How to compute a derived age attribute in postgresql? - postgresql

Question

I have the following person table with the dbirth column (date of birth). I need to compute the column age. I has been trying, but I have the following ERROR: generation expression is not immutable. I would greatly appreciate any help.
CREATE TABLE person
(
person_id SERIAL NOT NULL,
fname VARCHAR(50) NOT NULL,
lname VARCHAR(50) NOT NULL,
ssn CHAR(10) NOT NULL,
pnumber CHAR(12) NOT NULL,
dbirth DATE NOT NULL,
age integer GENERATED ALWAYS AS ( extract( year FROM CURRENT_DATE ) - extract( year FROM dbirth)) STORED,
address_id INTEGER NOT NULL,
sex_id INTEGER NOT NULL,
PRIMARY KEY ( person_id ),
FOREIGN KEY ( sex_id ) REFERENCES sex ( sex_id ),
FOREIGN KEY ( address_id ) REFERENCES address ( address_id )
);

Answer 1

The reason is EXTRACT function is not an IMMUTABLE.
Create your own custom IMMUTABLE function and use it in the generated column.
Example: Function to get_age as an Interval contains Year, Month and days
CREATE OR REPLACE FUNCTION get_age( birthday date )
RETURNS interval
AS $CODE$
BEGIN
RETURN age(birthday);
END
$CODE$
LANGUAGE plpgsql IMMUTABLE;
Then use the IMMUTABLE function in the generated column
CREATE TABLE person
(
person_id SERIAL NOT NULL,
fname VARCHAR(50) NOT NULL,
lname VARCHAR(50) NOT NULL,
ssn CHAR(10) NOT NULL,
pnumber CHAR(12) NOT NULL,
dbirth DATE NOT NULL,
age interval GENERATED ALWAYS AS (get_age(dbirth)) STORED,
address_id INTEGER NOT NULL,
sex_id INTEGER NOT NULL,
PRIMARY KEY ( person_id ),
FOREIGN KEY ( sex_id ) REFERENCES sex ( sex_id ),
FOREIGN KEY ( address_id ) REFERENCES address ( address_id )
);
Note: Age is an Interval datatype if you want to use it as an int modify the get_age function.

Answer 2

You do not want a generated column for age. Postgres only supports STORED generated columns, the problem comes in that the age column is NOT update unless dbirth is updated.
-- setup
create or replace function get_age( from_date timestamptz )
returns interval
language sql immutable
as $$
select age(from_date);
$$;
create table test_person
( person_id serial
, fname varchar(10) not null
, pnumber varchar(12) not null
, dbirth timestamptz(6) not null
, age interval generated always as (get_age(dbirth)) stored
);
insert into test_person(fname,pnumber,dbirth)
values ('MSam', '555-937-9292', '1995-06-23 08:00:01.344612'::timestamptz )
, ('WSam', '555-937-9292', '2019-03-21 15:32:18.863452'::timestamptz );
commit ;
select p.fname, p.pnumber,dbirth, p.age, current_timestamp
from test_person p
order by fname;
|fname |pnumber |dbirth |age |current_timestamp |
|----------|------------|-------------------|----------------------------|-------------------|
|MSam |555-937-9292|1995-06-23 03:00:01|25 years 24 days 15:59:58.65|2020-07-18 17:53:14|
|WSam |555-937-9292|2019-03-21 10:32:18|1 year 3 mons 27 days 08:27:|2020-07-18 17:53:14|
Now wait some time, does not have to be long, but noticeable.
Then run the exact same query.
select p.fname, p.pnumber,dbirth, p.age, current_timestamp
from test_person p
order by fname;
|fname |pnumber |dbirth |age |current_timestamp |
|----------|------------|-------------------|----------------------------|-------------------|
|MSam |555-937-9292|1995-06-23 03:00:01|25 years 24 days 15:59:58.65|2020-07-18 17:55:29|
|WSam |555-937-9292|2019-03-21 10:32:18|1 year 3 mons 27 days 08:27:|2020-07-18 17:55:29|
Notice in the above that although the current timestamp has changes (time has passed) the age has not. Have we just discovered "the Fountain Of Youth" So try a likely update, like phone number.
update test_person
set pnumber = '555-949-0070'
where fname = 'MSam';
select p.fname, p.pnumber,dbirth, p.age, current_timestamp
from test_person p
order by fname;
|fname |pnumber |dbirth |age |current_timestamp |
|----------|------------|-------------------|----------------------------|-------------------|
|MSam |555-949-0070|1995-06-23 03:00:01|25 years 24 days 15:59:58.65|2020-07-18 17:57:16|
|WSam |555-937-9292|2019-03-21 10:32:18|1 year 3 mons 27 days 08:27:|2020-07-18 17:57:16|
Perhaps a better solution would be:
Drop the Age column from the table completely.
Create a view having and have it generate the Age.
Write your queries against the view.
alter table test_person drop column age;
create view test_person_v as
( select t.*, age( current_timestamp, dbirth) age
from test_person t
);
);
select p.fname, p.pnumber,dbirth, p.age, current_timestamp
from test_person_v p
order by fname;
|fname |pnumber |dbirth |age |current_timestamp |
|----------|------------|-------------------|----------------------------|-------------------|
|MSam |555-949-0070|1998-06-23 03:00:01|22 years 25 days 15:37:45.54|2020-07-18 18:37:46|
|WSam |555-937-9292|2019-03-21 10:32:18|1 year 3 mons 28 days 08:05:|2020-07-18 18:37:46|
select p.fname, p.pnumber,dbirth, p.age, current_timestamp
from test_person_v p
order by fname;
|fname |pnumber |dbirth |age |current_timestamp |
|----------|------------|-------------------|----------------------------|-------------------|
|MSam |555-949-0070|1998-06-23 03:00:01|22 years 25 days 15:42:50.78|2020-07-18 18:42:52|
|WSam |555-937-9292|2019-03-21 10:32:18|1 year 3 mons 28 days 08:10:|2020-07-18 18:42:52|