SQL deleting records with group by multiple tables

Question

I am trying to delete duplicate records in a table but on if they are duplicate per a record from another.

The following query gets me the number of duplicate records per 'bodyshop'. Im trying to delete multiple invoices for each bodyshop.

SELECT
    inv.InvoiceNo, job.BodyshopId, COUNT(*)
FROM
    [Test].[dbo].[Invoices] as inv
  join [Test].[dbo].Repairs as rep on rep.Id = inv.RepairId
  join [Test].[dbo].Jobs as job on job.Id = rep.JobsId
GROUP BY
    inv.InvoiceNo, job.BodyshopId
HAVING 
    COUNT(*) > 1

I want the duplicate invoice numbers per bodyshop to be deleted but i do want the original one to remain.

InvoiceNo   BodyshopId  (No column name)
29737        16          2
29987        16          3
30059        16          2
23491        139         2
23608        139         3
23867        139         4
23952        139         3

I only want invoice number 29737 to be once against bodyshopid 16 etc.

Hope that makes sense

Thanks

Answer 1

You may run the following as two records are same so, Group by will return single row for same invoice:

DELETE FROM inv where id not in (
SELECT Max(inv.id) FROM (
SELECT
    inv.id, inv.InvoiceNo, job.BodyshopId, COUNT(*)
FROM
    [Test].[dbo].[Invoices] as inv
  join [Test].[dbo].Repairs as rep on rep.Id = inv.RepairId
  join [Test].[dbo].Jobs as job on job.Id = rep.JobsId
GROUP BY
    inv.InvoiceNo, job.BodyshopId
HAVING 
    COUNT(*) > 1
) TMP_TABLE )

id is the primary key.

General SQL. Modify if needed for sql-server.

Answer 2

Perhaps this :

with cte as (
SELECT
    inv.ID, inv.InvoiceNo, job.BodyshopId, rn = row_number() over (partition by inv.InvoiceNo, job.BodyshopId order by inv.InvoiceNo, job.BodyshopId)
FROM
    [Test].[dbo].[Invoices] as inv
  join [Test].[dbo].Repairs as rep on rep.Id = inv.RepairId
  join [Test].[dbo].Jobs as job on job.Id = rep.JobsId
)

delete t1
from [Test].[dbo].[Invoices] t1 inner join cte t2 on t1.ID = t2.ID 
where t2.rn > 1

Edit 1 - Your comments are trues. So a solution is to add an identity column to the invoice table. I've adapt my query.

To add / remove an identity column :

 alter table [Test].[dbo].[Invoices]  id int identity(1,1)
 drop  column id