police_data/main.py

import requests
import time
import sqlite3
import threading
import queue
from collections import deque
from typing import List

# --------------------------------
# Thread-Safe Leaky Bucket Rate Limiter
# --------------------------------
class LeakyBucketRateLimiter:
    def __init__(self, max_rate: float = 15.0):
        self.max_rate = max_rate
        self.timestamps: deque = deque()
        self.lock = threading.Lock()
        self.window: float = 1.0  # Sliding window size in seconds

    def acquire(self) -> None:
        with self.lock:
            now = time.monotonic()
            # Remove timestamps outside the current window
            while self.timestamps and self.timestamps[0] <= now - self.window:
                self.timestamps.popleft()

            # If limit reached, sleep until the oldest request exits the window
            if len(self.timestamps) >= self.max_rate:
                sleep_time = 1.0 - (now - self.timestamps[0])
                if sleep_time > 0:
                    time.sleep(sleep_time)
                now = time.monotonic()
                while self.timestamps and self.timestamps[0] <= now - self.window:
                    self.timestamps.popleft()

            self.timestamps.append(time.monotonic())

# --------------------------------
# Database Initialization
# --------------------------------
def init_db(db_path: str = "uk_crimes.db") -> sqlite3.Connection:
    conn = sqlite3.connect(db_path)
    conn.execute("""
        CREATE TABLE IF NOT EXISTS crimes (
            id INTEGER PRIMARY KEY,
            category TEXT,
            month TEXT,
            latitude REAL,
            longitude REAL,
            street_name TEXT,
            outcome_category TEXT
        )
    """)
    conn.commit()
    return conn

# --------------------------------
# UK Grid Polygon Generator
# --------------------------------
def generate_uk_polygons(lat_min: float = 49.5, lat_max: float = 61.0,
                         lng_min: float = -13.0, lng_max: float = 5.0,
                         step: float = 0.5) -> List[str]:
    polygons = []
    lats = [l / 10.0 for l in range(int(lat_min * 10), int(lat_max * 10), int(step * 10))]
    lns = [ln / 10.0 for ln in range(int(lng_min * 10), int(lng_max * 10), int(step * 10))]

    for i in range(len(lats) - 1):
        for j in range(len(lns) - 1):
            poly = f"{lats[i]},{lns[j]}:{lats[i]},{lns[j+1]}:{lats[i+1]},{lns[j+1]}:{lats[i+1]},{lns[j]}"
            polygons.append(poly)
    return polygons

# --------------------------------
# Producer Thread: Fetch & Queue
# --------------------------------
def producer_worker(polygons: List[str], date: str, limiter: LeakyBucketRateLimiter, data_queue: queue.Queue):
    total = len(polygons)
    for i, poly in enumerate(polygons):
        # Skip polygons exceeding the 4094 character limit [1]
        if len(poly) > 4090:
            continue

        # Enforce strict 15 requests/second limit [1]
        limiter.acquire()

        try:
            response = requests.get(
                "https://data.police.uk/api/crimes-street/all-crime",
                params={"date": date, "poly": poly},
                timeout=30
            )
            response.raise_for_status()
            data = response.json()

            # API returns 503 if a custom area contains >10,000 crimes [1]
            if not isinstance(data, list):
                print(f"[PRODUCER] Warning: Received non-list response for poly {poly[:30]}...")
                continue

            if not data:
                continue

            rows = []
            for crime in data:
                loc = crime.get("location", {})
                outcome = crime.get("outcome_status") or {}
                rows.append((
                    crime.get("id"),
                    crime.get("category"),
                    crime.get("month"),
                    float(loc.get("latitude", 0)),
                    float(loc.get("longitude", 0)),
                    loc.get("street", {}).get("name"),
                    outcome.get("category")
                ))

            # Push batch to memory queue (non-blocking if size allows)
            data_queue.put(rows)

        except requests.exceptions.RequestException as e:
            print(f"[PRODUCER] Network error for poly {poly[:30]}... | {e}")
        except Exception as e:
            print(f"[PRODUCER] Unexpected error: {e}")

        if (i + 1) % 100 == 0 or i == total - 1:
            print(f"[PRODUCER] Progress: {i+1}/{total} fetched | Queue size: {data_queue.qsize()}")

    # Signal completion to consumer
    data_queue.put(None)

# --------------------------------
# Consumer Thread: Queue to DB
# --------------------------------
def consumer_worker(data_queue: queue.Queue, conn: sqlite3.Connection):
    batch_size = 500
    batch = []

    while True:
        try:
            item = data_queue.get(timeout=5.0)
        except queue.Empty:
            # If queue is empty for 5s, assume production is done
            if not batch:
                break
            else:
                # Flush remaining batch
                conn.executemany(
                    "INSERT OR IGNORE INTO crimes (id, category, month, latitude, longitude, street_name, outcome_category) VALUES (?, ?, ?, ?, ?, ?, ?)",
                    batch
                )
                conn.commit()
                batch = []
                continue

        if item is None:
            # Flush last batch before exiting
            if batch:
                conn.executemany(
                    "INSERT OR IGNORE INTO crimes (id, category, month, latitude, longitude, street_name, outcome_category) VALUES (?, ?, ?, ?, ?, ?, ?)",
                    batch
                )
                conn.commit()
            data_queue.task_done()
            break

        batch.extend(item)

        if len(batch) >= batch_size:
            conn.executemany(
                "INSERT OR IGNORE INTO crimes (id, category, month, latitude, longitude, street_name, outcome_category) VALUES (?, ?, ?, ?, ?, ?, ?)",
                batch
            )
            conn.commit()
            batch = []

        data_queue.task_done()

# --------------------------------
# Main Orchestrator
# --------------------------------
def main():
    limiter = LeakyBucketRateLimiter(max_rate=15.0)
    conn = init_db("uk_crimes.db")
    data_queue = queue.Queue(maxsize=2000)  # Backpressure to prevent memory overflow
    date = "2024-01"

    print("Generating UK coverage polygons...")
    polygons = generate_uk_polygons()
    print(f"Total regions to process: {len(polygons)}\n")

    # Start threads
    prod_thread = threading.Thread(target=producer_worker, args=(polygons, date, limiter, data_queue), daemon=True)
    cons_thread = threading.Thread(target=consumer_worker, args=(data_queue, conn), daemon=True)

    cons_thread.start()
    prod_thread.start()

    # Wait for producer to finish
    prod_thread.join()
    print("\n[MAIN] Producer finished. Waiting for consumer to drain queue...")

    # Wait for consumer to finish
    cons_thread.join()

    total = conn.execute('SELECT COUNT(*) FROM crimes').fetchone()[0]
    print(f"[MAIN] Completed. Total unique records stored in uk_crimes.db: {total}")
    conn.close()

if __name__ == "__main__":
    main()