Is there a safe way to not have to group by a field when using an aggregate in another field? Here is my example
SELECT
C.CustomerName
,D.INDUSTRY_CODE
,CASE WHEN D.INDUSTRY_CODE IN ('003','004','005','006','007','008','009','010','017','029')
THEN 'PM'
WHEN UPPER(CustomerName) = 'ULINE INC'
THEN 'ULINE'
ELSE 'DR'
END AS BU
,ISNULL((SELECT SUM(GrossAmount)
where CONVERT(date,convert(char(8),InvoiceDateID )) between DATEADD(yy, DATEDIFF(yy, 0, GETDATE()) - 1, 0) and DATEADD(year, -1, GETDATE())),0) [PREVIOUS YEAR GROSS]
FROM factMargins A
LEFT OUTER JOIN dimDate B ON A.InvoiceDateID = B.DateId
LEFT OUTER JOIN dimCustomer C ON A.CustomerID = C.CustomerId
LEFT OUTER JOIN CRCDATA.DBO.CU10 D ON D.CUST_NUMB = C.CustomerNumber
GROUP BY
C.CustomerName,D.INDUSTRY_CODE
,A.InvoiceDateID
order by CustomerName
before grouping I was only getting 984 rows but after grouping by the A.InvoiceDateId field I am getting over 11k rows. The rows blow up since there are multiple invoices per customer. Min and Max wont work since then it will pull data incorrectly. Would it be best to let my application (crystal) get rid of the extra lines? Usually I like to have my base data be as close as possible to how the report will layout if possible.
Try moving the reference to InvoiceDateID to within an aggregate function, rather than within a selected subquery's WHERE clause.
In Oracle, here's an example:
with TheData as (
select 'A' customerID, 25 AMOUNT , trunc(sysdate) THEDATE from dual union
select 'B' customerID, 35 AMOUNT , trunc(sysdate-1) THEDATE from dual union
select 'A' customerID, 45 AMOUNT , trunc(sysdate-2) THEDATE from dual union
select 'A' customerID, 11000 AMOUNT , trunc(sysdate-3) THEDATE from dual union
select 'B' customerID, 12000 AMOUNT , trunc(sysdate-4) THEDATE from dual union
select 'A' customerID, 15000 AMOUNT , trunc(sysdate-5) THEDATE from dual)
select
CustomerID,
sum(amount) as "AllRevenue"
sum(case when thedate<sysdate-3 then amount else 0 end) as "OlderRevenue",
from thedata
group by customerID;
Output:
CustomerID | AllRevenue | OlderRevenue
A | 26070 | 26000
B | 12035 | 12000
This says:
For each customerID
I want the sum of all amounts
and I want the sum of amounts earlier than 3 days ago
Related
What I'm looking to do is quantify the total value of purchases and the number of months in which a purchase was made within three different timeframes by account. I only want to look at accounts who made a purchase between 1-1-2020 and 4-1-2021.
I'm wondering if there is a more streamlined way to pull in the fields I'm creating using CASE WHEN below (maybe through a series of queries to create the calculations and the left joining?). This query is taking extremely long to pull back, so I'd like to enhance this code where I can. All of my code and desired output is listed below. Thank you!
