create-python-tests

name	create-python-tests
description	Instructions on how to write Python unit tests to verify a Cortex data product's transformation logic, filters, and field mappings.

Creating Python Unit Tests for a Data Product

When a new data product is created or updated, you MUST create a corresponding Python unit test using pytest. This test ensures that the compiled SQL query correctly implements the business logic, transformations, joins, and filtering criteria documented in the implementation plan.

---

Workflow

Step 1: Analyze the Implementation Plan and Code

1. Retrieve Assumptions & Transformations: Read the implementation plan (from /create-data-product) to identify:
   • Which source tables are joined (e.g., KNA1 and ADRC).
   • What filtering rules are applied (e.g., mandt = '100', specific system versions).
   • What columns are expected and how they are transformed (e.g., coalesced fields, calculations).
2. Locate Scaffolding Files: Locate the custom data product's definition file (src/data_modules/<namespace>/<source>/products/<type>/definitions/*.js).
3. Leverage SAP & Domain Knowledge (CRITICAL): Apply your technical understanding of SAP DDIC structures (e.g., how client-specific data is separated, document lines vs headers, translation checks) and the business domain (e.g., Order-to-Cash, Procure-to-Pay, General Ledger) to design meaningful assertions. If any transformations or business rules are complex, custom, or ambiguous, you MUST ask the user for clarifying inputs on how those requirements should be verified in the unit tests before writing the test file.

Step 2: Scaffold the Test File

1. Test Location: Create the test file under tests/unit/<namespace>/ named test_<type>.py.
2. Template Reference: Use the base test template in test_template.py.md to structure the test file.

Step 3: Implement Test Assertions

Write assertions to verify the following from the compiled query:

1. Join Logic Verification: Parse the query to assert that the correct source tables are joined on the appropriate keys.
   • Example: Verify that KNA1 is joined with ADRC on kunnr.
2. Filter Logic Verification: Assert that filtering constraints from the user profile or assumptions are present in the SQL.
   • Example: Verify that mandt = '100' or mandt = "100" is enforced in the WHERE or ON clauses.
3. Field Transformations Verification: Verify that specific aliased or transformed fields exist in the SELECT statement.
   • Example: Verify that COALESCE is used on postal code fields or region fields.
4. Important Field Existence Verification: Verify that key or important fields (like primary keys, key foreign relations, and required business dimensions) are actively selected and projected in the final SQL statement.
   • Example: Verify that fields like customer_number_kunnr or valid_from_date are selected.
5. Custom Z-Fields Assertions: If the data product maps custom SAP fields (Z-fields, ZZ-fields, or YY-fields), write explicit assertions to verify that these custom columns are correctly projected in the SELECT query and adhere to the snake_case description mapping convention (e.g., asserting that 'zz_' or 'yy_' matches are present).

Step 4: Run the Unit Tests

1. Compile the workspace first: uv run cortex-build --config config/config.yaml.
2. Execute the unit tests using pytest. You MUST run the targeted test file for your custom data product to verify assertions quickly and efficiently:

   uv run pytest tests/unit/<namespace>/test_<type>.py -s
   

3. If any Test Fails: Fix the query definitions, configs, or test assertions and recompile/retest until the targeted test passes cleanly.
4. Full Suite Verification (CRITICAL): Once the targeted test passes, you MUST run the entire suite (uv run pytest) to ensure you haven't broken any other existing tests or core functionalities. ALL available unit tests must pass.

---

📊 Quality Gate Integration

Any Python unit tests created with this skill must be included in the Final Quality Report as part of the Quality Gate execution. Report the number of unit tests executed, passed, and failed.