Minimális API-hibareprodukció

API & Data Fixture Reproduction

When to Use

Use when reproducing backend API errors, webhook failures, ORM/SQL query bugs, pagination/filtering anomalies, serialization crashes, or permission edge cases.

Isolating the Minimal Request & Payload

1. Capture & Strip the HTTP Request

When starting from a complex frontend flow, HAR capture, or webhook log:

• Extract the exact HTTP method, URL path, query string, headers (Content-Type, Accept, Accept-Language, Authorization, Idempotency-Key, custom tenant headers), and request body.
• Replay the request in an isolated integration test or fetch script.
• Apply Delta Debugging to the payload and headers: strip optional fields and headers one group at a time until only the exact fields and values required to trigger the bug remain.

2. High-Yield Data Boundary & Edge-Case Triggers

When a bug report says "fails for certain records/users," check these data shapes first:

• Nullability & Empty Collections: null vs. undefined vs. "" (empty string) vs. [] (empty array); parent records with 0 children vs. 1 child vs. > page_size children; soft-deleted (deleted_at IS NOT NULL) or archived foreign-key relations.
• Unicode, Encoding & Special Characters: Multi-byte UTF-8 emojis (👨‍👩‍👧‍👦), accented characters (ő, ű, ß), right-to-left text, zero-width spaces (\u200B), SQL wildcard characters (%, _, \), regex metacharacters ([, (, *, +, ?), and quotes/newlines in CSV/JSON strings.
• Numeric & Financial Boundaries: 0, -0, negative numbers, floating-point precision (0.1 + 0.2), Number.MAX_SAFE_INTEGER (9007199254740991), BigInt serialization, and NaN / Infinity.
• Pagination & Sorting Ties: Duplicate sort timestamps (created_at ties across page boundaries causing skipped or duplicated rows in LIMIT/OFFSET or cursor pagination).

3. Build Minimal, Anonymized Database Fixtures

• Never copy raw production database dumps containing customer PII, real emails, passwords, or live API credentials into a test environment.
• Build a Minimal Synthetic Fixture (1–3 rows with anonymized/fake identifiers) that preserves the exact relational topology and column values responsible for the bug:
  • Example: If a query fails when a user belongs to two organizations where one subscription is canceled and the other is active, seed exactly 1 user + 2 org memberships with those two states.
• Verify that running the API endpoint or service method against this minimal fixture deterministically reproduces the exact SQL error, wrong aggregation, or 500 exception.