Snowflake: Sequence Support

Atlas now supports Snowflake sequences as first-class schema objects. Define them in HCL, and Atlas plans the right CREATE, ALTER, and DROP SEQUENCE statements. Tables, views, and tasks can declare sequence dependencies via depends_on.

Sequences in Snowflake generate unique, monotonically increasing (or decreasing) integer values, commonly used for surrogate keys and ordered ID columns. Atlas now manages sequences as first-class resources alongside your tables, views, and tasks.

Defining a Sequence

Use the sequence block to declare a sequence in your HCL schema. The schema attribute is required; start, increment, order, and comment are optional:

schema "sales" {}

sequence "customer_id_seq" {
  schema    = schema.sales
  start     = 1
  increment = 1
  comment   = "Surrogate key for customer records"
}

Using a Sequence as a Column Default

A typical pattern is to assign the next sequence value as a column default so new rows get unique IDs automatically. Declare depends_on on the table so Atlas creates the sequence before the table during migration:

table "customers" {
  schema = schema.sales
  column "id" {
    type    = NUMBER
    default = sql("customer_id_seq.NEXTVAL")
  }
  column "email" {
    type = VARCHAR(255)
  }
  depends_on = [sequence.customer_id_seq]
}

Generated SQL

Atlas generates CREATE SEQUENCE, ALTER SEQUENCE, and DROP SEQUENCE statements based on the diff between your desired and current state:

-- Create sequence "customer_id_seq"
CREATE SEQUENCE "sales"."customer_id_seq"
  START = 1
  INCREMENT = 1
  NOORDER
  COMMENT = 'Surrogate key for customer records';

Changing increment or order is handled with ALTER SEQUENCE SET:

-- Increase batch size for bulk loads
ALTER SEQUENCE "sales"."customer_id_seq" SET INCREMENT = 100;

-- Changing start requires a recreate (Snowflake limitation)
DROP SEQUENCE "sales"."customer_id_seq";
CREATE SEQUENCE "sales"."customer_id_seq" START = 1000001 INCREMENT = 1 NOORDER;