Skip to content

Core Concepts

ravikiranpagidi edited this page Jun 17, 2026 · 1 revision

Core Concepts

Domain Pack

A domain pack is a prebuilt enterprise dataset model such as banking, ecommerce, healthcare, telecom, or SaaS. It defines tables, columns, primary keys, foreign keys, distributions, and business behavior.

Scale Profile

A scale profile controls row counts with simple names: tiny, small, medium, and large. You can override row counts when you need a specific benchmark or test shape.

Relationship Graph

The relationship graph orders tables so parent tables are generated before child tables. This is how Great Generator avoids orphan foreign keys by default.

Realistic Value Generator

The realistic value layer creates names, emails, phones, addresses, companies, merchants, products, statuses, and domain-specific reference values. It improves descriptive columns while preserving keys and relationships.

CDC Layer

CDC simulation creates insert/update/delete-style records for pipeline testing, including event timestamps, ingestion timestamps, sequence numbers, late-arrival flags, and duplicate flags.

Anomaly Layer

The anomaly layer injects controlled data quality problems such as nulls, duplicates, invalid values, outliers, late-arriving records, and broken references. It is opt-in.

Pandas Engine

The pandas engine is ideal for local notebooks, unit tests, demos, CSV/JSON/Parquet files, and small-to-medium datasets.

Spark Engine

The Spark engine returns PySpark DataFrames and is designed for Databricks, lakehouse demos, Parquet/Delta outputs, and distributed generation.

Export Layer

Export helpers write table-per-folder outputs in CSV, JSON, Parquet, or Delta. Users can also write returned DataFrames using native pandas or Spark APIs.

Seed/Reproducibility

A seed makes generation deterministic. This is useful for tests, reproducible tutorials, demos, and research.

Clone this wiki locally