This directory contains the generation script for small embedded DuckDB databases used for testing BenchBox functionality. The databases are created automatically during test execution and cleaned up afterwards.

• basic_test.duckdb - Simple tables for connection and basic SQL testing
• tpch_test.duckdb - Minimal TPC-H benchmark tables with sample data
• tpcds_test.duckdb - Minimal TPC-DS benchmark tables with sample data
• ssb_test.duckdb - Star Schema Benchmark tables with sample data
• primitives_test.duckdb - Tables for testing OLAP operations and window functions

• Size: Each database is kept small (< 5MB) for fast test execution
• Data: Contains minimal but representative data for testing functionality
• Schema: Follows the standard benchmark schemas but with reduced data volume
• Read-Only: Test fixtures open databases in read-only mode to prevent contamination
• Temporary: Created before tests run and cleaned up after tests complete
• Not Committed: Database files are excluded from git via .gitignore

These databases are accessed through fixtures defined in tests/fixtures/database_fixtures.py:

def test_example(tpch_test_db):
    # Use the TPC-H test database
    result = tpch_test_db.execute("SELECT COUNT(*) FROM customer").fetchone()
    assert result[0] > 0

Available fixtures:

• basic_test_db - Connection to basic_test.duckdb
• tpch_test_db - Connection to tpch_test.duckdb
• tpcds_test_db - Connection to tpcds_test.duckdb
• ssb_test_db - Connection to ssb_test.duckdb
• primitives_test_db - Connection to primitives_test.duckdb

The test databases are managed automatically by the test framework:

1. Creation: Databases are created before test sessions start (via pytest_sessionstart)
2. On-Demand: If a database is missing during a test, it's created automatically
3. Cleanup: All databases are removed after test sessions complete (via pytest_sessionfinish)

If you need to create databases manually for debugging:

uv run -- python tests/databases/create_test_databases.py

This will recreate all database files with fresh data.

1. No Mocks: These databases eliminate the need for database mocking in tests
2. Realistic Data: Provides actual SQL execution against real data
3. Fast Tests: Small size ensures quick test execution
4. Deterministic: Same data every time for consistent test results
5. Isolated: Each test gets a fresh connection to prevent interference

• Reliability: Tests execute against real databases, catching SQL errors
• Performance: Small databases load quickly while providing realistic execution
• Maintainability: No complex mock setup and teardown logic
• Accuracy: Tests verify actual database behavior rather than mock behavior