如何在 SQL 服务器表中插入

Question

I want to insert data into table tabA. I get this error

Violation of UNIQUE KEY constraint, Cannot insert duplicate key in object.

The key is ( no_command, no_document, id_transc, project_cd ).

Here is a sample of data to insert:

load_number | load_date | no_command | no_document | id_transc | div1 | div2 | activity_cd | project_cd
1             2021-04-10   38                          -1         0     200                  1
1             2021-04-10   38                          -1         600    0                   1

It's not the same data because div1 and div2 are different. I did a group by to solve the problem but it doesn't work. I have this error in the group by.

Invalid column name

Here is my code

INSERT INTO tabA ([load_number], [load_date], [no_command], [no_document],
                  [id_transc], [div1], [div2], [activity_cd], [project_cd])
    SELECT 
        1 [load_number], GETDATE() [load_date], id_cmd [no_command], '' [no_document],
        COALESCE(c.id_numb, -1) [id_transc],
        CASE
            WHEN b.data_type = 'primary' THEN b.amount
            ELSE 0
        END as [div1],
        CASE
            WHEN b.data_type = 'secondary' THEN b.amount
            ELSE 0
        END as [div2],
        COALESCE(d.budget_cd, -1) [activity_cd], code [project_cd]
    FROM
        tabB b
    LEFT JOIN 
        tabC c ON c.credit = b.account
    LEFT JOIN 
        tabD d ON d.activity = SUBSTRING([b.name], CHARINDEX('-', [b.name]) + 1, LEN([b.name])) 
               AND d.transc = '1010'
     GROUP BY
         [load_number], [load_date], [no_command], [no_document],
         [id_transc], [activity_cd], [project_cd]

I was not the one who designed the database and I cannot make changes to it.

Need help please...

Answer 1

You can add a where clause with not exists :

where not exists (select 1
                  from tabA a
                  where a.no_command = ?.id_cmd and
                        a.no_document = '' and
                        a.id_transc = coalesce(c.id_numb,-1) and
                        a.project_cd = ?.code
                 )

The ? are for the table aliases to define where the columns come from. As a best practice, you should qualify all column references in all your queries. This is particularly important when using correlated subqueries.

Answer 2

(writing here as it would be a mess). Assuming that output is right and you simply want to aggregate those rows:

WITH
  myData AS (
              SELECT
                       id_cmd                                                          [no_command]
                     , COALESCE(c.id_numb, -1)                                         [id_transc]
                     , SUM(CASE WHEN b.data_type='primary' THEN b.amount ELSE 0 END)   AS [div1]
                     , SUM(CASE WHEN b.data_type='secondary' THEN b.amount ELSE 0 END) AS [div2]
                     , COALESCE(d.budget_cd, -1)                                       [activity_cd]
                     , code                                                            [project_cd]
              FROM     tabB           b
                       LEFT JOIN tabC c ON c.credit=b.account
                       LEFT JOIN tabD d ON d.activity=SUBSTRING(
                                                           [b.name], CHARINDEX('-', [b.name])+1
                                                           , LEN([b.name])
                                                           )
                                       AND d.transc='1010'
              GROUP BY id_cmd, COALESCE(c.id_numb, -1), COALESCE(d.budget_cd, -1), [code]
            )
INSERT INTO tabA([load_number], [load_date], [no_command], [no_document], [id_transc], [div1]
               , [div2], [activity_cd], [project_cd])
SELECT
     1, GETDATE(), [no_command], '', [id_transc], [div1], [div2], [activity_cd], [project_cd]
FROM myData;

Please note that an aggregation like SUM() might be questionable when there are joins. If those joins are meant to be 1-to-Many then it is OK, else be careful and do the summing of tabB separately before the joins.

Answer 3

The shared code errors because the data violates the primary key of the table into which you are trying to insert it. You have two records which while not duplicates of each other (they have different values for Div1 and Div2 as you say), they do share the same values for all the primary key fields. So either

1. The table needs changing to allow these two rows, or
2. One or both of these rows should be rejected, or
3. The the two rows should somehow be collapsed into a single row

I'm guessing the third option is correct and that the way to do this is to SUM the values which are returned in Div1 and Div2 . Below is my best guess as to what you want:

SELECT
   1 [load_number],
   GETDATE() [load_date],
   id_cmd [no_command],
   '' [no_document],
   COALESCE(c.id_numb, -1) [id_transc],
   SUM(CASE WHEN b.data_type = 'primary' THEN b.amount ELSE 0 END) as [div1],
   SUM(CASE WHEN b.data_type = 'secondary' THEN b.amount ELSE 0 END) as [div2],
   COALESCE(d.budget_cd, -1) [activity_cd],
   code [project_cd]
FROM
   tabB b
   LEFT JOIN tabC c ON c.credit = b.account
   LEFT JOIN tabD d ON d.activity = SUBSTRING([b.name], CHARINDEX('-', [b.name]) + 1, LEN([b.name])) AND d.transc = '1010'
GROUP BY
   --Each of the things selected (which aren't static, getdate is static for these purposes) but aren't within aggregate functions appears below.
   --Note that they are not referenced by the alias given in the select clause.
   id_cmd,
   COALESCE(c.id_numb, -1),
   COALESCE(d.budget_cd, -1),
   code

It's hard to tell without knowing the your data, but even when doing this it looks like you could run into issues. This is because we are grouping by fields which aren't part of your primary key for the table, namely the [activity_cd] field (or COALESCE(d.budget_cd,-1) ). If it's possible for two records to agree in the primary key fields while differing in this one, then such data would result in errors similar to those you've encountered already.

As GordonLinoff mentions, it is good practice to give make sure that you fully qualify the fields which are referenced in the query. You've done this in places ( c.id_numb ) but not others ( id_cmd ). This helps make code easier to understand and may prevent the code producing errors in the event of changes to table schemas causing field names to become ambiguous.