合并来自多个 CTE 的数据 PostgreSQL

Question

I have the following query:

CREATE TYPE udt_insert_count AS
(
    insert_proxy_count int,
    insert_city_count int,
    insert_isp_count int,
    update_proxy_count int
);

CREATE OR REPLACE FUNCTION fn_insert_proxies(isps json, cities json, proxies json)
RETURNS udt_insert_count AS $func$
DECLARE result_count udt_insert_count;
BEGIN
    WITH isp_count AS
    (
        INSERT INTO isp (id, name)
        SELECT * FROM json_populate_recordset(null::udt_isp, isps) as udt_isps
        ON CONFLICT (id)
        DO NOTHING RETURNING xmax
    ),
    city_count AS
    (
        INSERT INTO city (proxy_address, latitude, longitude, name, sub_division1, sub_division1_code, sub_division2, sub_division2_code, postal_code, accuracy_radius, timezone)
        SELECT * FROM json_populate_recordset(null::udt_city, cities)
        ON CONFLICT (proxy_address)
        DO NOTHING RETURNING xmax
    ),
    proxy_count AS
    (
        INSERT INTO proxy as p (address, port, country_code, type_id, access_type_id, provider_id, isp_id, speed, uptime, created_date, modified_date)
        SELECT *, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP FROM json_populate_recordset(null::udt_proxy, proxies)
        ON CONFLICT ON CONSTRAINT proxy_address_port_uc
        DO UPDATE SET speed = p.speed, uptime = p.uptime, modified_date = CURRENT_TIMESTAMP RETURNING xmax
    )
    
    SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)
    INTO result_count.insert_city_count
    FROM city_count;
    
    SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END), SUM(CASE WHEN xmax::text::int > 0 THEN 1 ELSE 0 END)
    INTO result_count.insert_proxy_count, result_count.update_proxy_count
    FROM proxy_count;
    
    SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)
    INTO result_count.insert_isp_count
    FROM isp_count;
    
    RETURN result_count;
END
$func$  LANGUAGE plpgsql;

Basically what I want to do is to gather the insert and update counts into three different CTEs isp_count, city_count and proxy_count and then I would like to select them into the declared type.

However when I execute this I get: relation "proxy_count" does not exist

And I am sure that this is not the way to do it.

Is there any way I could do this in 1 go? :

 SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)
 INTO result_count.insert_city_count
 FROM city_count;

 SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END), SUM(CASE WHEN xmax::text::int > 0 THEN 1 ELSE 0 END)
 INTO result_count.insert_proxy_count, result_count.update_proxy_count
 FROM proxy_count;

 SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)
 INTO result_count.insert_isp_count
 FROM isp_count;

Answer 1

Is there any reason not to just decouple the CTEs? You don't use them in the same SELECT statements.

CREATE OR REPLACE FUNCTION fn_insert_proxies(isps json, cities json, proxies json)
RETURNS udt_insert_count AS $func$
DECLARE result_count udt_insert_count;
BEGIN
    WITH city_count AS
    (
        INSERT INTO city (proxy_address, latitude, longitude, name, sub_division1, sub_division1_code, sub_division2, sub_division2_code, postal_code, accuracy_radius, timezone)
        SELECT * FROM json_populate_recordset(null::udt_city, cities)
        ON CONFLICT (proxy_address)
        DO NOTHING RETURNING xmax
    )
    SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)
    INTO result_count.insert_city_count
    FROM city_count;

    WITH proxy_count AS
    (
        INSERT INTO proxy as p (address, port, country_code, type_id, access_type_id, provider_id, isp_id, speed, uptime, created_date, modified_date)
        SELECT *, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP FROM json_populate_recordset(null::udt_proxy, proxies)
        ON CONFLICT ON CONSTRAINT proxy_address_port_uc
        DO UPDATE SET speed = p.speed, uptime = p.uptime, modified_date = CURRENT_TIMESTAMP RETURNING xmax
    )
    SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END),
           SUM(CASE WHEN xmax::text::int > 0 THEN 1 ELSE 0 END)
    INTO result_count.insert_proxy_count, result_count.update_proxy_count
    FROM proxy_count;

    WITH isp_count AS
    (
        INSERT INTO isp (id, name)
        SELECT * FROM json_populate_recordset(null::udt_isp, isps) as udt_isps
        ON CONFLICT (id)
        DO NOTHING RETURNING xmax
    )
    SELECT SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)
    INTO result_count.insert_isp_count
    FROM isp_count;

    RETURN result_count;
END
$func$  LANGUAGE plpgsql;

As an aside, I recommend using COUNT(*) FILTER (WHERE xmax = 0) instead of SUM(CASE WHEN xmax = 0 THEN 1 ELSE 0 END)

Answer 2

I found your question interesting, so I took a look in the CTE documentation , where it has an interesting line:

The sub-statements in WITH are executed concurrently with each other and with the main query. Therefore, when using data-modifying statements in WITH, the order in which the specified updates actually happen is unpredictable. All the statements are executed with the same snapshot (see Chapter 13), so they cannot “see” one another's effects on the target tables.

Unfortunately don't have access to PostgreSQL right now, but will update later if I can play with it.