Hi folks, I am experiencing an interesting behavior in PostgreSQL and would like to seek some clarification.
In the following snippet, I first add a column with a default value, then drop that default. However, when I query the table, the column still retains the dropped default for existing rows: SET client_min_messages=debug1; DROP TABLE IF EXISTS foo CASCADE; CREATE TABLE foo (id SERIAL PRIMARY KEY); INSERT INTO foo (id) SELECT generate_series(1, 10000); ALTER TABLE foo ADD COLUMN bar varchar(255) NOT NULL DEFAULT 'default'; ALTER TABLE foo ALTER COLUMN bar DROP DEFAULT; SELECT * from foo order by id desc limit 5; -- id | bar -- -------+--------- -- 10000 | default -- 9999 | default -- 9998 | default -- 9997 | default -- 9996 | default In this example, even after dropping the default value from the bar column, the rows that were previously inserted (prior to dropping the default) still show 'default' as their value in the bar column. It does not see that the table has been rewritten or rescanned, otherwise the debug1 messages would be triggered. Can anyone explain how PostgreSQL "knows about" the default value that has just been dropped and what is happened under the scenes? I am keen on a deep understanding on how Postgres achieves this. Here is what I could find in the docs, but it does not satisfy my question: > From PostgreSQL 11, adding a column with a constant default value no longer > means that each row of the table needs to be updated when the ALTER TABLE > statement is executed. Instead, the default value will be returned the next > time the row is accessed, and applied when the table is rewritten, making the > ALTER TABLE very fast even on large tables. https://www.postgresql.org/docs/current/ddl-alter.html#DDL-ALTER-ADDING-A-COLUMN
