Homework 14 — Cost & Governance

Task 1 — Detect the most expensive team last week

Write a single query that returns the one team with the highest cost in the last 7 days. Use the DuckDB _query_costs table (local) or SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY (Snowflake, optional).

The result must be a single row with at least two columns: team and a cost column.

Self-check:

• The query returns exactly 1 row.
• The team column contains a value extracted from the tag (not the raw tag string).
• The cost column is a numeric aggregate (SUM, not a raw row value).

Task 2 — Add a new role to grants.yml

Define a data_scientist role in governance/grants.yml with read-only (SELECT) access to fact_events AND read-only access to a hypothetical feature_store table. Reuse the existing YAML structure exactly — same indentation, same key names, same list syntax.

Self-check:

• python -c "import yaml; yaml.safe_load(open('governance/grants.yml'))" runs without error.
• The new role appears under the roles: key (not at the top level, and not under masking_policies:).
• Both grants specify only [SELECT] — not [SELECT, INSERT].

Task 3 — Audit-log a declined deletion request

Write a .jsonl entry for a GDPR erasure request that you decline. Invent a fictional public-figure account whose data falls under a journalistic exemption (GDPR Article 85). The entry must include these fields: ts, action: "gdpr_decline", subject, reason, legal_basis.

Append the entry to governance/gdpr/deletion_log.jsonl (do not replace the file; use open(..., 'a')).

Self-check:

• The entry parses cleanly: json.loads(line) returns a dict with all 5 required fields.
• The file has at least two lines (the lab-step-7 entry from the lab + your new entry).
• The action value is exactly "gdpr_decline" (not "gdpr_delete").