Custom Rules

Custom Rules

Use custom_rules when a column's business meaning or valid range cannot be inferred reliably from its name and type.

from great_generator import generate_from_schema

schema = {
    "customer_id": "string",
    "customer_name": "string",
    "age": "int",
    "salary": "double",
    "status": "string",
    "created_at": "datetime",
    "campaign_code": "string",
}

df = generate_from_schema(
    schema,
    rows=5000,
    custom_rules={
        "customer_id": {"prefix": "CUST"},
        "customer_name": {"type": "full_name"},
        "age": {"min": 25, "max": 65},
        "salary": {"min": 60000, "max": 180000},
        "status": {"weighted_values": {"Active": 0.8, "Inactive": 0.2}},
        "created_at": {"start": "2023-01-01", "end": "2024-12-31"},
        "campaign_code": {"pattern": "CMP-{index:06d}"},
    },
)

Supported rule keys

Key	Purpose
type	Semantic override
min, max	Numeric or age bounds
values	Allowed categories
weighted_values	Weighted categorical output
prefix	ID prefix
pattern	Format string with {index}
start, end	Date window
null_rate	Null rate for non-ID fields
unique	Validation expectation; IDs are unique by default

Inline rich schema metadata is planned. Today, keep the dtype schema and custom rules separate.