Question

I have a simple query. Something like this:

SELECT l.list_name, COUNT(order_id) 
FROM orders o JOIN lists l ON l.order_id=o.order_id 
    WHERE l.list_name LIKE 'orders_1%' or l.list_name LIKE 'orders_2%'
GROUP BY l.list_name

The situation looks like this: overight a stored procedure is updating the lists table, but it chops lists in parts if there are more than 1000 orders.

If I have 1200 orders with criteria or list ‘orders_1’, then my procedure will create two lists: ‘orders_1_1’ ‘and orders_1_2’, the first having 1,000 and second 200 orders.

So when I run my query to count those orders I will get results like so:

list_name                  count

orders_1                   100
orders_1_more_than_100_1   1000       
orders_1_more_than_100_2   200
orders_2                   400 
orders_3_1                 1000       
orders_3_2                 1000
orders_3_3                 420  
orders_3_more_than_100_1   1000       
orders_3_more_than_100_2   900
orders_3_more_than_200_1   1000       
orders_3_more_than_200_2   1000
orders_3_more_than_200_3   100      
orders_4                   200
orders_4_more_than_300     200 

The result I would like to get should look like this:

list_name                  count

orders_1                   100
orders_1_more_than_100     1200       
orders_2                   400 
orders_3                   2420 
orders_3_more_than_100     1900 
orders_3_more_than_200     2100     
orders_4                   200 
orders_4_more_than_300     200 

So that it will sum all lists that start the same.

Any ideas on that? 🙂

These are the exact values that I have in my list_names column:

WYS_AUT_PISMO_NR_6
WYS_AUT_PISMO_NR_5_POWYZEJ_240
WYS_AUT_PISMO_NR_5_DO_240
WYS_AUT_PISMO_NR_4_POWYZEJ_240_5
WYS_AUT_PISMO_NR_4_POWYZEJ_240_4
WYS_AUT_PISMO_NR_4_POWYZEJ_240_3
WYS_AUT_PISMO_NR_4_POWYZEJ_240_2
WYS_AUT_PISMO_NR_4_POWYZEJ_240_1
WYS_AUT_PISMO_NR_4_DO_240
WYS_AUT_PISMO_NR_3_POWYZEJ_240
WYS_AUT_PISMO_NR_3_DO_240
WYS_AUT_PISMO_NR_2_POWYZEJ_240
WYS_AUT_PISMO_NR_2_DO_240
WYS_AUT_PISMO_NR_1

What I want is to group them like so:

WYS_AUT_PISMO_NR_6
WYS_AUT_PISMO_NR_5_POWYZEJ_240
WYS_AUT_PISMO_NR_5_DO_240
WYS_AUT_PISMO_NR_4_POWYZEJ_240 /*here I must group those 5 lists*/
WYS_AUT_PISMO_NR_4_DO_240
WYS_AUT_PISMO_NR_3_POWYZEJ_240
WYS_AUT_PISMO_NR_3_DO_240
WYS_AUT_PISMO_NR_2_POWYZEJ_240
WYS_AUT_PISMO_NR_2_DO_240
WYS_AUT_PISMO_NR_1

Answer 1

This monstruous expression will isolate string starting from beginning of parameter to last _ if there are more than one underscores:

select case when len (l.list_name) - len (replace (l.list_name, '_', '')) > 1 
            then left(l.list_name, 
                      len (l.list_name) - charindex('_', reverse(l.list_name)))
            else l.list_name
        end

Alternatively you might strip ‘orders_’ from string, replace underscore with dot and convert it to float, then to int to remove decimals, and then back to string using this monstrosity:

select 'orders_' + cast (cast (cast (
        replace (substring (@str, 8, 100), '_', '.') 
        as float) as int) as varchar(100))

To avoid repeating this blobs, use derived table instead of lists:

SELECT l.TrimmedListName, COUNT(order_id) 
FROM orders o 
JOIN 
(
   select lists.*,
       -- Remove optional list continuation number
          case when len (list_name) - len (replace (list_name, '_', '')) > 1 
                then left(list_name, 
                          len (list_name) - charindex('_', reverse(list_name)))
                else list_name
            end AS TrimmedListName
     from lists
) l ON l.order_id=o.order_id 
WHERE (l.list_name LIKE 'orders_1%' or l.list_name LIKE 'orders_2%')
GROUP BY l.TrimmedListName