# Fetch and Rank Articles from a Paginated API Implement `top_articles(limit)` against a paginated HTTP endpoint. Page `p` returns JSON with `data` and `total_pages`. Each article may have `title`, `story_title`, and `num_comments`. Use `title` when it is non-empty, otherwise use `story_title`. Ignore an article when both are missing or empty. Treat a missing or null comment count as zero. Return up to `limit` selected titles ordered by descending comment count and then ascending title. A negative or zero limit returns an empty list without making a request. Discuss HTTP failure handling, pagination consistency, timeouts, and tests. Do not assume the endpoint always returns valid JSON or that every page succeeds. ### Constraints & Assumptions - Page numbers begin at one. - `total_pages` comes from the first successful page and is a positive integer. - The same article may appear twice only if the endpoint itself is inconsistent; state whether you deduplicate and by which stable key. ### Clarifying Questions to Ask - Is partial data acceptable if a later page fails? - Does the API provide a stable article ID and snapshot token? - Are retries safe and rate-limited? - Is title comparison case-sensitive? ### What a Strong Answer Covers - Boundary handling before network access - A small request helper with timeout and status checks - Validation of response shape on every page - Correct title fallback and sorting keys - Explicit all-or-nothing or partial-result semantics - Bounded retries, rate-limit awareness, and useful non-secret diagnostics - Unit tests with a mocked HTTP client ### Follow-up Questions - How would you fetch pages concurrently while respecting rate limits? - How would you process more articles than memory can hold? - How would you make results reproducible while the remote data changes?