It was hard to come up with an understandable title for this question.
I'll try to explain with an example.
First I had a simple table INFO in Oracle DB:
year type message
---- ---- -------
2001 1 cakes are yammy
2003 2 apples are dangerous
2012 2 bananas are suspicious
2005 3 cats are tricky
And I need to select newest messages of certain types (for example type = 1 or type = 2):
2001 1 cakes are yammy
2012 2 bananas are suspicious
So I used the query:
select * from INFO i
where year = (select max(year) from INFO i_last where i.type = i_last.type)
and i.type in (1, 2)
But now I need to add a new "quarter" column to my INFO table. And select the newest records by year and quarter.
year quarter type message
---- ------- ---- -------
2001 2 1 cakes are yammy
2012 3 2 onions are cruel
2012 1 2 bananas are suspicious
2005 1 3 cats are tricky
The newest records with type 1 or 2 will be:
2001 2 1 cakes are yammy
2012 3 2 onions are cruel
How should such query look like?
Analytic functions are your friend:
SELECT MAX( year ) KEEP ( DENSE_RANK LAST ORDER BY year ASC, quarter ASC, message ASC ) AS year,
MAX( quarter ) KEEP ( DENSE_RANK LAST ORDER BY year ASC, quarter ASC, message ASC ) AS quarter,
MAX( message ) KEEP ( DENSE_RANK LAST ORDER BY year ASC, quarter ASC, message ASC ) AS message,
type
FROM info
GROUP BY type;
SQLFIDDLE
Give this a shot:
SELECT i.year, i.quarter, i.type, i.message
FROM INFO i
JOIN INFO iy
ON i.type = iy.type
JOIN INFO iq
ON iq.type = iy.type
WHERE i.type IN (1,2)
GROUP BY i.year, i.quarter, i.type, i.message
HAVING i.year = MAX(iy.year) AND i.quarter = MAX(iq.quarter)
Related
I work with crashes and mileage for the same year which is Year in table. Crashes are are there for every record, but annual mileage is not. NULLs for mileage could be at the beginning or at the end of the time period for certain customer. Also, couple of annual mileage records can be missing as well. I do not know how to overcome this. I try to do it in CASE statement but then I do not know how to code it properly. Issue needs to be resolved in SQL and use SQL Server.
This is how the output looks like and I need to have mileage for every single year for each customer.
The info I am pulling from is proprietary database and the records themselves should be untouched as is. I just need code in query which will modify my current output to output where I have mileage for every year. I appreciate any input!
Year
Customer
Crashes
Annual_Mileage
2009
123
5
3453453
2010
123
1
NULL
2011
123
0
54545
2012
123
14
376457435
2013
123
3
63453453
2014
123
4
NULL
2015
123
15
6346747
2016
123
0
NULL
2017
123
2
534534
2018
123
7
NULL
2019
123
11
NULL
2020
123
15
565435
2021
123
12
474567546
2022
123
7
NULL
Desired Results
Year
Customer
Crashes
Annual_Mileage
2009
123
5
3453453
2010
123
1
175399 (prior value is taken)
2011
123
0
54545
2012
123
14
376457435
2013
123
3
63453453
2014
123
4
34900100 (avg of 2 adjacent values)
2015
123
15
6346747
2016
123
0
3440641 (avg of 2 adjacent values)
2017
123
2
534534
2018
123
7
534534 ( prior value is taken)
2019
123
11
549985 (avg of 2 adjacent values)
2020
123
15
565435
2021
123
12
474567546
2022
123
7
474567546 (prior value is taken)
SELECT Year,
Customer,
Crashes,
CASE
WHEN Annual_Mlg IS NOT NULL THEN Annual_Mlg
WHEN Annual_Mlg IS NULL THEN
CASE
WHEN PREV.Annual_Mlg IS NOT NULL
AND NEXT.Annual_Mlg IS NOT NULL
THEN ( PREV.Annual_Mlg + NEXT.Annual_Mlg ) / 2
ELSE 0
END
END AS Annual_Mlg
FROM #table
The above code doesn't work, but I just need to start somehow and that what I have currently.
I understand what I need to do I just do not know how to code it in SQL.
After i applied row_number () function i got this output for first 2 clients and for the rest of the 4 clients row_number() function gave correct output. i have no idea why is that. I thought may be because i used "full join" before to combine mileage and crashes table?
enter image description here
Your use of #table tells me that you're using MS SQL Server (a temporary table, probably in a stored procedure).
You want to:
select all the rows in #table
joined with the matching row (if any) for the previous year, and
joined with the matching row (if any) for the next year
Then it's easy. Assuming the primary key on your #table is composed of the year and customer columns, something like this ought to do you:
select t.year ,
t.customer ,
t.crashes ,
annual_milage = coalesce(
t.annual_milage ,
( coalesce( p.annual_mileage, 0 ) +
coalesce( n.annual_mileage, 0 )
) / 2
)
from #table t -- take all the rows
left join #table p on p.year = t.year - 1 -- with the matching row for
and p.customer = t.customer -- the previous year (if any)
left join #table n on n.year = t.year + 1 -- and the matching row for
and n.customer = t.customer -- the next year (if any)
Notes:
What value you default to if the previous or next year doesn't exist is up to you (zero? some arbitrary value?)
