PostgreSQL identity columns were introduced in version 10. They aim to replace the commonly used SERIAL type, as it not compliant with the SQL standard - and is error prone, as you have to manage the sequence associated to. An identity column is basically a generated integer type with a sequence built-in it knows everything about (it won't pick nextval if nextval already exists).

The bug:

• create table foo (id int primary key generated always as identity);
• Try to add an item in the foo
• Error: column id requires a value