Creating a temporary table to pull account numbers:
DROP TABLE IF EXISTS #accounts
SELECT DISTINCT s.account_no, c.code, c.code_desc
INTO #accounts
FROM sales AS s
LEFT JOIN customer AS c ON s.account_no = c.account_no
WHERE s.tran_date BETWEEN '2020-01-01' AND '2021-04-01'
GROUP BY s.account_no, c.code, c.code_desc;
Confirming row counts:
SELECT COUNT (*)
FROM #accounts
ORDER BY account_no;
Creating Sales and Sales period count columns for three timeframes:
SELECT
s.account_no, c.code, c.code_desc
SUM(CASE
WHEN s.tran_date BETWEEN '2020-01-01' AND '2021-04-01'
THEN VALUE_USD
END) AS Total_Spend_Pre,
SUM(CASE
WHEN s.tran_date BETWEEN '2021-04-01' AND '2022-03-31'
THEN VALUE_USD
END) Total_Spend_During,
SUM(CASE
WHEN s.tran_date > '2022-04-01'
THEN VALUE_USD
END) Total_Spend_Post,
COUNT(DISTINCT CASE WHEN s.tran_date BETWEEN '2020-01-01' AND '2021-04-01' THEN CONCAT(s.bk_month, s.bk_year) END) Pre_Periods,
COUNT(DISTINCT CASE WHEN s.tran_date BETWEEN '2021-04-01' AND '2022-03-31' THEN CONCAT(s.bk_month, s.bk_year) END) During_Periods,
COUNT(DISTINCT CASE WHEN s.tran_date > '2022-04-01' THEN CONCAT(s.bk_month, s.bk_year) END) Post_Periods
FROM
sales AS s
LEFT JOIN
customer AS c ON s.account_no = c.account_no
WHERE
c.account_no IN (SELECT DISTINCT account_no
FROM #accounts)
GROUP BY
s.account_no, c.code, c.code_desc;
Desired output:
account_no
code
code_desc
Total_Spend_Pre
Total_Spend_During
Total_Spend_Post
Pre_Periods
During_Periods
Post_Periods
25
1234
OTHER
1000
2005
500
2
14
5
11
5678
PC
500
100
2220
5
11
2
You may use your date ranges to join with dataset, and 'Tag' your result like below, this will result in 3 rows, for each group. If you need them in a single row, have PIVOTE over it
;With DateRanges AS (
SELECT CAST('2020-01-01' AS DATE) StartDate, CAST('2021-04-01' AS DATE) EndDate, 'Pre' Tag UNION
SELECT '2021-04-01', '2022-03-31', 'During' UNION
SELECT '2022-04-01', Null, 'Post'
)
SELECT s.account_no, c.code, c.code_desc, d.Tag,
SUM(VALUE_USD) AS Total_Spend,
COUNT(DISTINCT CONCAT(s.bk_month, s.bk_year)) RecordCount
FROM sales as s
LEFT JOIN customer as c
INNER JOIN DateRanges D ON s.tran_date BETWEEN D.StartDate AND ISNULL(D.EndDate,s.tran_date)
ON s.account_no = c.account_no
WHERE c.account_no IN (SELECT DISTINCT account_no FROM #accounts)
GROUP BY s.account_no, c.code, c.code_desc;
with [cte_accountActivityPeriods] as (
select [PeriodOrdinal] = 1, [PeriodName] = 'Total Spend Pre', [PeriodStart] = convert(date,'2020-01-01',23) , [PeriodFinish] = convert(date,'2021-03-31',23) union
select [PeriodOrdinal] = 2, [PeriodName] = 'Total Spend During', [PeriodStart] = convert(date,'2021-04-01',23) , [PeriodFinish] = convert(date,'2022-03-31',23) union
select [PeriodOrdinal] = 3, [PeriodName] = 'Total Spend Post', [PeriodStart] = convert(date,'2022-04-01',23) , [PeriodFinish] = convert(date,'9999-12-31',23)
)
, [cte_allsalesForActivityPeriod]
SELECT s.account_no, bk_month, bk_year, [PeriodOrdinal], s.tran_date, s.value_usd
FROM sales as s
cross join [cte_accountActivityPeriods]
on s.[tran_date] between [cte_ActivityPeriods].[PeriodStart] and [cte_ActivityPeriods].[PeriodFinish]
)
, [cte_uniqueAccounts] as ( /*Unique and qualifying Accounts*/
select distinct account_no from [cte_allsalesForActivityPeriod]
inner join #accounts accs on accs.[account_no] = [cte_allsalesForActivityPeriod].[account_no]
)
, [cte_AllSalesAggregatedByPeriod] as (
select account_no, [PeriodOrdinal], bk_month, bk_year, [PeriodTotalSpend] = sum([value_usd])
from [cte_allsalesForActivityPeriod]
