How to pull number of unique items sold in oracle sql - sql

Question

I have orders data for all items sold in my store for the past year. The table looks like this:
order_date item_id item_name order_quantity unit_price
--------------------------------------------------------------------
01-01-2017 a123 a234 2 10
04-02-2017 b123 b234 3 12
04-09-2017 c123 c234 1 15
04-10-2017 b123 b234 2 12
I need to pull data for number of unique items sold by week, month, and quarter. The query output should look like this:
timeline number unique_item_count
week 1 1
week 14 1
week 15 2
month 1 1
month 4 2
quarter 2 2
I've tried the following:
SELECT
TO_CHAR(c.ORDER_DAY, 'Q') AS QTR,
TO_CHAR(c.ORDER_DAY, 'MM') AS MNTH,
TO_CHAR(c.ORDER_DAY, 'WW') AS WK,
COUNT(DISTINCT CASE WHEN (c.QUANTITY*c.OUR_PRICE) > 0 THEN ITEM_ID ELSE NULL END) AS SALES_CNT
FROM TABLE c
GROUP BY TO_CHAR(c.ORDER_DAY, 'Q'), TO_CHAR(c.ORDER_DAY, 'MM'), TO_CHAR(c.ORDER_DAY, 'WW');
This works for weekly data, however, monthly and quarterly is just a sum of weekly numbers, which is incorrect in this case, since same items might be ordered in two different weeks, so the monthly number should be lower.
Is there a way to pull number of unique items purchased in each week, month, quarter?
Thanks!

Answer 1

You can do it without using the union of multiple statements by using UNPIVOT:
SQL Fiddle
Oracle 11g R2 Schema Setup:
CREATE TABLE table_c (
order_date,
item_id,
item_name,
order_quantity,
order_price
) AS
SELECT DATE '2017-09-01' + LEVEL * 3 - 3, 1, 'aaa', 1, 1 FROM DUAL CONNECT BY LEVEL <= 10
UNION ALL
SELECT DATE '2017-09-01' + LEVEL * 4 - 4, 2, 'bbb', 1, 1 FROM DUAL CONNECT BY LEVEL <= 10
UNION ALL
SELECT DATE '2017-09-01' + LEVEL * 5 - 5, 3, 'ccc', 1, 1 FROM DUAL CONNECT BY LEVEL <= 10
Query 1:
SELECT year,
timeline,
"number",
COUNT( DISTINCT item_id )
FROM (
SELECT TO_CHAR( order_date, 'WW' ) AS week,
TO_CHAR( order_date, 'MM' ) AS month,
TO_CHAR( order_date, 'Q' ) AS quarter,
EXTRACT( YEAR from order_date ) AS year,
item_id
FROM table_c
WHERE order_quantity > 0
AND order_price > 0
)
UNPIVOT ( "number" FOR timeline IN ( week AS 'week', month AS 'month', quarter AS 'quarter' ) )
GROUP BY year, timeline, "number"
ORDER BY year, timeline, "number"
Results:
| YEAR | TIMELINE | number | COUNT(DISTINCTITEM_ID) |
|------|----------|--------|------------------------|
| 2017 | month | 09 | 3 |
| 2017 | month | 10 | 2 |
| 2017 | quarter | 3 | 3 |
| 2017 | quarter | 4 | 2 |
| 2017 | week | 35 | 3 |
| 2017 | week | 36 | 3 |
| 2017 | week | 37 | 3 |
| 2017 | week | 38 | 3 |
| 2017 | week | 39 | 3 |
| 2017 | week | 40 | 2 |
| 2017 | week | 41 | 1 |
| 2017 | week | 42 | 1 |

Answer 2

You can do this with union all:
SELECT 'Q' as timeline, TO_CHAR(c.ORDER_DAY, 'Q') AS number,
COUNT(DISTINCT CASE WHEN (c.QUANTITY*c.OUR_PRICE) > 0 THEN ITEM_ID END) AS SALES_CNT
FROM TABLE c
GROUP BY TO_CHAR(c.ORDER_DAY, 'Q')
UNION ALL
SELECT 'MM' as timeline, TO_CHAR(c.ORDER_DAY, 'MM') AS number,
COUNT(DISTINCT CASE WHEN (c.QUANTITY*c.OUR_PRICE) > 0 THEN ITEM_ID END) AS SALES_CNT
FROM TABLE c
GROUP BY TO_CHAR(c.ORDER_DAY, 'MM')
UNION ALL
SELECT 'WW' as timeline, TO_CHAR(c.ORDER_DAY, 'WW') AS number,
COUNT(DISTINCT CASE WHEN (c.QUANTITY*c.OUR_PRICE) > 0 THEN ITEM_ID END) AS SALES_CNT
FROM TABLE c
GROUP BY TO_CHAR(c.ORDER_DAY, 'WW');

Answer 3

You can use GROUPING SETS to aggregate at various levels in a single query. Like so:
SELECT CASE
WHEN GROUPING (TO_CHAR (c.order_day, 'Q')) = 0 THEN 'Quarter'
WHEN GROUPING (TO_CHAR (c.order_day, 'MM')) = 0 THEN 'Month'
WHEN GROUPING (TO_CHAR (c.order_day, 'WW')) = 0 THEN 'Week'
ELSE '??'
END
timeline,
CASE
WHEN GROUPING (TO_CHAR (c.order_day, 'Q')) = 0 THEN TO_CHAR (c.order_day, 'Q')
WHEN GROUPING (TO_CHAR (c.order_day, 'MM')) = 0 THEN TO_CHAR (c.order_day, 'MM')
WHEN GROUPING (TO_CHAR (c.order_day, 'WW')) = 0 THEN TO_CHAR (c.order_day, 'WW')
ELSE '??'
END
"NUMBER",
COUNT (DISTINCT CASE WHEN (c.quantity * c.our_price) > 0 THEN item_id ELSE NULL END) unique_item_count
FROM c
GROUP BY GROUPING SETS (
(TO_CHAR (c.order_day, 'Q')),
(TO_CHAR (c.order_day, 'MM')),
(TO_CHAR (c.order_day, 'WW')))