Postgres function: return multiple tables

问题

Is it possible to return several result sets of different types from postgres function?

Something like:

CREATE OR REPLACE FUNCTION getUserById()
RETURNS setof ???
AS
$$
BEGIN

return query select id, name /* and other columns */ from users where id = 1;
return query select id, phone_number from user_phones where user_id = 1

END
$$
LANGUAGE plpgsql;

I don't want to use joins because several phones for user are possible. Also it would be great to avoid using cursors. It's possible in MS SQL and I want to do the same thing in postgres.

回答1:

After the seek, could not find any better solutions as use of cursors. I assume you have seen this one already: https://blog.dsl-platform.com/multiple-result-sets-alternatives-in-postgres-3/

CREATE FUNCTION load_page(_session INT) RETURNS setof refcursor AS
$$
DECLARE c_top_items refcursor;
DECLARE c_shopping_cart refcursor;
BEGIN
    OPEN c_top_items FOR
        SELECT t.name, t.description
        FROM top_item t
        ORDER BY t.popularity DESC
        LIMIT 10;
    RETURN NEXT c_top_items;
    OPEN c_shopping_cart FOR
        SELECT c.product_id, c.product_name, c.quantity
        FROM shopping_cart c
        WHERE c.session_id = _session
        ORDER BY c.id;
    RETURN NEXT c_shopping_cart;
END;
$$ LANGUAGE plpgsql;

And calling:

BEGIN;
SELECT load_page(mySession);
FETCH ALL IN "<server cursor 1>";
FETCH ALL IN "<server cursor 2>";
COMMIT;

回答2:

I don't want to use joins because several phones for user are possible.

This is not a reason to avoid JOINs in PostgreSQL. At all.

PostgreSQL allows you to aggregate the phone numbers into an array:

CREATE OR REPLACE FUNCTION getUserById()
RETURNS TABLE (
    id INTEGER,
    name TEXT,
    /* and other columns */
    phone_numbers TEXT[]
)
AS
$$
    select
         users.id,
         users.name,
         /* and other columns */
         -- Remove NULL because you get an array containing just NULL
         -- if user_phones doesn't contain any matching rows.
         array_remove(array_agg(user_phones.phone_number), NULL) as phone_numbers
    from users
    left join user_phones on user_phones.user_id = users.id
    where users.id = 1
    -- Note that grouping by a table's primary key allows you to use
    -- any column from that table in the select in PostgreSQL
    group by users.id
    ;
$$
LANGUAGE SQL
STABLE
;

This is much simpler and more intuitive.

You can switch to an inner join if it's okay to give back zero rows for a user without a phone number. In that case, you could drop the array_remove call.

I also added the STABLE specification to the function (since it doesn't modify any table data) and switched it to SQL instead of PGPLSQL (since it's just a single query). This will allow PG to optimize better; in particular, it can inline the query and push filters down in some cases. You may not even need a function, actually.

来源：https://stackoverflow.com/questions/36717138/postgres-function-return-multiple-tables

标签

postgresql

plpgsql

postgresql-9.3