group by s.account_no, [PeriodOrdinal], bk_month, bk_year
)
, [cte_PeriodAnalysis] as (
select account_no, [PeriodOrdinal], [ActivePeriods] = count(distinct concat(bk_month, bk_year))
from [cte_AllSalesAggregatedByPeriod]
group by s.account_no, [PeriodOrdinal]
)
, [cte_pivot_clumsily] as (
/* Aggregations already done - so simple pivot */
select [cte_uniqueAccounts].[account_no]
, [Total_Spend_Pre] = case when [SaleVal].[PeriodOrdinal] in (1) then [SaleVal].[PeriodTotalSpend] else 0 end
, [Total_Spend_During] = case when [SaleVal].[PeriodOrdinal] in (2) then [SaleVal].[PeriodTotalSpend] else 0 end
, [Total_Spend_Post] = case when [SaleVal].[PeriodOrdinal] in (3) then [SaleVal].[PeriodTotalSpend] else 0 end
, [Pre_Periods] = case when [SalePrd].[PeriodOrdinal] in (1) then [SalePrd].[ActivePeriods] else 0 end
, [During_Periods] = case when [SalePrd].[PeriodOrdinal] in (2) then [SalePrd].[ActivePeriods] else 0 end
, [Post_Periods] = case when [SalePrd].[PeriodOrdinal] in (3) then [SalePrd].[ActivePeriods] else 0 end
from [cte_uniqueAccounts]
left join [cte_AllSalesAggregatedByPeriod] [SaleVal] on [SaleVal].[account_no] = [cte_uniqueAccounts].[account_no]
left join [cte_PeriodAnalysis] [SalePrd] on [SalePrd].[account_no] = [cte_uniqueAccounts].[account_no]
)
select c.code, c.code_desc, [cte_pivot_clumsily].*
from [cte_pivot_clumsily]
LEFT JOIN customer as c
ON [cte_pivot_clumsily].account_no = c.account_no
I have two tables 1. Transactions and 2. Transaction type
e.g. of transaction table with dummy data
account_key Transaction_key date amount
1 11 03/22/0219 5000
1 12 03/23/2019 6000
1 13 03/22/2019 4000
1 14 03/23/2019 3000
e.g. of Transaction_type table with dummy data
transaction_key transaction_type
11 in
12 in
13 out
14 out
I have to find the ratio of sum of amount for 2 consecutive days of similar transaction type for the same account key. for eg (5000+6000)/(4000+3000)
the database is oracle and datatype is datetime
This is what I have tried
select t1.account_key,
trunc(t1.date),
sum(t1.amount) as in,
sum(t2.amount) as out
from transactions t1
inner join transactions t2 on t1.accountkey=t2.accountkey
where t1.date between '01-jun-2017' and '30-nov-2017'
and t2.date between '01-jun-2017' and '30-nov-2017'
and t1.transaction_key in (select a.transaction_key
from transaction_type a
where a.transaction type in 'in')
and t2.transaction_key in (select b.transaction_key
from transaction_type b
where b.transaction type in 'out')
group by t1.account_key,
trunc(t1.date)
having max(trunc(t1.date))-min(trunc(t1.date)) = 1
and max(trunc(t2.date))-min(trunc(t2.date)) = 1
You can use a WITH clause, also called a "common table expression" or "CTE" to break your problem down into manageable chunks.
Here is one way to do that:
WITH txn_by_date AS (
SELECT t.account_key,
t.txn_date,
tt.transaction_type,
sum(t.amount) amount
FROM txns t
INNER JOIN txn_types tt on tt.transaction_key = t.transaction_key
GROUP BY t.account_key, t.txn_date, tt.transaction_type )
SELECT tin.account_key,
tin.txn_date,
tin.amount amount_in,
tout.txn_date,
tout.amount amount_out,
tin.amount / tout.amount ratio
FROM txn_by_date tin
INNER JOIN txn_by_date tout ON tout.txn_date = tin.txn_date + 1
AND tout.transaction_type = 'out'
AND tout.account_key = tin.account_key
WHERE tin.transaction_type = 'in';
The CTE first computes the total txn amount by account, transaction type, and day. Once you have that, the main SELECT gets each 'in' total, joins it with the 'out' total from the consecutive day, and the computes the ratio.
