Question

I have and existing database, which I have migrated with SQLAlchemy to a new PostgreSQL database.

I moved all primary keys with the same values as before. Now I have tables filled with data, but the associated sequences starts from 1. I have pk values stored 1 to 2000.

Now, when I try to save something with Django, I have the

duplicate key value violates unique constraint
regarding to the Primary Key.

How can I modify the sequence start values or escape this situation?
My current solution is:

conn = psycopg2.connect(...)
    for table_name in table_names:

        cursor = conn.cursor()
        cursor.execute("""
        SELECT setval('%s_id_seq', (SELECT COALESCE(MAX(id),0)+1 FROM %s));
        """% (table_name, table_name))

It works for me, but I don’t like it.

Answer 1

Ways to set / reset a sequence in PostgreSQL

(not necessarily to max(id)).

• There’s the simple way you have in the question. You can set the sequence to start at an arbitrary number with setval():

  SELECT setval('tbl_id_seq');
  

• Then there’s the standard SQL way with ALTER SEQUENCE doing the same:

  ALTER SEQUENCE myseq RESTART WITH 1;
  

• If you like to restart your sequences at numbers other than the default 1:

  CREATE TABLE foo(id serial, a text);      -- creates sequence "foo_id_seq"
  INSERT INTO foo(a) VALUES('a');           -- seq. starts with --> 1
  
  ALTER SEQUENCE foo_id_seq START WITH 10;  -- doesn't restart sequence
  INSERT INTO foo(a) VALUES('b');           --> 2
  
  ALTER SEQUENCE foo_id_seq RESTART;        -- restarts sequence
  INSERT INTO foo(a) VALUES('c');           --> 10
  

• And there is another way, when you empty a table with TRUNCATE:

  TRUNCATE foo RESTART IDENTITY;
  

  Implicitly executes ALTER SEQUENCE foo_id_seq RESTART;