Is the previous/next year guaranteed to be the current year +/- 1?
If not, you may have to use derived tables as the source for the
prev/next data, selecting the closest previous/next year (that sort
of thing rather complicates the query significantly).
Edited To Note:
If you have discontiguous years for each customer such that the "previous" and "next" years for a given customer are not necessarily the current year +/- 1, then something like this is probably the most straightforward way to find the previous/next year.
We use a derived table in our from clause, and assign a sequential number in lieu of year for each customer, using the ranking function row_number() function. This query, then
select row_nbr = row_number() over (
partition by x.customer
order by x.year
) ,
x.*
from #table x
would produce results along these lines:
row_nbr
customer
year
...
1
123
1992
...
2
123
1993
...
3
123
1995
...
4
123
2020
...
1
456
2001
...
2
456
2005
...
3
456
2020
...
And that leads us to this:
select year = t.year ,
customer = t.customer ,
crashes = t.crashes ,
annual_mileage = coalesce(
t.mileage,
coalesce(
t.annual_mileage,
(
coalesce(p.annual_mileage,0) +
coalesce(n.annual_mileage,0)
) / 2
),
)
from (
select row_nbr = row_number() over (
partition by x.customer
order by x.year
) ,
x.*
from #table x
) t
left join #table p on p.customer = t.customer and p.row_nbr = t.row_nbr-1
left join #table n on n.customer = t.customer and n.row_nbr = t.row_nbr+1
In a firebird database with a table "Sales", I need to select the first sale of all customers. See below a sample that show the table and desired result of query.
---------------------------------------
SALES
---------------------------------------
ID CUSTOMERID DTHRSALE
1 25 01/04/16 09:32
2 30 02/04/16 11:22
3 25 05/04/16 08:10
4 31 07/03/16 10:22
5 22 01/02/16 12:30
6 22 10/01/16 08:45
Result: only first sale, based on sale date.
ID CUSTOMERID DTHRSALE
1 25 01/04/16 09:32
2 30 02/04/16 11:22
4 31 07/03/16 10:22
6 22 10/01/16 08:45
I've already tested following code "Select first row in each GROUP BY group?", but it did not work.
In Firebird 2.5 you can do this with the following query; this is a minor modification of the second part of the accepted answer of the question you linked to tailored to your schema and requirements:
select x.id,
x.customerid,
x.dthrsale
from sales x
join (select customerid,
min(dthrsale) as first_sale
from sales
group by customerid) p on p.customerid = x.customerid
and p.first_sale = x.dthrsale
order by x.id
The order by is not necessary, I just added it to make it give the order as shown in your question.
With Firebird 3 you can use the window function ROW_NUMBER which is also described in the linked answer. The linked answer incorrectly said the first solution would work on Firebird 2.1 and higher. I have now edited it.
Search for the sales with no earlier sales:
SELECT S1.*
FROM SALES S1
LEFT JOIN SALES S2 ON S2.CUSTOMERID = S1.CUSTOMERID AND S2.DTHRSALE < S1.DTHRSALE
WHERE S2.ID IS NULL
Define an index over (customerid, dthrsale) to make it fast.
in Firebird 3 , get first row foreach customer by min sales_date :
SELECT id, customer_id, total, sales_date
FROM (
SELECT id, customer_id, total, sales_date
, row_number() OVER(PARTITION BY customer_id ORDER BY sales_date ASC ) AS rn
FROM SALES
) sub
WHERE rn = 1;
İf you want to get other related columns, This is where your self-answer fails.
select customer_id , min(sales_date)
, id, total --what about other colums
from SALES
group by customer_id
So simple as:
select CUSTOMERID min(DTHRSALE) from SALES group by CUSTOMERID
I need to count a value (M_Id) at each change of a date (RS_Date) and create a column grouped by the RS_Date that has an active total from that date.
So the table is:
Ep_Id Oa_Id M_Id M_StartDate RS_Date
--------------------------------------------
1 2001 5 1/1/2014 1/1/2014
1 2001 9 1/1/2014 1/1/2014
1 2001 3 1/1/2014 1/1/2014
1 2001 11 1/1/2014 1/1/2014
1 2001 2 1/1/2014 1/1/2014
1 2067 7 1/1/2014 1/5/2014
1 2067 1 1/1/2014 1/5/2014
1 3099 12 1/1/2014 3/2/2014
1 3099 14 2/14/2014 3/2/2014
1 3099 4 2/14/2014 3/2/2014
So my goal is like
RS_Date Active
-----------------
1/1/2014 5
1/5/2014 7
3/2/2014 10
If the M_startDate = RS_Date I need to count the M_id and then for
each RS_Date that is not equal to the start date I need to count the M_Id and then add that to the M_StartDate count and then count the next RS_Date and add that to the last active count.
