SQL Segmentation - sql

I have this table
Customer
Amount
Date
Period
Group
77766
50
2022-02-28
1
2
77766
20
2022-03-31
2
2
77766
30
2022-04-30
3
3
12345
50
2022-02-28
1
1
12345
20
2022-03-31
2
2
12345
30
2022-04-30
3
3
and I'm trying to achieve this segmentation model where I assign a Group to its corresponding Period per Customer
Customer
Amount
Date
Period
Group
Group_Period1
Group_Period2
Group_Period3
77766
50
2022-02-28
1
2
2
2
3
77766
20
2022-03-31
2
2
2
2
3
77766
30
2022-04-30
3
3
2
2
3
12345
50
2022-02-28
1
1
1
2
3
12345
20
2022-03-31
2
2
1
2
3
12345
30
2022-04-30
3
3
1
2
3
I tried case function but it didn't work
select a.*,
case when a.Period=1 then Group end Grupo1 ,
case when a.Period=2 then Group2 end Grupo2 ,
case when a.Period=3 then Group3
end Grupo3
from a
I got this output:
Customer
Amount
Date
Period
Group
Group_Period1
Group_Period2
Group_Period3
77766
50
2022-02-28
1
2
2
null
null
77766
20
2022-03-31
2
2
null
2
null
77766
30
2022-04-30
3
3
null
null
3
12345
50
2022-02-28
1
1
1
null
null
12345
20
2022-03-31
2
2
null
2
null
12345
30
2022-04-30
3
3
null
null
3
Can anybody guide me to achieve the expected? Thank you in advice

You can use window function such as MAX() OVER () with conditionals for values of Period column such as
SELECT *,
MAX(CASE WHEN Period=1 THEN `Group` END)
OVER (PARTITION BY Customer) AS Group_Period1,
MAX(CASE WHEN Period=2 THEN `Group` END)
OVER (PARTITION BY Customer) AS Group_Period2,
MAX(CASE WHEN Period=3 THEN `Group` END)
OVER (PARTITION BY Customer) AS Group_Period3
FROM t -- replace with your table's name
where grouping is figured out through use of PARTITION BY clause
Demo

One approach can be via inline sub-query.
select *,
(select group_1 from cust a where a.customer = b.customer and period=1) Group_Period1,
(select group_1 from cust a where a.customer = b.customer and period=2) Group_Period2,
(select group_1 from cust a where a.customer = b.customer and period=3) Group_Period3
from cust b;
Fiddle here

Related

Oracle 11g count distinct products group by user

I have a table with following fields
ID
PRODUCT
STATUS
USERID
Date
1
100
1
10
01-01-2023
2
101
1
10
01-01-2023
2
102
2
10
01-01-2023
3
100
2
20
02-01-2023
4
102
1
30
02-01-2023
4
100
1
10
03-01-2023
Desired output
Distinct product scan by each userid between 01-01-2023 to 03-01-2023 and count by status for each user
USERID
PRODUCT SCAN
STATUS1 CNT
STATUS2 CNT
10
3
2
1
20
1
0
1
30
1
1
0
Use COUNT(DISTINCT column_name).
Query
select userid, count(distinct product) as productScan
from table_name
where date between '2023-01-01' and '2023-01-03'
group by userid;

