STUFF XML PATH + Partition By

Question

Post reply

STUFF XML PATH + Partition By

JeremyU

SSChasing Mays

Points: 640
More actions
February 18, 2023 at 12:08 am
Go to Answer
#4151248
Hello,
I need to concatenate multiple rows into one based on a condition:
DC, UPC, MASTERCHAINNAME
DC01, 9780061491900, TARGET
DC01, 9780061491900, WALMART
DC01, 9780061491900, COTSCO
So I want it to be
DC, UPC, MASTERCHAINNAME
DC01, 9780061491900, TARGET, WALMART, COSTCO
I read it is possible to do it with the help of STUFF XML PATH function, which does concat all fields, but it doesn't group them therefore I get all the chains in one column and it doesn't look if it applies to that DC and UPC#. I need a function like PARTITION BY that can filter MASTERCHAINNMAES by DC & UPC.
```
SELECT DC, UPC, stuff((select ', ' + MASTERCHAINNAME 
              from pre_filtering
              for xml path ('') 
             ), 1, 2, '') as MASTERCHAINNAME
FROM pre_filtering
ORDER BY DC, UPC
```
Right now it just gives me all chain names and that is wrong (without looking at UNIT# (UPC) and DC.
Does someone know how this function can work?
- This topic was modified 1 year, 10 months ago by JeremyU. Reason: SQL
- This topic was modified 1 year, 10 months ago by JeremyU.

Jonathan AC Roberts

SSCoach

Points: 18362

February 18, 2023 at 3:28 am

Answer

#4151286

Try this instead, it splits it into 2 separate statements:

IF OBJECT_ID('tempdb..#pre_filtering') IS NOT NULL
    DROP TABLE #pre_filtering;

;WITH pre_filtering AS 
(
    SELECT DISTINCT 
           a.[DC], 
           a.[UPC], 
           b.[STARTDATE] AS 'MIN START POG DATE', 
           c.[ENDDATE] AS 'MAX START POG DATE', 
           a.[MASTERCHAINNAME]
      FROM [snapshot] a
      LEFT JOIN (SELECT DISTINCT 
                        UPC, 
                        MASTERCHAINNAME, 
                        STARTDATE, 
                        ROW_NUMBER() OVER(PARTITION BY UPC, MASTERCHAINNAME ORDER BY STARTDATE ASC) RN
                   FROM [snapshot]
                ) b 
             ON a.UPC = b.UPC
            AND a.[MASTERCHAINNAME] = b.[MASTERCHAINNAME]
      LEFT JOIN (SELECT DISTINCT 
                        UPC, 
                        MASTERCHAINNAME, 
                        ENDDATE, 
                        ROW_NUMBER() OVER(PARTITION BY UPC, MASTERCHAINNAME ORDER BY ENDDATE DESC) RN
                   FROM [snapshot]
                ) c ON a.UPC = c.UPC
            AND a.[MASTERCHAINNAME] = c.[MASTERCHAINNAME]
          WHERE b.RN = '1'
            AND c.RN = '1'
)
SELECT * 
  INTO #pre_filtering  
  FROM pre_filtering
;

SELECT a.DC, 
       a.UPC, 
       MIN(a.[MIN START POG DATE]) OVER(PARTITION BY a.DC, a.UPC) AS 'MIN START POG DATE', 
       MAX(a.[MAX START POG DATE]) OVER(PARTITION BY a.DC, a.UPC) AS 'MAX START POG DATE', 
       STUFF((SELECT ', ' + b.MASTERCHAINNAME
               FROM #pre_filtering   b
              WHERE b.DC  = a.DC
                AND b.UPC = a.UPC
                FOR xml PATH('')
              ), 1, 2, '') AS MASTERCHAINNAME  -- doesn't understand that some UPCs + DCs are not connected to MASTERCHAINNAMES
  FROM #pre_filtering a
 GROUP BY DC, UPC, [MIN START POG DATE], [MAX START POG DATE]
 ORDER BY DC, UPC
;

Viewing 15 posts - 1 through 15 (of 18 total)

You must be logged in to reply to this topic. Login to reply

Jonathan AC Roberts SSCoach Points: 18362 More actions · Answer 1

Create some test data in a consumable format (you should have done this):

DROP TABLE IF EXISTS #pre_filtering;

CREATE TABLE #pre_filtering (
    DC VARCHAR(50),
    UPC VARCHAR(50),
    MASTERCHAINNAME VARCHAR(50)
);

INSERT INTO #pre_filtering (DC, UPC, MASTERCHAINNAME)
VALUES ('DC01', '9780061491900', 'TARGET'),
       ('DC01', '9780061491900', 'WALMART'),
       ('DC01', '9780061491900', 'COSTCO')
;

With SQL 2019 you can use STRING_AGG instead:

SELECT DC, 
       UPC, 
       STRING_AGG(MASTERCHAINNAME, ',') as MASTERCHAINNAME
  FROM #pre_filtering
 GROUP BY DC, UPC
 ORDER BY DC, UPC
;

JeremyU SSChasing Mays Points: 640 More actions · Answer 2

I have SSMS v18.5.1. When I tried to do it, it gave me an error namely 'STRING_AGG' is not a recognized built-in function name.