Here is a full example, with test data (also expressed using the WITH clause):
with txns (account_key, Transaction_key, txn_date, amount) AS
(
SELECT 1, 11, TO_DATE('03/22/2019','MM/DD/YYYY'), 5000 FROM DUAL UNION ALL
SELECT 1, 12, TO_DATE('03/23/2019','MM/DD/YYYY'), 6000 FROM DUAL UNION ALL
SELECT 1, 13, TO_DATE('03/22/2019','MM/DD/YYYY'), 4000 FROM DUAL UNION ALL
SELECT 1, 14, TO_DATE('03/23/2019','MM/DD/YYYY'), 3000 FROM DUAL ),
txn_types ( transaction_key, transaction_type ) AS
(
SELECT 11, 'in' FROM DUAL UNION ALL
SELECT 12, 'out' FROM DUAL UNION ALL
SELECT 13, 'in' FROM DUAL UNION ALL
SELECT 14, 'out' FROM DUAL ),
txn_by_date AS (
SELECT t.account_key,
t.txn_date,
tt.transaction_type,
sum(t.amount) amount
FROM txns t
INNER JOIN txn_types tt on tt.transaction_key = t.transaction_key
GROUP BY t.account_key, t.txn_date, tt.transaction_type )
SELECT tin.account_key,
tin.txn_date,
tin.amount amount_in,
tout.txn_date,
tout.amount amount_out,
tin.amount / tout.amount ratio
FROM txn_by_date tin
INNER JOIN txn_by_date tout ON tout.txn_date = tin.txn_date + 1
AND tout.transaction_type = 'out'
AND tout.account_key = tin.account_key
WHERE tin.transaction_type = 'in';
+-------------+-----------+-----------+------------+------------+-------+
| ACCOUNT_KEY | TXN_DATE | AMOUNT_IN | TXN_DATE_1 | AMOUNT_OUT | RATIO |
+-------------+-----------+-----------+------------+------------+-------+
| 1 | 22-MAR-19 | 9000 | 23-MAR-19 | 9000 | 1 |
+-------------+-----------+-----------+------------+------------+-------+
ALTERNATE VERSION, FOR ANCIENT VERSIONS OF ORACLE
For really old versions of Oracle, you may need to avoid both the WITH clause and ANSI-style joins. Here is the above query rewritten to avoid those features.
SELECT tin.account_key,
tin.txn_date,
tin.amount amount_in,
tout.txn_date,
tout.amount amount_out,
tin.amount / tout.amount ratio
FROM ( SELECT t.account_key,
t.txn_date,
tt.transaction_type,
sum(t.amount) amount
FROM txns t,
txn_types tt
WHERE tt.transaction_key = t.transaction_key
GROUP BY t.account_key, t.txn_date, tt.transaction_type ) tin,
( SELECT t.account_key,
t.txn_date,
tt.transaction_type,
sum(t.amount) amount
FROM txns t,
txn_types tt
WHERE tt.transaction_key = t.transaction_key
GROUP BY t.account_key, t.txn_date, tt.transaction_type ) tout
WHERE tin.transaction_type = 'in'
AND tout.transaction_type(+) = 'out'
AND tout.account_key(+) = tin.account_key;
I would try to use to Temporary Tables. Store the sums of each in Transaction type in two temp tables, then use them to calculate your ratio. I have not tested this query but this is basically what you are looking for:
CREATE PRIVATE TEMPORARY TABLE as In_TempTable
Select T2.Transaction_Type, sum(T1.amount) In_Total
from Transactions T1
left join Transaction_Type T2 on T2.Transaction_key = T1.Transaction_Key
where T1.date between '01-jun-2017' and '30-nov-2017'
AND T2.Transactio_Type = 'In'
group by T2.Transaction_Type;
CREATE PRIVATE TEMPORARY TABLE as Out_TempTable
Select T2.Transaction_Type, sum(T1.amount) Out_Total
from Transactions T1
left join Transaction_Type T2 on T2.Transaction_key = T1.Transaction_Key
where T1.date between '01-jun-2017' and '30-nov-2017'
AND T2.Transactio_Type = 'Out'
group by T2.Transaction_Type;
Select Sum(a.In_Total)/Sum(b.Out_Total)
from In_TempTable a
Full Outer Join Out_TempTable b on b.Transaction_Type = a.Transaction_Type
I would personally change this:
T1.date between '01-jun-2017' and '30-nov-2017'
To something like this
T1.date between startdate and startdate+2
Depending on your needs of course and you will have to declare your startdate
I have a simple table of Orders like:
Order
=========
Date (DATETIME)
Price (INT)
And I would like to know how much money we earned per every month and output should look like:
January-2018 : 100
February-2018: 200
...
January-2019: 300
...