Group by one column and return several columns on multiple conditions - T-SQL

I have two tables which I can generate with SELECT statements (joining multiple tables) as follows:
Table 1:
ID
Site
type
time
1
Dallas
2
01-01-2021
2
Denver
1
02-01-2021
3
Chicago
1
03-01-2021
4
Chicago
2
29-11-2020
5
Denver
1
28-02-2020
6
Toronto
2
11-05-2019
Table 2:
ID
Site
collected
deposited
1
Denver
NULL
29-01-2021
2
Denver
01-04-2021
29-01-2021
3
Chicago
NULL
19-01-2020
4
Dallas
NULL
29-01-2019
5
Winnipeg
13-02-2021
17-01-2021
6
Toronto
14-02-2020
29-01-2020
I would like the result to be grouped by Site, having on each column the COUNT of type=1 , type=2, deposited and collected, all of the 4 columns between a selected time interval. Example: (interval between 01-06-2020 and 01-06-2021:
Site
type1
type2
deposited
collected
Dallas
0
1
0
0
Denver
1
0
2
1
Chicago
1
1
0
0
Toronto
0
0
0
0
Winnipeg
0
0
1
1
How about union all and aggregation?
select site,
sum(case when type = 1 then 1 else 0 end) as type_1,
sum(case when type = 2 then 1 else 0 end) as type_2,
sum(deposited) as deposited, sum(collected) as collected
from ((select site, type, 0 as deposited, 0 as collected
from table1
) union all
(select site, null,
(case when deposited is not null then 1 else 0 end),
(case when collected is not null then 1 else 0 end)
from table2
)
) t12
group by site;
Combine your tables 1 and 2 with a join on Site
Use COUNT(CASE WHEN type = 1 then 1 END) as type1 and a similar construct for type 2
Use COUNT(CASE WHEN somedate BETWEEN '2020-06-01' and '2021-06-01' then 1 END) as ... for your dates

Get TOP 1 row from multiple tables based on date

I have five tables in a SQL Server database. These tables are listed below and I want to select Data from these tables according to date. I tried searching but could not find solution for multiple tables. Please help
TABLE1
id PId CId
----------- ---------- ------------
1 P001 1
2 P002 2
3 P003 4
4 P004 5
TABLE2
id CId CNo ConId
----------- ---------- ------------ ----------
1 1 1 C123
2 1 2 PA444
3 1 3 PA456
4 2 1 AUX2398
5 2 2 AUX2345
6 4 1 PA123
7 5 1 C234
TABLE3
id CId CNo Label Date
----------- ---------- ------------ ---------- ----------
1 1 1 A 1/1/2000
2 1 2 A 15/10/2020
3 1 3 A 20/10/2020
4 2 1 A 15/10/2020
5 2 2 A 20/10/2020
6 4 1 A 20/10/2020
7 5 1 A 27/10/2020
TABLE4
id CId CNo Label Date
----------- ---------- ------------ ---------- ----------
1 1 1 B 20/10/2020
2 1 2 B 27/10/2020
3 1 3 B 22/10/2020
4 2 1 B 22/10/2020
5 2 2 B 26/10/2020
6 4 1 B 22/10/2020
7 5 1 B 30/10/2020
TABLE5
id CId CNo Label Date
----------- ---------- ------------ ---------- ----------
1 1 1 C 26/10/2020
2 1 2 C 1/1/2000
3 1 3 C 23/10/2020
4 2 1 C 25/10/2020
5 2 2 C 30/10/2020
6 4 1 C 25/10/2020
7 5 1 C 1/1/2000
I want to select Label and Date from Table 3, 4 and 5 where Date is >1/1/2000 and < than and close to 24/10/2020 and grouped according to PId, CId, ConId and CNo.
Desired result:
PId CId ConId CNo Label Date
-------- ---------- ---------- -------- --------- ----------
P001 1 C123 1 B 20/10/2020
P001 1 PA444 2 A 15/10/2020
P001 1 PA456 3 C 23/10/2020
P002 2 AUX2398 1 B 22/10/2020
P002 2 AUX2345 2 A 20/10/2020
P003 4 PA123 1 B 22/10/2020
P004 5 C234 1 - -
Any help will be appreciated. Thank you.
You can achieve this with a couple of CTE's; the first forms a UNION of TABLE3, TABLE4 and TABLE5; the second generates a row number based on the Date descending for each partition of PId, CId, ConId and CNo. We then select all rows from the second CTE where the row number is 1:
WITH CTE AS (
SELECT * FROM Table3 WHERE date > '2000-01-01'
UNION ALL
SELECT * FROM Table4 WHERE date > '2000-01-01'
UNION ALL
SELECT * FROM Table5 WHERE date > '2000-01-01'
),
CTE2 AS (
SELECT t1.PId, t1.CId, t2.ConId, t2.CNo, CTE.Label, CTE.Date,
ROW_NUMBER() OVER (PARTITION BY t1.PId, t1.CId, t2.ConId, t2.CNo ORDER BY CTE.Date DESC) AS rn
FROM TABLE1 t1
JOIN TABLE2 t2 ON t2.CId = t1.CId
LEFT JOIN CTE ON CTE.Cid = t2.CId AND CTE.CNo = t2.CNo AND CTE.Date < '2020-10-24'
)
SELECT PId, CId, ConId, CNo, Label, Date
FROM CTE2
WHERE rn = 1
ORDER BY PId, CId, CNo
Output:
PId CId ConId CNo Label Date
P001 1 C123 1 B 2020-10-20
P001 1 PA444 2 A 2020-10-15
P001 1 PA456 3 C 2020-10-23
P002 2 AUX2398 1 B 2020-10-22
P002 2 AUX2345 2 A 2020-10-20
P003 4 PA123 1 B 2020-10-22
P004 5 C234 1 - -
Demo on dbfiddle