Jonathan AC Roberts SSCoach Points: 18362 More actions · Answer 3

JeremyU wrote:

I have SSMS v18.5.1. When I tried to do it, it gave me an error namely 'STRING_AGG' is not a recognized built-in function name.

It's the version of SQL Server not the version of SSMS you need.

Try: SELECT @@VERSION

Jonathan AC Roberts SSCoach Points: 18362 More actions · Answer 4

You need a GROUP BY:

SELECT DC, UPC, stuff((select ', ' + MASTERCHAINNAME 
              from pre_filtering b
             where b.DC = a.DC
               and b.UPC = a.UPC
              for xml path ('') 
             ), 1, 2, '') as MASTERCHAINNAME
FROM pre_filtering a
GROUP BY DC, UPC
ORDER BY DC, UPC
;

This reply was modified 1 year, 10 months ago by Jonathan AC Roberts. Reason: Added where b.DC = a.DC and b.UPC = a.UPC

JeremyU SSChasing Mays Points: 640 More actions · Answer 5

JeremyU

SSChasing Mays

Points: 640

February 18, 2023 at 2:13 am

#4151265

Jonathan it is Microsoft SQL Server 2016

Jonathan AC Roberts SSCoach Points: 18362 More actions · Answer 6

JeremyU wrote:

Jonathan it is Microsoft SQL Server 2016

You asked the question in the SQL 2019 forum.

Try the code in the last comment I made.

JeremyU SSChasing Mays Points: 640 More actions · Answer 7

Jonathan thanks for your help. I appreciate it.

There is an issue here. It just prints all the chain names and doesn't differentiate that some UPCs are not available or available for different locations.

Issue_04

Whereas this unit is available only at those locations

Issue_06

I don't know if you want to see my whole query, but this is it:

I am just trying to list in which CHAINS items are available, but the code as it is right now just lists all the chains without looking at DC and UPC (item#).

;with pre_filtering AS (
SELECT distinct a.[DC], a.[UPC]
      ,b.[STARTDATE] AS 'MIN START POG DATE'
      ,c.[ENDDATE] AS 'MAX START POG DATE'
      ,a.[MASTERCHAINNAME]    
  FROM [snapshot] a
  LEFT OUTER JOIN (SELECT distinct UPC, MASTERCHAINNAME, STARTDATE, ROW_NUMBER() OVER ( PARTITION BY UPC, MASTERCHAINNAME ORDER BY STARTDATE ASC) RN FROM  [snapshot]) b
ON a.UPC=b.UPC
AND a.[MASTERCHAINNAME]=b.[MASTERCHAINNAME]
LEFT OUTER JOIN (SELECT distinct UPC, MASTERCHAINNAME, ENDDATE, ROW_NUMBER() OVER ( PARTITION BY UPC, MASTERCHAINNAME ORDER BY ENDDATE DESC) RN FROM  [snapshot]) c
ON a.UPC=c.UPC
AND a.[MASTERCHAINNAME]=c.[MASTERCHAINNAME]
WHERE b.RN='1' and c.RN='1'

)


SELECT DC, UPC, MIN([MIN START POG DATE]) OVER (PARTITION BY UPC) AS 'MIN START POG DATE'
,MAX([MAX START POG DATE]) OVER (PARTITION BY UPC) AS 'MAX START POG DATE'

,stuff((select ', ' + MASTERCHAINNAME 
              from pre_filtering
              for xml path ('') 
             ), 1, 2, '') as MASTERCHAINNAME  -- doesn't understand that some UPCs + DCs are not connected to MASTERCHAINNAMES
FROM pre_filtering
GROUP BY DC, UPC, [MIN START POG DATE], [MAX START POG DATE]
ORDER BY DC, UPC
;

Jonathan AC Roberts SSCoach Points: 18362 More actions · Answer 8

Try this:

