Grab one record per ID with multiple Lead values - sql

Question

I have a table like this:
ID | Val | Quantity
----------------------
1 | A | 11
1 | B | 15
1 | B | 19
1 | Z | 45
2 | D | 4
2 | E | 25
2 | F | 13
2 | Y | 2
3 | G | 10
3 | H | 15
3 | I | 19
I want to select the top record for each ID ordered by VAL, Quantity AND add the next 2 Val/Quantity within the sort as columns to that row. My expected output look like this:
ID | Val | Quantity | VAL2 | Quantity2 | VAL3 | Quantity3
-------------------------------------------------------------------
1 | A | 11 | B | 15 | B | 19
2 | B | 15 | D | 4 | E | 25
3 | C | 19 | G | 10 | H | 15
I've almost done it using lead, but I don't know how to get rid of the rest of the records in my data-set, as I only want the top.
SELECT ID,
VAL,
Quantity,
lead(VAL,1) over (order by VAL, Quantity ASC) as Val2,
lead(Quantity,1) over (order by VAL, Quantity ASC) as Quantity2,
lead(VAL,2) over (order by VAL, Quantity ASC) as Val3,
lead(Quantity,2) over (order by VAL, Quantity ASC) as Quantity3,
FROM MY_TABLE
order by VAL, Quantity ASC
How can I only select the top record for each ID, while maintaining the lead records? Or is there a more elegant/efficient way to do this?

Answer 1

From your question it seems the expected output should actually be:
ID VAL QUANTITY VAL2 QUANTITY2 VAL3 QUANTITY3
1 A 11 B 15 B 19
2 D 4 E 25 F 13
3 G 10 H 15 I 19
You can get this result with a CTE which generates the LEAD values, as well as a ROW_NUMBER for each set of values. You can then select the first row for each ID from the CTE:
WITH CTE AS (
SELECT ID,
Val, Quantity,
LEAD(Val) OVER (PARTITION BY ID ORDER BY Val, Quantity) AS Val2,
LEAD(Quantity) OVER (PARTITION BY ID ORDER BY Val, Quantity) AS Quantity2,
LEAD(Val, 2) OVER (PARTITION BY ID ORDER BY Val, Quantity) AS Val3,
LEAD(Quantity, 2) OVER (PARTITION BY ID ORDER BY Val, Quantity) AS Quantity3,
ROW_NUMBER() OVER (PARTITION BY ID ORDER BY Val, Quantity) AS rn
FROM MY_TABLE
)
SELECT ID, Val, Quantity, Val2, Quantity2, Val3, Quantity3
FROM CTE
WHERE rn = 1
Demo on SQLFiddle

Answer 2

You may use ROW_NUMBER to define the order of the rows within the same ID
with t as (
select ID, VAL, QUANTITY,
row_number() over (partition by ID order by VAL, QUANTITY) as rn
from tab)
select *
ID V QUANTITY RN
---------- - ---------- ----------
1 A 11 1
1 B 15 2
1 B 19 3
1 Z 45 4
2 D 4 1
...
In the next step use PIVOT to get the best three values in one row.
with t as (
select ID, VAL, QUANTITY,
row_number() over (partition by ID order by VAL, QUANTITY) as rn
from tab)
select *
from t
PIVOT
(MAX(VAL) as VAL, MAX(QUANTITY) as QUANTITY FOR RN IN (1 as "COL1" ,2 as "COL2",3 as "COL3")
)
ID C COL1_QUANTITY C COL2_QUANTITY C COL3_QUANTITY
---------- - ------------- - ------------- - -------------
1 A 11 B 15 B 19
2 D 4 E 25 F 13
3 G 10 H 15 I 19
If the standar pivot column naming is not fine, simple add a next query and rename the columns.
Note that this query return the same result as apternative approach base on multiple LEAD columns, but you get a bit better flexibility if you plan to vary the number of traced columns.