I can get the basic counts with something like
(Case when M_StartDate <= RS_Date
then [m_Id] end) as Test.
But I am stuck as how to get to the result I want.
Any help would be greatly appreciated.
Brian
-added in response to comments
I am using Server Ver 10
If using SQL SERVER 2012+ you can use ROWS with your the analytic/window functions:
;with cte AS (SELECT RS_Date
,COUNT(DISTINCT M_ID) AS CT
FROM Table1
GROUP BY RS_Date
)
SELECT *,SUM(CT) OVER(ORDER BY RS_Date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS Run_CT
FROM cte
Demo: SQL Fiddle
If stuck using something prior to 2012 you can use:
;with cte AS (SELECT RS_Date
,COUNT(DISTINCT M_ID) AS CT
FROM Table1
GROUP BY RS_Date
)
SELECT a.RS_Date
,SUM(b.CT)
FROM cte a
LEFT JOIN cte b
ON a.RS_DAte >= b.RS_Date
GROUP BY a.RS_Date
Demo: SQL Fiddle
You need a cumulative sum, easy in SQL Server 2012 using Windowed Aggregate Functions. Based on your description this will return the expected result
SELECT p_id, RS_Date,
SUM(COUNT(*))
OVER (PARTITION BY p_id
ORDER BY RS_Date
ROWS UNBOUNDED PRECEDING)
FROM tab
GROUP BY p_id, RS_Date
It looks like you want something like this:
SELECT
RS_Date,
SUM(c) OVER (PARTITION BY M_StartDate ORDER BY RS_Date ROWS UNBOUNDED PRECEEDING)
FROM
(
SELECT M_StartDate, RS_Date, COUNT(DISTINCT M_Id) AS c
FROM my_table
GROUP BY M_StartDate, RS_Date
) counts
The inline view computes the counts of distinct M_Id values within each (M_StartDate, RS_Date) group (distinctness enforced only within the group), and the outer query uses the analytic version of SUM() to add up the counts within each M_StartDate.
Note that this particular query will not exactly reproduce your example results. It will instead produce:
RS_Date Active
-----------------
1/1/2014 5
1/5/2014 7
3/2/2014 8
3/2/2014 2
This is on account of some rows in your example data with RS_Date 3/2/2014 having a later M_StartDate than others. If this is not what you want then you need to clarify the question, which currently seems a bit inconsistent.
Unfortunately, analytic functions are not available until SQL Server 2012. In SQL Server 2010, the job is messier. It could be done like this:
WITH gc AS (
SELECT M_StartDate, RS_Date, COUNT(DISTINCT M_Id) AS c
FROM my_table
GROUP BY M_StartDate, RS_Date
)
SELECT
RS_Date,
(
SELECT SUM(c)
FROM gc2
WHERE gc2.M_StartDate = gc.M_StartDate AND gc2.RS_Date <= gc.RS_Date
) AS Active
FROM gc
If you are using SQL 2012 or newer you can use LAG to produce a running total.
https://msdn.microsoft.com/en-us/library/hh231256(v=sql.110).aspx
I have a requirement to print the corresponding highest year and highest quarter for a given column.
Input is in a table:
cityprogram year quarter
=========== ==== =======
Abc 1998 1
Abc 1999 4
Abc 1999 4
Abc 1998 3
xyz 1998 4
xyz 1998 1
xyz 2000 3
It should print
Abc 1999 4
xyz 2000 3
I tried many joins, max conditions, I seem to get quarter 4 and 4 for both of them :( thanks
Use a window function like ROW_NUMBER in a common-table-expression:
WITH CTE AS(
SELECT [cityprogram], [year], [quarter],
RN = ROW_NUMBER() OVER (
PARTITION BY [cityprogram]
ORDER BY [year] DESC, [quarter] DESC)
FROM dbo.TableName
)
SELECT [cityprogram], [year], [quarter]
FROM CTE
WHERE RN = 1
DEMO
CITYPROGRAM YEAR QUARTER
Abc 1999 4
xyz 2000 3
ROW_NUMBER returns only one row per group even if there are ties(cityprograms with the same highest year+quarter). If you then want to show all highest you can replace ROW_NUMBER with DENSE_RANK.
Say I have a table
StoreID TotalSales Month Year
-- ---------- ----- ----
1 10 1 2012
2 2 1 2012
3 15 1 2012
1 4 2 2012
2 5 2 2012
I need: For each unique "Month/Year", grab the top two StoreID's with the highest Sales.
I'm at a loss on how to do this. I tried with a cross apply but that doesn't seem to work. This is all way over my head so hopefully someone can give me a nudge in the right direction.
This query uses Common Table Expression and Window Function to be able to get all the columns within the row. It works on SQL Server 2005 and up
WITH records
AS
(
SELECT StoreID, TotalSales , Month, Year,
DENSE_RANK() OVER (PARTITION BY Month, Year
ORDER BY TotalSales DESC) rn
FROM tableName
)
SELECT StoreID, TotalSales , Month, Year
FROM records
WHERE rn <= 2
SQLFiddle Demo