MS Access merge two tables

I need to create a new table base on two tables. I have a database in ms access and need to migrate.
tableT tableS
ID CustID DATE Exp1 Exp2 ID CustID DATE Tem1 Tem2
-------------------------------- ---------------------------------
1 1 1/1/00 5 5 1 1 1/1/00 3 4
2 2 1/1/00 1 3 2 2 1/1/00 5 0
3 1 3/1/00 3 2 3 1 5/1/00 0 3
4 3 4/1/00 4 1 4 3 6/1/00 0 0
Desired output table tableNew:
ID CustID DATE Exp1 Exp2 Tem1 Tem2
---------------------------------------------
1 1 1/1/00 5 5 3 4
2 2 1/1/00 1 3 5 0
3 1 3/1/00 3 2
4 3 4/1/00 4 1
5 1 5/1/00 0 3
6 3 6/1/00 0 0
If I use outer join, I will not get the output I need.
Any idea or help.
You want a full join. You can emulate this in MS Access using:
select t.CustID, t.DATE, t.Exp1, t.Exp2, s.tem1, s.tem2
from TableT as t left outer join
tableS as s
on t.CustId = s.CustId and t.date = s.date
union all
select s.CustID, s.DATE, null, null, s.tem1, s.tem2
from tableS as s left outer join
tableT as t
on t.CustId = s.CustId and t.date = s.date
where t.CustId is null;

Multiple times Grouping with where in SQL Server

I have a source table like this
ProductName SaleReceipt SaleCode
--------------------------------
F-Apple 1001 1
F-Orange 1002 2
G-Rice 1003 3
G-Barile 1005 4
G-Oats 1006 1
V-Carrot 1007 4
V-Cabbage 1008 3
V-Potato 1009 1
V-Tomato 1010 1
Chocolate 1011 4
Cookies 1012 1
Cakes 1013 2
I need to create a report like this
30 Day delay 60 Day Delay 90 day delay 120 day delay
Fruits 1 1 0 0
Grains 1 0 1 1
Vegetables 2 0 1 1
Other category 1 1 0 1
The conditions to create the report are:
All ProductName start with F is grouped as fruits
All ProductName start with G is grouped as Grains
All ProductName start with V is grouped as Vegetables
Other product name go in the other category
30 day delay: count of (SaleReceipt) with a SaleCode=1
60 day delay: count of (SaleReceipt) with a SaleCode=2
90 day delay: count of (SaleReceipt) with a SaleCode=3
120 day delay: count of (SaleReceipt) with a SaleCode>=4
I could not find how to do grouping two times. I am using SQL Server 2014.
You can use case in a group by. Or use a lookup table:
select coalesce(c.name, 'Other category'),
sum(case when salescode = 1 then 1 else 0 end) as salecode_1,
sum(case when salescode = 2 then 1 else 0 end) as salecode_2,
sum(case when salescode = 3 then 1 else 0 end) as salecode_3,
sum(case when salescode = 4 then 1 else 0 end) as salecode_4
from t left join
(values ('F-%', 'Fruit'), ('G-%' , 'Grain'), ('V-%', 'Vegetable')
) c(pattern, name)
on t.productname like c.pattern
group by coalesce(c.name, 'Other category');