scd2-bq-engine 0.1.0

BigQuery SCD Type 2 dimension builder with SQLX generation

SCD2 BigQuery Engine

Version: 0.1.0
License: MIT
Status: ✅ Production Ready

Overview

BigQuery-first SCD Type 2 dimension builder with SQLX template generation for Dataform/dbt.

Generates production-ready SQLX files that implement full SCD Type 2 logic including:

• Hash-based change detection
• Effective date tracking
• Soft deletes
• Late arrival handling
• BigQuery optimizations (partitioning, clustering)

Why This Module?

• dbt snapshots are dbt-locked and Snowflake-optimized
• Existing SCD2 tutorials use stored procedures (hard to version control)
• No declarative, config-driven SCD2 solution for BigQuery

Features (v0.1.0 MVP)

• ✅ SQLX template generator for standard SCD2 pattern
• ✅ Hash-based change detection (MD5/SHA256)
• ✅ Effective dating with late arrival handling
• ✅ Soft delete support
• ✅ YAML config → SQLX output
• ✅ CLI + Python API

Quick Start

Installation

pip install scd2-bq-engine

Initialize Configuration

# Create a new dimension configuration
scd2-bq init dim_employee \
    --source-table project.dataset.stg_employees \
    --business-keys employee_number \
    --tracked-columns first_name,last_name,job_code,department

Generate SQLX

# Generate SCD2 dimension SQLX file
scd2-bq generate \
    --config dim_employee_config.yaml \
    --output-file dim_employee.sqlx

Python API

from scd2_bq_engine import SCD2Generator, SCD2Config

# Create configuration
config = SCD2Config(
    dimension_name="dim_employee",
    source_table="project.dataset.stg_employees",
    business_keys=["employee_number"],
    tracked_columns=["first_name", "last_name", "job_code"]
)

# Generate SQLX
generator = SCD2Generator(config)
generator.write_sqlx("dim_employee.sqlx")

Directory Structure

scd2-bq-engine/
├── src/scd2_bq_engine/
│   ├── generator.py          # Core SQLX generator
│   ├── templates/            # Jinja2 SQLX templates
│   └── validators.py         # Config validation
├── tests/
├── examples/
└── docs/

License

MIT - see LICENSE