Filter Open Restaurants Through Existing Domain APIs

Problem Statement

At DoorDash, presenting an accurate feed of open restaurants requires coordinating multiple downstream microservices. You are tasked with implementing the core orchestration logic for a service that retrieves and filters restaurants for a given city at a specific Unix timestamp.

You are provided with two external domain client interfaces:

1. DirectoryClient

   • get_restaurants_by_city(city: str) -> list[dict]
   • Returns restaurant summaries: [{"id": str, "name": str, "city": str, "active": bool}, ...]

2. HoursClient

   • is_open(restaurant_id: str, timestamp_utc: int) -> bool
   • Returns True if the store is open at the given timestamp, False otherwise. This call is network-intensive and may raise exceptions (e.g., TimeoutError or ConnectionError).

Implement the class OpenRestaurantFinder with the following requirements:

• Pre-filtering & Short-circuiting: Do not call HoursClient for restaurants where active is False or whose city does not match the requested city (case-insensitive comparison).
• Bounded Concurrency: Query the HoursClient concurrently for valid candidates using at most max_workers parallel workers.
• Failure Handling: If HoursClient.is_open raises an error for any restaurant, treat that restaurant as closed (False) rather than failing the entire request.
• Order Preservation: The returned list of open restaurants must maintain the relative ordering provided by the DirectoryClient.
• Safe In-Memory Caching: Cache the boolean result of (restaurant_id, timestamp_utc) calls in memory so repeated requests with identical parameters do not make duplicate downstream calls to HoursClient.

Class Interface

class OpenRestaurantFinder:
    def __init__(self, directory_client, hours_client, max_workers: int = 5):
        """
        Initialize with downstream API clients and concurrency limit.
        """
        pass

    def get_open_restaurants(self, city: str, timestamp_utc: int) -> list[dict]:
        """
        Returns only active, open restaurants for the specified city at timestamp_utc,
        preserving original ranking order.
        """
        pass

Constraints

• 1 = 0
• 1 <= max_workers <= 20
• Number of restaurants returned by directory: 0 <= N <= 10,000
• The cache should be thread-safe if your implementation uses multi-threading.

Examples

Example 1

Directory response for city = "San Francisco":

[
    {"id": "r1", "name": "Tartine", "city": "San Francisco", "active": True},
    {"id": "r2", "name": "Bob's Donuts", "city": "San Francisco", "active": False},
    {"id": "r3", "name": "Arsicault", "city": "San Francisco", "active": True},
    {"id": "r4", "name": "Che Fico", "city": "Oakland", "active": True}
]

HoursClient.is_open(restaurant_id, 1700000000) results:

• "r1": True
• "r3": False
• "r2" and "r4" should never be queried because "r2" is inactive and "r4" is in Oakland.

Output:

[
    {"id": "r1", "name": "Tartine", "city": "San Francisco", "active": True}
]

Example 2 (Downstream Failure)

Directory response for city = "Seattle":

[
    {"id": "s1", "name": "Pike Bagels", "city": "Seattle", "active": True},
    {"id": "s2", "name": "Canlis", "city": "Seattle", "active": True}
]

HoursClient.is_open calls:

• "s1" raises TimeoutError
• "s2" returns True

Output:

[
    {"id": "s2", "name": "Canlis", "city": "Seattle", "active": True}
]