I have the following SQL:
SELECT
DATENAME(MONTH , DATEADD(M, t.MDate , -1 ) ) + '-' + CAST(t.YDate as NVARCHAR(20)),
SUM(Price)
FROM
(
SELECT
DATEPART(YYYY, o.Date) as YDate,
DATEPART(MM, o.Date) as MDate,
Price
FROM [Order] o
) t
GROUP BY t.YDate, t.MDate
ORDER BY t.YDate, t.MDate
Is it ok or may be there is a better approach ?
format() is the way to go. I would recommend:
select format(o.date, 'MMMM-yyyy') as Monthyear,
sum(o.price) as total
from orders o
group by format(datenew, 'MMMM-yyyy')
order by min(o.date)
If you are using Sql server 2012 or + , you can make use of format function as well.
select cast('2018-02-23 15:34:09.390' as datetime) as datenew, 100 as price into #temp union all
select cast('2018-02-24 15:34:09.390' as datetime) as datenew, 300 as price union all
select cast('2018-03-10 15:34:09.390' as datetime) as datenew, 500 as price union all
select cast('2018-03-11 15:34:09.390' as datetime) as datenew, 700 as price union all
select cast('2019-02-23 15:34:09.390' as datetime) as datenew, 900 as price union all
select cast('2019-02-23 15:34:09.390' as datetime) as datenew, 1500 as price
select Monthyear, total from (
select format(datenew, 'MMMM-yyyy') Monthyear, format(datenew, 'yyyyMM') NumMMYY ,sum(price) total from #temp
group by format(datenew, 'MMMM-yyyy') , format(datenew, 'yyyyMM') ) test
order by NumMMYY
Since format will convert it to nvarchar, I had to create one more column in the subquery to order it properly. If you don't mind one more column you don't have to use subquery, and you can order it with another column. I could not find any other way (except for case statement which is not a very good idea) to do it in the same select
Output:
Monthyear Total
February-2018 400
March-2018 1200
February-2019 2400
Maybe just with this?
with Orders as (
select convert(date,'2018-01-15') as [date], 10 as Price union all
select convert(date,'2018-01-20'), 20 union all
select convert(date,'2018-01-30'), 30 union all
select convert(date,'2018-02-15'), 20 union all
select convert(date,'2018-03-15'), 40 union all
select convert(date,'2018-03-20'), 50
)
select
datename(month,o.Date) + '-' + datename(year,o.Date) + ': ' +
convert(varchar(max),SUM(Price))
FROM [Orders] o
group by datename(month,o.Date) + '-' + datename(year,o.Date)
order by 1 desc
I don't get well why are you removing one month in your query, I am not doing it. You can test it here: https://rextester.com/GRKK80105
I have a table Supplier with two columns, TotalStock and Date. I'm trying to write a single query that will give me stock totals by week / month / year for a list of suppliers.
So results will look like this..
SUPPLIER WEEK MONTH YEAR
SupplierA 50 100 2000
SupplierB 60 150 2500
SupplierC 15 25 200
So far I've been playing around with multiple selects but I can't get any further than this:
SELECT Supplier,
(
SELECT Sum(TotalStock)
FROM StockBreakdown
WHERE Date >= '2014-5-12'
GROUP BY Supplier
) AS StockThisWeek,
(
SELECT Sum(TotalStock)
FROM StockBreakdown
WHERE Date >= '2014-5-1'
GROUP BY Supplier
) AS StockThisMonth,
(
SELECT Sum(TotalStock)
FROM StockBreakdown
WHERE Date >= '2014-1-1'
GROUP BY Supplier
) AS StockThisYear
This query throws an error as each individual grouping returns multiple results. I feel that I'm close to the solution but can't work out where to go
You don't have to use subqueries to achieve what you want :
SELECT Supplier
, SUM(CASE WHEN Date >= CAST('2014-05-12' as DATE) THEN TotalStock END) AS StockThisWeek
, SUM(CASE WHEN Date >= CAST('2014-05-01' as DATE) THEN TotalStock END) AS StockThisMonth
, SUM(CASE WHEN Date >= CAST('2014-01-01' as DATE) THEN TotalStock END) AS StockThisYear
FROM StockBreakdown
GROUP BY Supplier
You may need to make the selects for the columns return only a single result. You could try this (not tested currently):
SELECT Supplier,
(
SELECT TOP 1 StockThisWeek FROM
(
SELECT Supplier, Sum(TotalStock) AS StockThisWeek
FROM StockBreakdown
WHERE Date >= '2014-5-12'
GROUP BY Supplier
) tmp1
WHERE tmp1.Supplier = Supplier
) AS StockThisWeek,
(
SELECT TOP 1 StockThisMonth FROM
(
SELECT Supplier, Sum(TotalStock) AS StockThisMonth
FROM StockBreakdown
WHERE Date >= '2014-5-1'
GROUP BY Supplier
) tmp2
WHERE tmp2.Supplier = Supplier
) AS StockThisMonth,
...
