Generate Postgres test data with no database

To generate test data, you usually need a database first. That order is backwards. Generation needs one input: the schema. And a schema is just text. If you can paste your CREATE TABLE statements, you can generate data.

Weavori's DDL paste mode reads CREATE TABLE statements from a file or stdin and generates rows from them. No connection string, no server to start, no Docker. Output is a JSON array (default) or CSV on stdout.

Generate from a schema file

Save your schema to schema.sql, then run:

--rows 100 generates 100 rows per table, written to stdout as JSON. Add --format csv for CSV:

Or read the DDL from stdin and pipe it:

What it parses

Paste mode reads CREATE TABLE statements, including schema-qualified names (public.users), IF NOT EXISTS, and inline or table-level PRIMARY KEY, FOREIGN KEY ... REFERENCES, UNIQUE, and CHECK constraints.

Here is a schema you can paste as-is:

CREATE TABLE users (
  id         bigint PRIMARY KEY,
  first_name text NOT NULL,
  last_name  text NOT NULL,
  email      text NOT NULL,
  created_at timestamptz NOT NULL
);
 
CREATE TABLE orders (
  id         bigint PRIMARY KEY,
  user_id    bigint NOT NULL REFERENCES users(id),
  status     text NOT NULL,
  created_at timestamptz NOT NULL
);

Column names steer the values. first_name produces names, email produces emails, status draws from realistic statuses, and created_at produces timestamps. These are the same schema rules the live-database path uses, so the data is shaped by your schema, not by a config file you write.

What comes out

Tables generate in foreign-key order: users before orders, and every orders.user_id references a real generated users.id. With more than one table, CSV labels each block:

--- public.users ---
id,first_name,last_name,email,created_at
1,Maya,Okafor,m.okafor@example.com,2026-02-11T08:14:52Z
2,Daniel,Weiss,d.weiss@example.com,2026-03-02T17:40:19Z

--- public.orders ---
id,user_id,status,created_at
1,1,paid,2026-03-05T10:22:01Z
2,1,shipped,2026-03-19T09:03:44Z

When you want this

• Before the database exists. Design the schema first, look at the data, and fix names and types while they are still cheap to change.
• Fixtures without a container. Generate JSON or CSV in the pipeline, then load it into the database your tests already run against.
• Demos and screenshots. You need realistic rows on a screen, not a provisioned environment.

Write into a database instead

If you do have a Postgres instance, skip stdout and stream the rows in:

--create-tables is on by default, so Weavori creates the tables in the target first, then writes the rows with COPY.

Compared to writing the INSERTs

The alternative is a seed script you maintain by hand. Both work. They differ in who keeps the foreign keys straight:

Try it

Paste any CREATE TABLE block you already have. A snippet from a design doc, a migration file, or the schema section of a ticket is enough.

• Generate command reference
• DDL paste mode and the end-to-end workflow
• How to generate realistic test data for PostgreSQL
• Postgres test data without seed scripts