Katkat
Back

Find Strictly Decreasing Forecast Windows for Equatorial Cities

VersemedicalSoftware Engineer · Onsite
MediumSoftware EngineerOnsiteSoftware Engineering Fundamentals

Last updated

This problem tests the ability to handle API integration, data stream processing, and sliding window logic while managing error handling and rate-limiting constraints.

Problem Description

Build a service function get_decreasing_windows(city_id) that identifies all 8-hour intervals where the temperature forecast is strictly decreasing.

Requirements

  1. Data Access: Query a PostgreSQL database to retrieve the latitude/longitude for a given city_id. Only proceed if the city is equatorial (latitude between -5 and 5 degrees).
  2. Data Acquisition: Fetch the next 72 hours of hourly forecast data from an external Weather API. The API returns a list of objects: {'timestamp': 'YYYY-MM-DDTHH:MM:SSZ', 'temp': float}.
  3. Resilience: The API is unstable. Implement a retry mechanism with exponential backoff for 5xx errors. If the API returns a 429 (Too Many Requests), respect the Retry-After header.
  4. Logic: Identify all 8-hour sliding windows (consecutive timestamps) where temp[i] > temp[i+1] > ... > temp[i+7].
  5. Validation: Ensure data continuity; if an hour is missing in the API response, that window cannot be considered valid.

Function Signature

def get_decreasing_windows(city_id: int) -> List[Tuple[str, str]]:
    # Returns a list of (start_timestamp, end_timestamp) tuples

Example

Input sequence (temps): [30, 29, 28, 27, 26, 25, 24, 23, 22, 24]

  • Window 1 (indices 0-7): [30, 29, 28, 27, 26, 25, 24, 23] (Strictly decreasing: Yes)
  • Window 2 (indices 1-8): [29, 28, 27, 26, 25, 24, 23, 22] (Strictly decreasing: Yes)
  • Window 3 (indices 2-9): [28, 27, 26, 25, 24, 23, 22, 24] (Strictly decreasing: No) Result: [('2023-10-01T00:00:00Z', '2023-10-01T07:00:00Z'), ('2023-10-01T01:00:00Z', '2023-10-01T08:00:00Z')]

Edge Cases

  • Missing data points (gaps in time).
  • City outside equatorial range.
  • Fewer than 8 hours of total data returned.

Hints

Hint 1

Use a helper function to validate the strictly decreasing condition using a simple loop or a deque to maintain a sliding window.

Hint 2

Implement an exponential backoff decorator or a simple loop that checks the response status code before parsing the JSON body.

Hint 3

To handle gaps, parse the ISO timestamps into Python datetime objects and compare consecutive timestamps to ensure the delta is exactly one hour.

Follow-up questions

  • How would you modify this solution if the city list was massive and you needed to perform this calculation for all equatorial cities every hour?
  • What strategies would you implement to cache the weather results to minimize API costs and stay within rate limits?
  • How would you mock the external Weather API to write unit tests for the retry logic and edge cases like missing timestamps?