Add foreign key constraint to an existing table column, referencing a column in a newly created table

Question

Let's say I have one table defined as this

CREATE TABLE sales (
 agent_id integer references agents(agent_id),
 sale_date date,
 amout numeric(10,2)
);

Then an ETL process fills this table with data

INSERT INTO sales VALUES
 (1, '2019-01-01', 1031.11),
 (1, '2019-01-02', 525.44),
 (1, '2019-01-03', 323.99);

But later, I need to add a table with actual processed dates, such as

CREATE TABLE dates (
 date_idx date primary key,
 year integer not null,
 month integer not null,
 day integer not null
);

and want to add a foreign key constraint on the old table:

ALTER TABLE sales
 ADD CONSTRAINT sales_date_fk FOREIGN KEY (sale_date)
 REFERENCES dates (date_idx);

Naturally, I get the following error:

ERROR: insert or update on table "sales" violates foreign key constraint "sales_date_fk" Detail: Key (sale_date)=(2019年01月01日) is not present in table "dates".

I know I can work around this by deleting/truncating all data in sales or prefilling dates before adding the constraint, but I would prefer not to if I can avoid it.

Is there any way to accomplish this?

I'm using Postgres 11, but this fiddle running 9.6 shows what I've tried to explain here.

Answer 1

You can use the NOT VALID modifier when creating a foreign key to stop it verifying existing data. For instance:

ALTER TABLE sales
ADD CONSTRAINT sales_date_fk FOREIGN KEY (sale_date)
REFERENCES dates (date_idx) NOT VALID;

I would not recommend this though, unless something prevents you from populating the new child table so you have no choice.

This is postgres specific syntax, for instance the equivalent in SQL Server is WITH NOCHECK, and some others may not support the option at all, so take care if trying to support multiple database backends.

Answer 2

Select all distinct date from sales tables and insert data into the dates tables. After populating data into that table try to create the foreign key constraint.