How to return ID after inserting into table?

Rob8111

Hi, I'm inserting some data into table1 from table2 using the query below inside plsql.

insert into table1 (

col1

, col2

, clo3

)

select

col1

, col2

, col3

from table2;

I want to return the ID of newly inserted row from table1.

I tried,

'RETURNING ID INTO VARIBLAE'. But this is not working in 'insert into select' query.

Is there any other way?

BluShadow

Unfortunately the RETURNING clause doesn't work directly with an INSERT ... SELECT ..., but instead works with an INSERT ... VALUES ... (also consider that your SELECT statement may query more than 1 row, so selecting a single ID to a variable isn't usually going to work anyway)

There's a workaround where you can collect the values to be inserted from the select into a collection type and then use a FORALL PL/SQL statement with an INSERT ... VALUES ... RETURNING BULK COLLECT INTO <collection type> which then gives you a collection with all the 'ID' values you want to capture.

So what exactly are you trying to achieve? Inserting a single row of data and capture the ID? or Insert multiple rows and capture all the ID's?

BluShadow

e.g.

SQL> create table t(id number, text varchar2(20))
  2  /

Table created.

SQL> create sequence tSeq
  2  /

Sequence created.

SQL> create trigger tT before insert on t for each row
  2  begin
  3    :new.id := tSeq.nextval;
  4  end;
  5  /

Trigger created.

SQL> create type tIds as table of number;
  2  /

Type created.

SQL> create type tNames as table of varchar2(20)
  2  /

Type created.

SQL> declare
  2    ids   tIds;
  3    names tNames;
  4  begin
  5    select ename
  6    bulk collect into names
  7    from   emp;
  8    --
  9    forall n in 1 .. names.count
 10      insert into t(text)
 11      values (names(n))
 12      returning id bulk collect into ids;
 13    --
 14    for i in 1 .. ids.count
 15    loop
 16      dbms_output.put_line('Inserted '|| to_char(ids(i)));
 17    end loop;
 18  end;
 19  /
Inserted 1
Inserted 2
Inserted 3
Inserted 4
Inserted 5
Inserted 6
Inserted 7
Inserted 8
Inserted 9
Inserted 10
Inserted 11
Inserted 12
Inserted 13
Inserted 14

PL/SQL procedure successfully completed.

Obviously this uses more PGA memory as you're collecting things in memory rather than doing stuff directly with SQL.

Rob8111

Hi @BluShadow, actually I want to insert a single row and return its id.

Alex Nuijten

Then use

INSERT INTO <table> VALUES (<values>) RETURNING ID into <variable>

Just like @BluShadow already mentioned.

BluShadow

Ok, so the key thing is...

Inserting single values and getting the ID is easily done with INSERT ... VALUES ... RETURNING n into var

Inserting from a SELECT and getting the ID or multiple ID's requires the workaround.

Oracle just doesn't support the RETURNING clause for INSERT ... SELECT ...

If you were just inserting a single column you could workaround like...

declare
  id number;
begin
  insert into t(text)
    values ((select ename from emp where empno = 7788))
    returning id into id;
  dbms_output.put_line('Inserted ' || to_char(id));
end;
/

... but if you're inserting multiple columns then it becomes more messy doing it that way...

SQL> create table t(id number, ename varchar2(20), deptno number)
  2  /

Table created.

SQL> create sequence tSeq
  2  /

Sequence created.

SQL> create trigger tT before insert on t for each row
  2  begin
  3    :new.id := tSeq.nextval;
  4  end;
  5  /

Trigger created.

SQL> declare
  2    id number;
  3  begin
  4    insert into t(ename, deptno)
  5      values ((select ename from emp where empno = 7788), (select deptno from emp where empno = 7788))
  6      returning id into id;
  7    dbms_output.put_line('Inserted ' || to_char(id));
  8  end;
  9  /
Inserted 1

PL/SQL procedure successfully completed.

I'm sure there must be some other funky workaround to achieve it...

Rob8111

HI @Alex Nuijten,

INSERT INTO <table> VALUES (<values>) RETURNING ID into <variable>

this will not work with INSERT INTO <table1> SELECT FROM <table2>.

BluShadow

https://community.oracle.com/tech/developers/discussion/comment/16834960#Comment_16834960

Correct. As already discussed. I'm sure Alex already knows that too.

It's the way Oracle is.

mathguy

Instead of using multiple scalar subqueries (and scanning the source data multiple times), and also instead of creating a record type, it seems more natural to create local scalar variables for the columns you are copying from table2 to table1.

Then select the columns from table2 into these local variables (a SELECT ... INTO ... FROM table2 WHERE ... statement), and use the variables in an INSERT INTO table1 (...) VALUES ( ... ) statement. Something like this:

create table t(id number generated always as identity, ename varchar2(20), deptno number)
/

Table T created.



declare
  l_ename  t.ename%type;
  l_deptno t.deptno%type;
  l_id     number;
begin
  select  ename, deptno
    into  l_ename, l_deptno
    from  scott.emp
    where empno = 7788;
  insert into t (ename, deptno) values (l_ename, l_deptno) returning id into l_id;
  dbms_output.put_line ('Newly generated id:  ' || l_id);
end;
/

Newly generated id:  1


PL/SQL procedure successfully completed.




select * from t;

   ID ENAME                    DEPTNO
----- -------------------- ----------
    1 SCOTT                        20