;WITH pre_filtering AS 
(
    SELECT DISTINCT 
           a.[DC], 
           a.[UPC], 
           b.[STARTDATE] AS 'MIN START POG DATE', 
           c.[ENDDATE] AS 'MAX START POG DATE', 
           a.[MASTERCHAINNAME]
      FROM [snapshot] a
      LEFT JOIN (SELECT DISTINCT 
                        UPC, 
                        MASTERCHAINNAME, 
                        STARTDATE, 
                        ROW_NUMBER() OVER(PARTITION BY UPC, MASTERCHAINNAME ORDER BY STARTDATE ASC) RN
                   FROM [snapshot]
                ) b 
             ON a.UPC = b.UPC
            AND a.[MASTERCHAINNAME] = b.[MASTERCHAINNAME]
      LEFT JOIN (SELECT DISTINCT 
                        UPC, 
                        MASTERCHAINNAME, 
                        ENDDATE, 
                        ROW_NUMBER() OVER(PARTITION BY UPC, MASTERCHAINNAME ORDER BY ENDDATE DESC) RN
                   FROM [snapshot]
                ) c ON a.UPC = c.UPC
            AND a.[MASTERCHAINNAME] = c.[MASTERCHAINNAME]
          WHERE b.RN = '1'
            AND c.RN = '1'
)
SELECT a.DC, 
       a.UPC, 
       MIN(a.[MIN START POG DATE]) OVER(PARTITION BY a.UPC) AS 'MIN START POG DATE', 
       MAX(a.[MAX START POG DATE]) OVER(PARTITION BY a.UPC) AS 'MAX START POG DATE', 
       STUFF((SELECT ', ' + b.MASTERCHAINNAME
               FROM pre_filtering b
              WHERE b.DC  = a.DC
                AND b.UPC = a.UPC
                FOR xml PATH('')
              ), 1, 2, '') AS MASTERCHAINNAME  -- doesn't understand that some UPCs + DCs are not connected to MASTERCHAINNAMES
  FROM pre_filtering a
 GROUP BY DC, UPC, [MIN START POG DATE], [MAX START POG DATE]
 ORDER BY DC, UPC
;

JeremyU SSChasing Mays Points: 640 More actions · Answer 9

Something must be incorrect. It just keeps running and running. There are only around 249K rows I believe and it is running for pretty long time already and no output.

Jonathan AC Roberts SSCoach Points: 18362 More actions · Answer 10

If you want to make it even faster you can add an index to the temporary table:

IF OBJECT_ID('tempdb..#pre_filtering') IS NOT NULL
    DROP TABLE #pre_filtering;

;WITH pre_filtering AS 
(
    SELECT DISTINCT 
           a.[DC], 
           a.[UPC], 
           b.[STARTDATE] AS 'MIN START POG DATE', 
           c.[ENDDATE] AS 'MAX START POG DATE', 
           a.[MASTERCHAINNAME]
      FROM [snapshot] a
      LEFT JOIN (SELECT DISTINCT 
                        UPC, 
                        MASTERCHAINNAME, 
                        STARTDATE, 
                        ROW_NUMBER() OVER(PARTITION BY UPC, MASTERCHAINNAME ORDER BY STARTDATE ASC) RN
                   FROM [snapshot]
                ) b 
             ON a.UPC = b.UPC
            AND a.[MASTERCHAINNAME] = b.[MASTERCHAINNAME]
      LEFT JOIN (SELECT DISTINCT 
                        UPC, 
                        MASTERCHAINNAME, 
                        ENDDATE, 
                        ROW_NUMBER() OVER(PARTITION BY UPC, MASTERCHAINNAME ORDER BY ENDDATE DESC) RN
                   FROM [snapshot]
                ) c ON a.UPC = c.UPC
            AND a.[MASTERCHAINNAME] = c.[MASTERCHAINNAME]
          WHERE b.RN = '1'
            AND c.RN = '1'
)
SELECT * 
  INTO #pre_filtering  
  FROM pre_filtering
;

CREATE NONCLUSTERED INDEX IX_#pre_filtering_1 ON #pre_filtering(DC, UPC) 
       INCLUDE (MASTERCHAINNAME, [MIN START POG DATE], [MAX START POG DATE])
;

SELECT a.DC, 
       a.UPC, 
       MIN(a.[MIN START POG DATE]) OVER(PARTITION BY a.DC, a.UPC) AS 'MIN START POG DATE', 
       MAX(a.[MAX START POG DATE]) OVER(PARTITION BY a.DC, a.UPC) AS 'MAX START POG DATE', 
       STUFF((SELECT ', ' + b.MASTERCHAINNAME
               FROM #pre_filtering   b
              WHERE b.DC  = a.DC
                AND b.UPC = a.UPC
                FOR xml PATH('')
              ), 1, 2, '') AS MASTERCHAINNAME  -- doesn't understand that some UPCs + DCs are not connected to MASTERCHAINNAMES
  FROM #pre_filtering a
 GROUP BY DC, UPC, [MIN START POG DATE], [MAX START POG DATE]
 ORDER BY DC, UPC
;

JeremyU SSChasing Mays Points: 640 More actions · Answer 11

Jonathan thank you!! It works. I appreciate your help!!

Jeff Moden SSC Guru Points: 1004406 More actions · Answer 12

NM.

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)

Jeff Moden SSC Guru Points: 1004406 More actions · Answer 13

NM.

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)