sql: insert object spread over multiple tables - sql

Question

There are already posts, for example this one, which state that "naive" inheritance in SQL, namely using one table per class level, is a common thing. Example
create table parent
( id integer primary key
, more-parent-attributes
);
create table child
( id integer primary key references parent(id) on delete cascade
, more-child-attributes
);
My question is only how to insert a child in an idiomatic ANSI SQL way
into the table. The foreign key constraint makes the requirement that we first
insert a new row into parent and then a new row into child, using the id
of the parent row. I don't know how to do this (get this id) safely and portably, and
using only one request.
Hint: I'm rather a beginner and don't know imperative SQL programming--just in
case there is an obvious imperative solution.

Answer 1

You must execute two insert.
The first insert add row in parent table, the second insert add row in the child table.
Two insert operations can be grouped in the same transaction.
To get the correct inserted id in the parent table you must get a select id from parent.
Show below:
Step 1:
INSERT INTO parent (id, more att) values (your ID, other values)
Pay attention about ID value, you can use newid() (Sql server) uuid() (mySql) or autoincremental integer field
Step 2:
You retrieve your key querying your parent table with a functional key.
SELECT id FROM parent where functional_key satisfacted
For example, if I store in my parent table a list of employes, a functional key can be register number.
So your query becomes:
SELECT id FROM parent WHERE register_no = 'YOUR_REGISTER_NUMBER'
Step 3:
INSERT INTO child (id, fk_parent, other fields) values(id, fk_parent, other fields)
The fk_parent field must be valued with the result of Step 2.
In this step you can:
value fk_parent with a variable or you can use a subquery (step 2) in your insert statement.

Answer 2

I ended up doing something similar. You need to have some identifying piece of data that you can insert into the Parent in order to get the Id. If you're using this in some kind of application then you can use a GUID. In my application I used a concatenation of source columns that I knew would produce a unique value.
CREATE TABLE Parent
(
Id INT IDENTITY NOT NULL PRIMARY KEY
,SourceId VARCHAR(50) NOT NULL
);
CREATE TABLE Child
(
ParentId INT NOT NULL REFERENCES Parent (Id)
,Data VARCHAR(20)
);
-- Some procedure inserts the unique value
INSERT INTO Parent (SourceId) VALUES ('UNIQUE VALUE');
-- Another procedure inserts data using the unique value
DECLARE #Id INT;
SELECT #Id = Id FROM Parent WHERE SourceId = 'UNIQUE VALUE';
INSERT INTO Child (ParentId, Data) VALUES (#Id, 'Some Data');

Answer 3

Scope_Identity() is what you looking for:
DECLARE #Id INT
INSERT INTO parent (more-parent-attributes) values (.....)
SET #Id = Scope_Identity()
INSERT INTO child (parent(id), more-child-attributes) SELECT #Id, ....more-child-attributes
Scope_Identity() returns identity column in the same scope. It means that Parent key should be Identity column:
id int IDENTITY(1,1)PRIMARY KEY
and I think this is the case as if you were deciding what is the Parent key id, you would use the same for child insert.