This selects the supplier and then tries to create two columns StockThisWeek and StockThisMonth by selecting the first entry from the select you created before. As through the GROUP BY there should only be one entry per supplier, so you don't lose and data.
I have 2 seperate select statements, using aggregate functions in each. I would like to be able to take the results and combine them.
table_a
id int
entered_date datetime (holds utc stamp)
balance money
group_id int
table_b
id int
entered_date date
balance money
transaction_type int
query 1:
select convert(date,entered_date), sum(balance) as earned
from table_a
where group_id in (1, 2, 3, 4)
group by convert(date,entered_Date)
query 2:
select convert(date,entered_date), sum(balance) as spent
where transaction_type = 2
group by convert(date,entered_Date)
results:
query 1:
2012-05-13, 5000
2012-05-14, 12000
...
query 2:
2012-05-13, 9000
2012-05-14, 55856
...
I would like to return one row for each record without using temp tables. The result set should have a date, then earned vs. spent. I have a report running using union to get the totals and that is fine, but i need to generate a result set with 1 record and a earned vs against line. Any help with this is appreciated.
Try:
;With AllDates AS
(
select convert(date,entered_date) As EnteredDate
from table_a
where group_id in (1, 2, 3, 4)
group by convert(date,entered_Date)
UNION
select convert(date,entered_date)
from table_b
where transaction_type = 2
group by convert(date,entered_Date)
)
, AllEarned AS (
select convert(date,entered_date) AS EnteredDate, sum(balance) as Earned
from table_a
where group_id in (1, 2, 3, 4)
group by convert(date,entered_Date)
)
, AllSpent AS (
select convert(date,entered_date) AS EnteredDate, sum(balance) as Spent
from table_b
where transaction_type = 2
group by convert(date,entered_Date)
)
SELECT
d.EnteredDate, e.Earned, s.Spent
FROM AllDates d
LEFT OUTER JOIN AllEarned e ON d.EnteredDate=e.EnteredDate
LEFT OUTER JOIN AllSpent s ON d.EnteredDate=s.EnteredDate
ORDER BY 1,2,3
You can combine these using logic, assuming that both are from the same table
(the second query is missing the from statement):
select convert(date,entered_date),
sum(case when group_id in (1, 2, 3, 4) then balance end) as earned,
sum(case when transaction_type = 2 then balance end) as spend
from table_a
group by convert(date,entered_Date)
SELECT
CASE WHEN a.a_date IS NULL THEN b.a_date ELSE a.a_date END as a_data,
a.earned,
b.spent
FROM
(select
convert(date,entered_date) as a_date,
sum(balance) as earned
from table_a
where group_id in (1, 2, 3, 4)
group by entered_Date) A
FULL OUTER JOIN
(select
convert(date,entered_date) as a_date,
sum(balance) as spent
from table_a
where transaction_type = 2
group by entered_Date) B
ON A.a_date=b.a_date
Or using FULL OUTER JOIN if there are data that don't meet both conditions. And using CASE WHEN a.a_date IS NULL THEN b.a_date ELSE a.a_date END as a_data
Assuming earned amounts are from table_a and spent amounts are from table_b,
; WITH a AS (
select entered_date=convert(date,entered_date), balance as earned, 0 AS spent
from table_a
where group_id in (1, 2, 3, 4)
UNION ALL
select entered_date=convert(date,entered_date), 0 AS earned, balance as spent
from table_b
where transaction_type = 2
)
SELECT entered_date
, earned=SUM(earned)
, spent=SUM(spent)
FROM a
GROUP BY entered_date