dfgenerator 0.1.1

Deterministic test DataFrame builder with optional PySpark export

dfgenerator

Lightweight helpers to build deterministic test datasets for Pandas and PySpark, with Faker-powered generators, YAML-based reusable configs, and JSON Schema support.

Features

• 🎲 Deterministic data generation with seed support for reproducible tests
• 📊 Multiple input formats: YAML configs, JSON Schema, Spark/Iceberg DataFrames, or programmatic Python API
• 🌍 Multi-locale support with Faker integration (50+ locales)
• 🔗 Relational data with primary/foreign key support for joinable datasets
• 📤 Multiple export formats: CSV, JSON, Pandas DataFrame, PySpark DataFrame
• 🎯 Smart inference from JSON Schema formats, property names, and existing Spark DataFrames
• ⚡ Spark/Iceberg support - Generate test data from production schemas
• 🚀 CLI and Python API for flexible usage

Quick Start

From Spark/Iceberg DataFrame

from pyspark.sql import SparkSession
from dfgenerator import build_from_spark

# Load production Iceberg table
spark = SparkSession.builder.getOrCreate()
prod_df = spark.read.format("iceberg").load("prod.users")

# Generate test data with same schema
test_data = build_from_spark(prod_df, rows=100, seed=42)

# Export for testing
test_data.to_file("test_users.csv")
test_df = test_data.to_spark(spark)

From JSON Schema

# Generate data from a JSON Schema file
dfgenerator --schema user.schema.json --rows 100 --seed 42 --output users.csv

from dfgenerator import build_from_jsonschema

schema = {
    "type": "object",
    "properties": {
        "id": {"type": "integer"},
        "email": {"type": "string", "format": "email"},
        "age": {"type": "integer", "minimum": 18, "maximum": 65}
    }
}

dataset = build_from_jsonschema(schema, rows=100, seed=42)
dataset.to_file("users.csv")

From YAML Config

dfgenerator --config people.yaml --rows 50 --output people.json

Programmatic API

from dfgenerator import DataFrameBuilder, RowTemplate, ColumnSpec, sequence, choices

template = RowTemplate([
    ColumnSpec("id", sequence(1)),
    ColumnSpec("city", choices(["Paris", "Berlin", "Boston"])),
])

dataset = DataFrameBuilder().from_template(template, 10).build()
df = dataset.to_pandas()

Architecture

dfgenerator uses a lightweight adapter pattern for extensibility:

• Core Domain: Pure business logic for data generation
• Input Adapters: YAML, JSON Schema, Spark DataFrames (easily extensible!)
• Output Adapters: CSV, JSON, Pandas, PySpark

Want to add support for Protobuf, Avro, or GraphQL schemas? Just create a new adapter!

See docs/ARCHITECTURE.md for details.

Documentation

• User Guide - Complete documentation and examples
• Architecture - Design patterns and extensibility guide