Relationship Aware Data Generation

Generate Related Tables from Your Schemas

After generate_from_schema, use generate_relational when your lower environment needs several connected tables.

from great_generator import generate_relational

data = generate_relational(
    tables={
        "customers": {
            "schema": "customer_id int primary key, customer_name string, email string",
            "rows": 1000,
        },
        "orders": {
            "schema": "order_id int primary key, customer_id int references customers.customer_id, order_amount double, order_date date",
            "rows": 5000,
        },
    },
    engine="pandas",
)

customers_df = data["customers"]
orders_df = data["orders"]

Output model

The result is a dictionary keyed by table name:

{
    "customers": customers_df,
    "orders": orders_df,
}

This makes storage optional and flexible:

customers_df.to_parquet("customers.parquet", index=False)
orders_df.to_parquet("orders.parquet", index=False)

In a Spark notebook, select engine="spark" to receive Spark DataFrames and use normal Spark writers for Delta, Snowflake, Azure SQL, cloud storage, or catalog tables.

Referential integrity by default

Parent tables are generated before child tables. Child foreign keys are selected from valid parent keys, so this relationship remains valid:

orders.customer_id -> customers.customer_id

Orphan keys are not introduced unless an explicit anomaly configuration requests them.

Relationship declarations

Relationships can be declared inline:

customer_id int references customers.customer_id

or separately:

data = generate_relational(
    tables={
        "customers": "customer_id int primary key, customer_name string",
        "orders": "order_id int primary key, customer_id int, order_amount double",
    },
    relationships=["orders.customer_id -> customers.customer_id"],
    rows={"customers": 1000, "orders": 5000},
)

When to use each API

• Use generate_from_schema for one table.
• Use generate_relational for your own connected tables.
• Use generate_domain later for ready-made demonstrations and learning datasets.
• Use dimensional and Data Vault generators later when you specifically need those modeling patterns.