hep with CTE in Postgresql - join tables-CodePudding

i have following query.

with sh as
(select payor, name_policy, program, count(distinct event_id) as sevents from cart_item_funder_policy_worker where event_Status like 'Show%' 
    and event_time between '2022-04-01' and '2022-09-30' group by payor, name_policy, program),
qty_s as
(select payor, name_policy, program, sum(item_qty) as qty_show from cart_item_funder_policy_worker where event_Status like 'Show%' 
    and event_time between '2022-04-01' and '2022-09-30' group by payor, name_policy, program),
ns as
(select payor, name_policy, program, count(distinct event_id) as nsevents from cart_item_funder_policy_worker where event_Status like 'No Show%' 
    and event_time between '2022-04-01' and '2022-09-30' group by payor, name_policy, program),
qty_ns as
(select payor, name_policy, program, sum(item_qty) as qty_no_show from cart_item_funder_policy_worker where event_Status like 'No Show%' 
    and event_time between '2022-04-01' and '2022-09-30' group by payor, name_policy, program)
    
select sh.payor, sh.name_policy, sh.program, sh.sevents, qty_s.qty_show, ns.nsevents, qty_ns.qty_no_show
from sh
full join ns on sh.name_policy = ns.name_policy
full join qty_s on qty_s.name_policy = sh.name_policy
full join qty_ns on qty_ns.name_policy = ns.name_policy;

as you can see that i only want to see the (count of event_id and sum of item_qty) grouped by payor, name_policy, program. but the above query groups the result by sh.sevents, qty_s.qty_show, ns.nsevents, qty_ns.qty_no_show as well.

can you please provide any assistance in figuring out how i can avoid grouping of sh.sevents, qty_s.qty_show, ns.nsevents, qty_ns.qty_no_show in my query.

Table definition:

CREATE TABLE cart_item_funder_policy_worker
( payor          VARCHAR
, name_policy    VARCHAR
, program        VARCHAR
, item_qty       INTEGER
, event_id       INTEGER
, event_date     DATE
, event_status   VARCHAR
);

thanks

CodePudding user response：

It looks like you're nearly there and you can simplify your CTEs to perform both the COUNT and SUM aggregates over the same data.

There doesn't appear to be a need for a FULL JOIN here. When will you have data in sh or ns that doesn't exist in cart_item_funder_policy_worker? LEFT JOIN will produce a row in the result set where you have data in cart_item_funder_policy_worker but not in sh or ns. Or just INNER JOIN/JOIN if you don't care about joining against NULL.

I didn't test it but I believe you can get the aggregates down to one block with something like this:

WITH data AS
(
    SELECT payor
         , name_policy
         , program
         , item_qty
         , event_id
         , event_status
      FROM cart_item_funder_policy_worker
     WHERE event_time BETWEEN '2022-04-01' AND '2022-09-30'
)
, aggs AS
(
    SELECT d.name_policy
         , COUNT(DISTINCT d.event_id) FILTER (WHERE d.event_status LIKE 'Show%') AS sevents
         , COUNT(DISTINCT d.event_id) FILTER (WHERE d.event_status LIKE 'No Show%') AS nsevents
         , SUM(d.item_qty) FILTER (WHERE d.event_status LIKE 'Show%') AS qty_show
         , SUM(d.item_qty) FILTER (WHERE d.event_status LIKE 'No Show%') AS qty_no_show
      FROM data d
     GROUP BY d.payor, d.name_policy, d.program
)
SELECT DISTINCT 
       d.payor
     , d.name_policy
     , d.program
     , a.sevents
     , a.qty_show
     , a.nsevents
     , a.qty_no_show
  FROM data d
  LEFT JOIN aggs a ON d.name_policy = a.name_policy
;

Or something like this:

SELECT all of the data you're going to be aggregating over in the first CTE, then calculate your aggregates in a separate CTE. In your aggregate portions, SELECT only your JOIN fields and the aggregates. Finally, grab your desired result set.

WITH data AS
(
    SELECT payor
         , name_policy
         , program
         , item_qty
         , event_id
         , event_status
      FROM cart_item_funder_policy_worker
     WHERE event_time BETWEEN '2022-04-01' AND '2022-09-30'
)
, sh AS
(
    SELECT d.name_policy
         , COUNT(DISTINCT d.event_id) AS sevents
         , SUM(d.item_qty) AS qty_show
      FROM data d
     WHERE d.event_status LIKE 'Show%'
     GROUP BY d.payor, d.name_policy, d.program
)
, ns AS
(
    SELECT d.name_policy
         , COUNT(DISTINCT d.event_id) AS nsevents
         , SUM(d.item_qty) AS qty_no_show
      FROM data d
     WHERE d.event_status LIKE 'No Show%'
     GROUP BY d.payor, d.name_policy, d.program
)
SELECT DISTINCT 
       d.payor
     , d.name_policy
     , d.program
     , sh.sevents
     , sh.qty_show
     , ns.nsevents
     , ns.qty_no_show
  FROM data d
  LEFT JOIN sh ON d.name_policy = sh.name_policy
  LEFT JOIN ns ON d.name_policy = ns.name_policy
;