COUNT number of rows in a GROUP on higher aggregate level - sql-server

Question

I am trying to find out how many rows of a certain item exist in the table, e.g. in the following example for itemID 1 I need the result 5 (not 3, which is what I currently get). I am tempted to add TransactionID into the PARTITION BY clause, but that results in Msg 8120 since the query does not GROUP by TransactionID. Well, if it did then getting that count would be easy, but I do not want to group on Transaction Level. What can I do to get that ItemCount right? It must be so easy but I am banging my head.
DECLARE #t TABLE (TransactionID INT PRIMARY KEY IDENTITY, CustomerID INT, ItemID INT);
INSERT INTO #t (CustomerID, ItemID)
VALUES
(1, 1),
(2, 1),
(3, 1),
(4, 2),
(1, 1),
(2, 2),
(3, 3),
(4, 4),
(1, 1);
SELECT
CustomerID,
ItemID,
Rows = COUNT(*),
ItemRowCount = COUNT(*) OVER (PARTITION BY ItemID)
FROM
#t
GROUP BY
CustomerID,
ItemID
ORDER BY
ItemID,
CustomerID;
EDIT: I was overaggregating, I guess. Sebastian Meine got me on the track and his answer is right so I accepted it. However, this subquery works for my:
SELECT
CustomerID,
ItemID,
Rows = COUNT(*),
ItemRowCount = (SELECT COUNT(*) FROM #t x WHERE t.ItemID = x.ItemID)
FROM
#t t
GROUP BY
CustomerID,
ItemID
ORDER BY
ItemID,
CustomerID;

Answer 1

You need to pull your outer group count out of the actual group by query. The easiest way to do that is like this:
SQL Fiddle
MS SQL Server 2008 Schema Setup:
CREATE TABLE dbo.tbl (TransactionID INT PRIMARY KEY IDENTITY, CustomerID INT, ItemID INT);
INSERT INTO dbo.tbl (CustomerID, ItemID)
VALUES
(1, 1),
(2, 1),
(3, 1),
(4, 2),
(1, 1),
(2, 2),
(3, 3),
(4, 4),
(1, 1);
Query 1:
SELECT *,SUM(Rows)OVER(PARTITION BY ItemId) ItemCnt
FROM(
SELECT
CustomerID,
ItemID,
Rows = COUNT(*)
FROM
dbo.tbl
GROUP BY
CustomerID,
ItemID
)X
ORDER BY
ItemID,
CustomerID
Results:
| CUSTOMERID | ITEMID | ROWS | ITEMCNT |
----------------------------------------
| 1 | 1 | 3 | 5 |
| 2 | 1 | 1 | 5 |
| 3 | 1 | 1 | 5 |
| 2 | 2 | 1 | 2 |
| 4 | 2 | 1 | 2 |
| 3 | 3 | 1 | 1 |
| 4 | 4 | 1 | 1 |
Notice that I add the inner counts together instead of recounting from scratch.

Answer 2

You can use simple count and group by:
SELECT
ItemID,
ItemRowCount = COUNT(1)
FROM
#t
GROUP BY
ItemID
ORDER BY
ItemID
or if you need attach total rows count and item row count to every row:
SELECT
CustomerID,
ItemID,
Rows = COUNT(1) over (),
ItemRowCount = COUNT(1) OVER (PARTITION BY ItemID)
FROM
#t
ORDER BY
ItemID,
CustomerID;

Answer 3

To find out how many rows of a certain item exist in the table, we may not need CustomerId. Use following query -
DECLARE #t TABLE (TransactionID INT PRIMARY KEY IDENTITY, CustomerID INT, ItemID INT);
INSERT INTO #t (CustomerID, ItemID)
VALUES
(1, 1),
(2, 1),
(3, 1),
(4, 2),
(1, 1),
(2, 2),
(3, 3),
(4, 4),
(1, 1);
;WITH cte AS(
SELECT itemId, ROW_NUMBER() OVER (PARTITION BY itemId ORDER BY itemId DESC) AS row_cnt FROM #t
)
SELECT itemId, MAX(row_cnt) row_count FROM cte GROUP BY itemId
It will return -
itemId row_count
1 5
2 2
3 1
4 1
And if is case you need customerId, the use -
;WITH cte AS(
SELECT customerId, itemId, ROW_NUMBER() OVER (PARTITION BY itemId ORDER BY itemId DESC) AS row_cnt FROM #t
)
SELECT customerId, itemId, MAX(row_cnt) item_count FROM cte GROUP BY CustomerID, itemId
It will return -
customerId itemId item_count
1 1 5
2 1 2
3 1 3
2 2 1
4 2 2
3 3 1
4 4 1

Answer 4

You can use a CTE or join to a sub-table (like this)
SELECT
tbl.CustomerID,
tbl.ItemID,
Rows = COUNT(*),
ItemRowCount
FROM tbl
JOIN (SELECT ItemID, Count(*) as ItemRowCount
FROM tbl
GROUP BY ItemID) t ON tbl.ItemID = t.ItemID
GROUP BY
tbl.CustomerID,
tbl.ItemID,
ItemRowCount
ORDER BY
tbl.ItemID,
CustomerID;
or this
SELECT
tbl.CustomerID,
tbl.ItemID,
Rows = COUNT(*),
MAX(ItemRowCount)
FROM tbl
JOIN (SELECT ItemID, Count(*) as ItemRowCount
FROM tbl
GROUP BY ItemID) t ON tbl.ItemID = t.ItemID
GROUP BY
tbl.CustomerID,
tbl.ItemID
ORDER BY
tbl.ItemID,
CustomerID;