INSERT到另一个表中的表和UPDATE外键（Sql Server 2008）

Question

I have two tables, TableA and TableB like this:

TableA
-------------------------------------------
|  id  |  some_data  | new_FK_column_on_B |
| ---- | ----------- | ------------------ |
|   1  |     ...     |        null        |
|  ... |     ...     |        null        |
|  999 |     ...     |        null        |
-------------------------------------------

TableB
----------------------------
|  id  |  some_other_data  |
| ---- | ----------------- |
|      |                   |
----------------------------

At the moment, TableB is empty, and FK column in TableA is null for all rows. I need to write one-time initializing scrit to populate TableB and initialize FK column for some rows (criterial, not for all) in TableA by identifiers from rows, inserted in TableB .

I know two ways to do this:

1) using while and scope_identity() , inserting new row into TableB and updating TableA on each iteration, while exists rows in TableA , which should be updated

while (exists (select 1 from TableA where [condition]))
begin
    insert into TableB (some_other_data) values ('some_other_data')

    update TableA set new_FK_column_on_B
    where id = (select top 1 id from TableA where [condition])
end

2) create temp column in TableB , storing id of row in TableA , for which it was inserted, and then update TableA using join

alter table TableB add temp int
go

insert into TableB (some_other_data, temp) select 'some_other_data', id from TableA where [condition]

update TableA
set new_FK_column_on_B = b.id
from TableB as b
join TableA as a on a.id = b.temp

alter table TableB drop column temp

Also I was trying to use somehow output from insert like this, but it's syntax is incorrect:

update TableA
set new_FK_column_on_B =
(
    select insertedId from 
    (
        insert into TableB (some_other_data)
        output inserter.id as insertedId
        values ('some_other_data')
    )
)
where [condition]

Is there any easier way to do this whithout using while or modifing any table?

Answer 1

I found this question when searching for a solution a similar case. The only difference was that I wanted to fill TableB with an row for each row in TableA (no where-clause).

I found a solution to the third option you suggeted (using output from insert), however, since you are using INSERT with data from a SELECT, you cannot use the OUTPUT clause. This pointed me in the right direction:

DECLARE @MyTableVar table(tableBId int, tableAId int)

MERGE INTO TableB
using TableA AS AP
on 1=0
WHEN NOT MATCHED THEN
   Insert(some_other_data) 
           Values('some_other_data')
           Output inserted.ID, AP.ID INTO @MyTableVar;

update TableA set new_FK_column_on_B = (select tableBId from @MyTableVar where tableAId = TableA.ID)

Be aware that executing this a seconds time will create new entries in TableB. If you only want to create new rows in TableB, where there is no foreign key set in TableA, you can use this script:

DECLARE @MyTableVar TABLE(tableBId int, tableAId int)

MERGE INTO TableB AS B
USING TableA AS AP
ON A.new_FK_column_on_B = B.id
WHEN NOT MATCHED THEN
   INSERT(some_data) 
           VALUES(AP.some_data)
           OUTPUT inserted.ID, AP.ID INTO @MyTableVar;

UPDATE TableA SET new_FK_column_on_B = (
   SELECT tableBId
   FROM @MyTableVar
   WHERE tableAId = TableA.ID )
WHERE TableA.new_FK_column_on_B IS NULL;

Answer 2

You can do all this as set operations:

insert into b(some_data)
    select distinct some_data
    from a;

update a
    set new_FK_column_on_B = b.id
    from a join
         b
         on a.some_data = b.some_data;

This assumes that the id column in b is declared as identity() , so it gets assigned automatically. This is a good idea, but if you want to do this manually, then the first query would be something like:

insert into b(some_data)
    select row_number() over (order by (select null)), some_data
    from (select distinct some_data
          from a
         ) a;

There is no need for a while loop.

Answer 3

DISCLAIMER: This is not tested at all, I wrote this in notepad:

DECLARE @TableAValues TABLE (IdTableA int, SomeData varchar)

INSERT INTO @TableAValues(IdTableA, SomeData)
    SELECT id, 'some_data' FROM TableA      

DECLARE @TableBIds TABLE (IdTableB int)

INSERT INTO TableB(SomeData)
OUTPUT INSERTED.ID INTO @TableBIds
SELECT SomeData FROM @TableAValues

UPDATE ta
SET ta.new_FK_column_on_B = tbi.IdTableB
FROM dbo.TableA AS ta
INNER JOIN @TableAValues AS tav ON ta.id = tav.IdTableA -- used in case more records were added to table in the interim.
LEFT OUTER JOIN @TableBIds tbi On tav.RowNum = tbi.RowNum

Note: I am using In memory tables, but if you were concerned about memory usage, you could probably just switch those out for temp tables on disk.

The idea I was going for here:

1. Grab the rows from table A (ID + data) that we will use to populate B and store them (I am using an In memory table)
2. Insert those rows into B and store the corresponding ID of B (again in an in memory table)
3. I assume the order of rows in both in memory tables will now match, so the idea is that we can join both in memory tables together on row number to get Table A Id with it's Table B Id - which we use to update table A's foreign key.

I'd be very surprised if my code sample above works as is, but hopefully the idea will be useful, if not my execution.