improving similarity cache

lib/services/similarity-check/similarityCache.js

@@ -1,116 +1,92 @@
-import crypto from 'crypto';
-
-const retention = 60 * 60 * 1000;
 /**
- * Internal cache storage.
- * Maps a SHA-256 hash (string) to its expiry timestamp (number in ms).
- * @type {Map<string, number>}
- */
-const entries = new Map();
-
-/**
- * Reference to the currently scheduled cleanup timer.
- * @type {NodeJS.Timeout | null}
- */
-let timer = null;
-
-/**
- * Generate a SHA-256 hash from a list of input strings.
- * Null or undefined values are ignored.
+ * Similarity cache
  *
- * @param {...(string|null|undefined)} strings - Input values to hash
+ * Maintains an in-memory Set of content hashes to detect whether a listing
+ * (identified by a tuple of title, price and address) has been seen before.
+ *
+ * Design notes:
+ * - The cache is refreshed periodically from persistent storage. To avoid
+ *   modification-during-iteration issues, the refresh builds a new Set and
+ *   atomically swaps the reference instead of mutating in place.
+ * - Hashing ignores null/undefined values but preserves falsy-yet-valid values
+ *   like 0. Non-string values are coerced to strings before hashing.
+ *
+ * This module has no persistence of its own; it relies on
+ * getAllEntriesFromListings() for data hydration.
+ * @module similarityCache
+ */
+import crypto from 'crypto';
+import { getAllEntriesFromListings } from '../storage/listingsStorage.js';
+
+/** @type {number} Refresh interval in milliseconds (defaults to one hour). */
+const reloadCycle = 60 * 60 * 1000; // every hour, refresh
+
+/**
+ * Internal cache of content hashes for known listings.
+ *
+ * Each entry is an SHA-256 hex digest produced by toHash(title, price, address).
+ * @type {Set<string>}
+ */
+let cache = new Set();
+
+// Periodically refresh the cache from storage
+setInterval(() => {
+  initSimilarityCache();
+}, reloadCycle);
+
+/**
+ * Initialize or refresh the similarity cache from persistent storage.
+ *
+ * Reads all stored listings via getAllEntriesFromListings(), computes a hash for
+ * each, and swaps the in-memory Set atomically to avoid in-place mutations that
+ * could interfere with concurrent iteration.
+ *
+ * This function is idempotent and safe to call at any time.
+ * @returns {void}
+ */
+export const initSimilarityCache = () => {
+  const allEntries = getAllEntriesFromListings();
+  const newCache = new Set();
+  for (const entry of allEntries) {
+    newCache.add(toHash(entry?.title, entry?.price, entry?.address));
+  }
+  // Atomic swap to avoid mutating the cache while it may be iterated elsewhere
+  cache = newCache;
+};
+
+/**
+ * Check if a listing is already known and add it to the cache if not.
+ *
+ * The listing is identified by the combination of its title, price and
+ * address. Null/undefined fields are ignored during hashing. Falsy-but-valid
+ * values (e.g., price 0) are preserved.
+ *
+ * @param {Object} params - Listing fields
+ * @param {string|undefined|null} params.title - The listing title
+ * @param {string|undefined|null} params.address - The listing address
+ * @param {number|string|undefined|null} params.price - The listing price
+ * @returns {boolean} true if the entry already existed in the cache (duplicate), otherwise false
+ */
+export const checkAndAddEntry = ({ title, address, price }) => {
+  const hash = toHash(title, price, address);
+  if (cache.has(hash)) {
+    return true;
+  }
+  cache.add(hash);
+  return false;
+};
+
+/**
+ * Generate an SHA-256 hash from a list of input values.
+ * Null or undefined values are ignored. Falsy but valid values like 0 are preserved.
+ * Non-string values are coerced to strings prior to hashing.
+ *
+ * @param {...(string|number|null|undefined)} strings - Input values to hash
  * @returns {string} Hexadecimal hash
  */
 function toHash(...strings) {
-  return crypto.createHash('sha256').update(strings.filter(Boolean).join('|')).digest('hex');
-}
-
-/**
- * Cleanup expired cache entries and schedule the next cleanup run.
- * This function is invoked automatically by scheduled timers.
- *
- * @private
- */
-function runCleanup() {
-  const now = Date.now();
-  for (const [hash, expiry] of entries) {
-    if (expiry <= now) entries.delete(hash);
-  }
-  scheduleNext();
-}
-
-/**
- * Find the soonest expiry timestamp among all cache entries
- * and schedule a one-shot timer that will trigger at that time.
- * Cancels any existing timer before scheduling a new one.
- *
- * @private
- */
-function scheduleNext() {
-  if (timer) {
-    clearTimeout(timer);
-    timer = null;
-  }
-  let next = Infinity;
-  const now = Date.now();
-  for (const expiry of entries.values()) {
-    if (expiry > now && expiry < next) next = expiry;
-  }
-  if (next !== Infinity) {
-    timer = setTimeout(runCleanup, Math.max(0, next - now));
-  }
-}
-
-/**
- * Add or refresh a cache entry for the given title and address.
- * The entry will automatically expire after the configured retention window.
- *
- * @param {string} title - The title used to build the cache key
- * @param {string} address - The address used to build the cache key
- */
-export function addCacheEntry(title, address) {
-  const hash = toHash(title, address);
-  const expiry = Date.now() + retention;
-  entries.set(hash, expiry);
-  scheduleNext();
-}
-
-/**
- * Check if a cache entry with the same title and address exists
- * and is still valid (not expired).
- *
- * @param {string} title - The title used to build the cache key
- * @param {string} address - The address used to build the cache key
- * @returns {boolean} True if a valid cache entry exists, false otherwise
- */
-export function hasSimilarEntries(title, address) {
-  const hash = toHash(title, address);
-  const expiry = entries.get(hash);
-  if (expiry == null) return false;
-  if (expiry <= Date.now()) {
-    entries.delete(hash);
-    scheduleNext();
-    return false;
-  }
-  return true;
-}
-
-/**
- * Stop any scheduled cleanup timers and prevent further automatic cleanup.
- * Entries that are already in the cache will remain until removed manually
- * or until cleanup is started again by adding new entries.
- */
-export function stopCacheCleanup() {
-  if (timer) clearTimeout(timer);
-  timer = null;
-}
-
-/**
- * this is only for test purposes
- */
-export function invalidateAllForTest() {
-  for (const key of entries.keys()) {
-    entries.set(key, 0);
-  }
-  runCleanup();
+  const normalized = strings
+    .filter((v) => v !== null && v !== undefined)
+    .map((v) => (typeof v === 'string' ? v : String(v)));
+  return crypto.createHash('sha256').update(normalized.join('|')).digest('hex');
 }

lib/services/storage/listingsStorage.js

@@ -310,8 +310,8 @@ export const deleteListingsByJobId = (jobId) => {
   if (!jobId) return;
   return SqliteConnection.execute(
     `DELETE
-                                   FROM listings
-                                   WHERE job_id = @jobId`,
+     FROM listings
+     WHERE job_id = @jobId`,
     { jobId },
   );
 };
@@ -332,3 +332,13 @@ export const deleteListingsById = (ids) => {
     ids,
   );
 };
+
+/**
+ * Return all listings with only the fields: title, address, and price.
+ * This is the single helper requested for simple consumers.
+ *
+ * @returns {{title: string|null, address: string|null, price: number|null}[]}
+ */
+export const getAllEntriesFromListings = () => {
+  return SqliteConnection.query(`SELECT title, address, price FROM listings`);
+};