Question

I have a table which contains the following columns and I am trying to split it as follows. How do I write a migration script for the following transformation in Oracle

source:

create table abc (id pk, col1, col2, col3, col4, col5, col6)

targets:

create table def (id pk, col1, col2)
create table ghi (id pk, def_id fk, value)

Given this starting point …

insert into table abc values (1, 1, 2, 3, 4, 5, 6)
insert into table abc values (2, 7, 8, 9, 10, 11, 12)

… the post translation the data gets mapped as follows

def contains (1, 1, 2)
ghi contains (1, 1, 3)
ghi contains (2, 1, 4)
ghi contains (3, 1, 5)
ghi contains (4, 1, 6)

def contains (2, 7, 8)
ghi contains (5, 2, 9)
ghi contains (6, 2, 10)
ghi contains (7, 2, 11)
ghi contains (8, 2, 12)

The most important restriction is that I want to generate rows in ghi only if the corresponding column values in abc are not null

Answer 1

I think the easiest way to do this is with two selects from your source table. The first one is a cinch:

insert into def
select id, col1, col2 from abc

The second one uses the INSERT ALL syntax:

INSERT ALL
   WHEN col3 is not null THEN
      INTO ghi values (some_seq.nextval, id, col3)
   WHEN col4 is not null THEN
      INTO ghi values (some_seq.nextval, id, col4)
   WHEN col5 is not null THEN
      INTO ghi values (some_seq.nextval, id, col5)
   WHEN col6 is not null THEN
      INTO ghi values (some_seq.nextval, id, col6)
   SELECT id, col3, col4, col5, col6
      FROM abc;

Note, I am presuming a sequence SOME_SEQ as the source of primary key values for GHI. You may want to have some other mechanism.

---

Hmmm, having run my code I get this result:

SQL> select * from ghi;

        ID     DEF_ID      VALUE
---------- ---------- ----------
         1          1          3
         2          2          9
         1          1          4
         2          2         10
         1          1          5
         2          2         11
         1          1          6
         2          2         12

8 rows selected.

SQL> 

As you can see, SOME_SEQ has only incremented for the two rows in ABC not the eight rows in GHI. This makes sense because two calls to NEXTVAL in the same statement return the same value, but I was hoping that each INSERT would count separately.

Oh well. This means for your migration you cannot enforce the primary key on GHI until after you’ve populated it. So

1. disable the primary key (or create the table without one).
2. populate it with data from ABC
3. populate the primary key column (in whatever fashion)
4. enable the primary key constraint

Alternatively you need to find some other way to populate the GHI ID column.