SQL Server Switch Based on grouping of previous values - sql

Question

I have a table of order information in the following format:
╔══════════════╦══════╦════════════════╦═══════════╦═════════╦══════════╗
║ Order Number ║ Line ║ Item ║ Warehouse ║ Carrier ║ Quantity ║
╠══════════════╬══════╬════════════════╬═══════════╬═════════╬══════════╣
║ 255 ║ 1 ║ STUFFED-ANIMAL ║ WH1 ║ UPS ║ 3 ║
║ 256 ║ 1 ║ BLOCKS ║ WH2 ║ FEDEX ║ 1 ║
║ 257 ║ 1 ║ DOLL ║ WH1 ║ UPS ║ 1 ║
║ 257 ║ 2 ║ DRESS ║ WH1 ║ UPS ║ 3 ║
║ 257 ║ 3 ║ SHOES ║ WH2 ║ UPS ║ 1 ║
║ 258 ║ 1 ║ CHAIR ║ WH3 ║ FEDEX ║ 1 ║
║ 258 ║ 2 ║ CHAIR ║ WH3 ║ UPS ║ 2 ║
╚══════════════╩══════╩════════════════╩═══════════╩═════════╩══════════╝
I am trying to query it in such a way that I partition it into groups based on a unique combination of columns.
In my example, I would like the following result:
╔════════════════╦══════╦════════════════╦═══════════╦═════════╦══════════╗
║ Package-Number ║ Line ║ Item ║ Warehouse ║ Carrier ║ Quantity ║
╠════════════════╬══════╬════════════════╬═══════════╬═════════╬══════════╣
║ 255 ║ 1 ║ STUFFED-ANIMAL ║ WH1 ║ UPS ║ 3 ║
║ 256 ║ 1 ║ BLOCKS ║ WH2 ║ FEDEX ║ 1 ║
║ 257-1 ║ 1 ║ DOLL ║ WH1 ║ UPS ║ 1 ║
║ 257-1 ║ 2 ║ DRESS ║ WH1 ║ UPS ║ 3 ║
║ 257-2 ║ 3 ║ SHOES ║ WH2 ║ UPS ║ 1 ║
║ 258-1 ║ 1 ║ CHAIR ║ WH3 ║ FEDEX ║ 1 ║
║ 258-2 ║ 2 ║ CHAIR ║ WH3 ║ UPS ║ 2 ║
╚════════════════╩══════╩════════════════╩═══════════╩═════════╩══════════╝
To break it down I would like to do the following:
If the order number, warehouse, and carrier are the same that is one 'partition.' If there is only one partition then we just leave the order number as the package number, otherwise we break it down into packages. These packages are numbered by the same set of values being the same, but now have a number denoting which package it is.
I was looking into using row_number() over (partition by... that I found after searching for similar issues but I don't think it is exactly what I'm looking for.
Could someone point me in the right direction?

Answer 1

This is tricky. Having count(distinct) as a window function would help. But there is a convenient trick using the sums of dense_rank()s.
So, I think this does what you want:
select (case when seqnum_asc + seqnum_desc - 1 > 1 -- more than 1 distinct value
then concat(ordernumber, '-', seqnum_asc)
else concat(ordernumber, '') -- just to convert the value to a string
end) as packagenumber,
t.*
FROM (select t.*,
dense_rank() over (partition by ordernumber order by warehouse, carrier) as seqnum_asc,
dense_rank() over (partition by ordernumber order by warehouse desc, carrier desc) as seqnum_desc
from mytable t
) t;
Here is a db<>fiddle.
Note: This does not take into account the ordering by the line number -- because your question doesn't mention that at all. If you only want adjacent rows with the same value to be included in each group, then ask a new question with appropriate sample data and desired results.

Answer 2

Here is an option using Dense_Rank() instead of Row_Number()
Example
Select [Package-Number] = concat([Order Number]
,left(nullif(count(*) over (partition by [Order Number] ),1),0)
+dense_rank() over (partition by [Order Number],warehouse,carrier order by Line)*-1 )
,Line
,Item
,Warehouse
,Carrier
,Quantity
From YourTable
Returns