plpgsql VOLATILE
Usage:
select * from generic_func(array['tbl_a','tbl_b','tbl_c','tbl_d'])
Result:
id name weight age
-- ---- ------ ---
2 ABC 11 112
2 CBC 11 112
2 BBC 11 112
2 DBC 11 112
and
select * from generic_func(array['tbl_a'])
Result:
id name weight age
-- ---- ------ ---
2 ABC 11 112
I've been trying to use code which finds the count of elements in a table and stores it in a local variable. I basically just want to check the existence of a record, so if there is any easier way to do this.
Here is an example I found of storing the result of a query in a variable (link):
CREATE FUNCTION checklist( d SMALLINT )
RETURNING VARCHAR(30), VARCHAR(12), INTEGER;
DEFINE name VARCHAR(30);
DEFINE dept VARCHAR(12);
DEFINE num INTEGER;
SELECT mgr_name, department,
CARDINALITY(direct_reports)
FROM manager INTO name, dept, num
WHERE dept_no = d;
IF num > 20 THEN
EXECUTE FUNCTION add_mgr(dept);
ELIF num = 0 THEN
EXECUTE FUNCTION del_mgr(dept);
ELSE
RETURN name, dept, num;
END IF;
END FUNCTION;
But when I try to create my own version of this, I get a syntax error. I have no idea what the problem is.
CREATE FUNCTION test ()
RETURNING INTEGER AS num1;
DEFINE l_count INTEGER;
CREATE TEMP TABLE t_queued_calls (
session_id DEC(18,0) PRIMARY KEY,
calling_number NVARCHAR(50)
) WITH NO LOG;
INSERT INTO t_queued_calls VALUES (123456, '5555555555');
SELECT COUNT(*) FROM t_queued_calls INTO l_count WHERE session_id = 123456;
DROP TABLE t_queued_calls;
END FUNCTION;
The position of the INTO clause is wrong in both functions. The INTO clause goes after the select-list (the list of expressions after the keyword SELECT) and before the FROM clause (see the Informix "Guide to SQL: Syntax" manual on the SELECT statement), as in this code:
CREATE PROCEDURE test()
RETURNING INTEGER AS num1;
DEFINE l_count INTEGER;
CREATE TEMP TABLE t_queued_calls (
session_id DEC(18,0) PRIMARY KEY,
calling_number NVARCHAR(50)
) WITH NO LOG;
INSERT INTO t_queued_calls VALUES (123456, '5555555555');
SELECT COUNT(*) INTO l_count FROM t_queued_calls WHERE session_id = 123456;
DROP TABLE t_queued_calls;
RETURN l_count;
END PROCEDURE;
Also, the first function as shown in the question has the same problem with the ordering of the clauses. Also, it does not always RETURN a value, and the original version of the second function never returns a value (although it says it will).
The could be related to the fact the insert dont have the columns name
adapt your_column1, your_column2 to your table schema
INSERT INTO t_queued_calls (your_column1, your_column2) VALUES (123456, '5555555555');
SELECT COUNT(*) FROM t_queued_calls INTO l_count WHERE session_id = 123456;
And/Or the number of column from the select don't match the number and type in insertt ... you select un field only but insert two field
and select into is strange select format ...normally is insert into but select don't use into clause
I found the correct syntax from this question: [Use MERGE statement in a stored procedure with Informix
SELECT COUNT(*) INTO l_count FROM t_queued_calls WHERE session_id = 123456;
I am having trouble merging a table with a collection.
Let's say I have a table emp.
Here is my PL/SQL code snippet.
TYPE empcol is table of emp%ROWTYPE INDEX BY BINARY_INTEGER;
tmpemp empcol;
-- Code here to load data from a CSV file into tmpemp collection
-- tmpemp(1).emp_id := parsedstring
-- etc.
MERGE INTO emp A using tmpemp B ON A.emp_id = B.emp_id
WHEN MATCHED THEN UPDATE SET A.fname = B.fname, A.lname = B.lname
WHEN NOT MATCHED THEN INSERT (emp_id, fname, lname) VALUES (b.emp_id, b.fname, b.lname);
Compiler doesn't like it. Its throwing ORA-0942 - Table or View doesn't exist.
What am I doing wrong? or How can I do this better.
Thanks a lot for any help you can provide.
PL/SQL types like emp%ROWTYPE or TABLE OF ... INDEX BY ... cannot be used in SQL queries.
The type must be declared as SQL type (not as PL/SQL type) to be used in SQL query.
Try this approach (example):
create table emp(
firstname varchar2(100),
salary number
);
insert into emp values( 'John', 100 );
commit;
create type my_emp_obj is object(
firstname varchar2(100),
salary number
);
/
create type my_emp_obj_table is table of my_emp_obj;
/
declare
my_emp_tab my_emp_obj_table;
begin
null;
my_emp_tab := my_emp_obj_table( my_emp_obj( 'John', 200 ), my_emp_obj( 'Tom', 300 ));
MERGE INTO emp
USING ( SELECT * FROM TABLE( my_emp_tab )) src
ON ( emp.firstname = src.firstname )
WHEN MATCHED THEN UPDATE SET salary = src.salary
WHEN NOT MATCHED THEN INSERT VALUES( src.firstname, src.salary );
end;
/
select * from emp;
FIRSTNAME SALARY
----------------------- ----------
John 200
Tom 300