Supported Schema Input Types

Supported Schema Input Types

This matrix describes the current code, not aspirational support.

Schema Input Type	Example	Best For	Status
Plain Python mapping	{"name": "string", "age": "int"}	Fast table generation	Supported
Rich inline metadata	{"age": {"type": "int", "min": 18}}	Embedded business rules	Partial: use separate custom_rules today
Pandas dtype mapping	df.dtypes.to_dict()	Pandas and notebooks	Supported
Pandas DataFrame	empty or populated frame	Preserve Pandas dtypes	Supported
Compact DDL	"id int, name string"	SQL-like definitions	Supported
Full SQL DDL	CREATE TABLE ...	Database teams	Planned
PySpark StructType	StructType([...])	Spark and lakehouse platforms	Supported for common scalar fields
PySpark DataFrame	Spark DataFrame	Infer schema and SparkSession	Supported
Great Generator TableSchema	typed object	Extensions	Supported
Great Generator DomainSchema	multi-table metadata	Schema-based related tables	Supported
JSON Schema	object with properties	APIs and contracts	Planned
YAML schema profile	schema file	Reusable configuration	Planned
Column list	["name", "age"]	Fast prototypes	Planned
SQLAlchemy model	ORM class	Backend teams	Planned
Pydantic model	BaseModel	API teams	Planned
Dataclass	Python dataclass	Typed Python	Planned

JSON, TOML, and simple YAML dataset recipes are supported by generate_from_recipe. That is separate from schema input support.

Return behavior

• Python, compact DDL, and Pandas inputs return Pandas by default.
• A Spark context or engine="spark" returns a Spark DataFrame.
• A PySpark DataFrame provides its own SparkSession.
• DomainSchema returns a dictionary of DataFrames.

See the individual pages for examples and limitations.