Get the table name from another table and query on it in Redshift - amazon-web-services

Question

I want to select all data from a table name where table name needs to extract from another table every time.
select * from (select max(table_name) from tableA)
I am trying to write stored procedure for it.
Below is the syntax.
CREATE OR REPLACE PROCEDURE db.stproced() AS $$
declare tablename VARCHAR(200);
BEGIN
select max(table_name) into tableA from easybuy.current_portfolio_table_name;
EXECUTE 'select * from ' || tablename || ' ;'
RETURN;
END;
$$ LANGUAGE plpgsql;
CALL easybuy.stproced();
Above code is executing fine but it is not printing the records which should be come from EXECUTE statement.

Answer 1

This cannot be done in plain SQL like your example. This is because the query is compiled and the table target needs to known a compilation.
This leads to needing 2 SQL statements to perform this action. As you mention a stored procedure can run a number of SQL statements so is a good idea. In many cases this will work except a stored procedure cannot produce data out the JDBC/ODBC connection (AFAIK). A stored procedure can fill a table with the results or fill a cursor but in both cases you will need to select or fetch to see these in your bench. Again you are back to needing 2 statements - executing the stored procedure and grabbing the results (select or fetch).
You could set up a wrapper around Redshift that takes some "special" command and maps it to the 2 SQL statement and otherwise just passes SQL through. This can work and there are available tools that work like this.
Some benches have the ability to configure macros that you could map to perform the 2 statements in question. This could be a route to look into.
If you explain the overarching problem you are trying to solve there may be other routes to achieve this goal.
==============================================================
Adding a stored procedure example that will perform the desired operation.
First let's set up some dummy tables:
create table test1 (tname varchar(16));
insert into test1 values
('test2'),
('b123'),
('c123');
create table test2 (UUID varchar(16), Key varchar(16), Value varchar(16));
insert into test2 values
('a123', 'Key1', 'Val1'),
('b123', 'Key2', 'Val2'),
('c123', 'Key3', 'Val3');
Next we create the stored procedure:
CREATE OR REPLACE procedure indirect_table(curs1 INOUT refcursor)
AS
$$
DECLARE
row record;
BEGIN
select into row max(tname) as tname from test1;
OPEN curs1 for EXECUTE 'select * from ' || row.tname || ';';
END;
$$ LANGUAGE plpgsql;
A quick explainer - this procedure first takes the name of a cursor as an argument and defines a record for storing the result of a query. Into this record is stored the max table name from table test1. This record is used to construct a query that uses this record's value in the FROM clause. This constructed query is run into a cursor where the results wait for a fetch request.
So the last step is to call the procedure and fetch the results. These are the only steps that will be needed in your script once the procedure is saved (committed).
call indirect_table('mycursor');
fetch all mycursor;
This will produce the desired output into the users bench. (Note that "fetch all" is not supported on a single node cluster and "fetch 1000" will be needed in such a case.)