Need to shorten the SQL Query - snowflake-cloud-data-platform

Question

Query #1:
SELECT
id,
SUM(spent_time) AS spent_time_01,
COUNT(vpno) AS spend_time_cnt_01
FROM
(SELECT
LEAD(create_ts) OVER (PARTITION BY id ORDER BY CAST(vpno AS INT)) AS lead,
DATEDIFF('second', create_ts::timestamp, LEAD::timestamp) AS spent_time,
*
FROM
table_1) A
WHERE
item_category = 'A'
AND pid IS NOT NULL
GROUP BY
id
Query #2:
SELECT
id,
SUM(spent_time) AS spent_time_02,
COUNT(vpno) AS spend_time_cnt_02
FROM
(SELECT
LEAD(create_ts) OVER (PARTITION BY id ORDER BY CAST(vpno AS INT)) AS lead,
DATEDIFF('second', create_ts::timestamp, LEAD::timestamp) AS spent_time,
*
FROM
table_1) A
WHERE
item_category = 'B'
AND pid IS NOT NULL
GROUP BY
id
I tried the below but I'm not getting the correct results. I think I'm missing something in the query.
SUM(CASE WHEN pid IS NOT NULL
AND vpno IS NOT NULL
AND item_category = 'A' THEN spent_time END) AS spent_time_01
,COUNT_IF(vpno IS NOT NULL
AND item_category = 'A'
AND pid IS NOT NULL) AS spend_time_cnt
Someone kindly guide me on this.

Answer 1

Common part to be turn into CTE and item_category moved as conditional aggregation:
WITH cte AS (
SELECT *,
LEAD(create_ts) OVER (PARTITION BY id ORDER BY vpno::INT) AS lead,
DATEDIFF('second', create_ts::timestamp, LEAD::timestamp) AS spent_time
FROM table_1
)
SELECT id,
SUM(CASE WHEN item_category = 'A' THEN spent_time END) AS spent_time_01,
COUNT(CASE WHEN item_category = 'A' THEN vpno END) AS spend_time_cnt_01,
SUM(CASE WHEN item_category = 'B' THEN spent_time END) AS spent_time_02,
COUNT(CASE WHEN item_category = 'B' THEN vpno END) AS spend_time_cnt_02
FROM cte
WHERE pid IS NOT NULL
AND item_category IN ('A', 'B')
GROUP BY id;

Answer 2

Well if we are trying for Code Golf (smaller code) Lukasz answer can be rewritten a couple line less, and smaller by moving the lead value into the only place it is used, and not have a CTE as it's only used once, and just have it as a subselect. Moving the WHERE filters into the select, and swapping from CASE to IFF. But these are points I have made before..
Also by only asking for the columns you want in the sub-select, the compile can be faster on cold meta data. but those are details that don't matter here.
SELECT id,
SUM(iff(item_category = 'A', spent_time, null)) AS spent_time_01,
COUNT(iff(item_category = 'A', vpno, null)) AS spend_time_cnt_01,
SUM(iff(item_category = 'B', spent_time, null)) AS spent_time_02,
COUNT(iff(item_category = 'B', vpno, null)) AS spend_time_cnt_02
FROM (
SELECT id, item_category,
DATEDIFF('second', create_ts::timestamp, (LEAD(create_ts) OVER (PARTITION BY id ORDER BY vpno::INT))::timestamp) AS spent_time
FROM table_1
WHERE pid IS NOT NULL
AND item_category IN ('A', 'B')
)
GROUP BY id;
But I do tend to prefer CTE for readability and I don't like over long lines, so even I will give my points vote to Lukasz's answer.