fix: memory leak in InMemorySessionService.deleteSession() - clean up empty parent maps

[YuqiGuo105]

Problem

Fixes #687 #687

InMemorySessionService stores sessions in a three-level nested ConcurrentHashMap:

sessions: appName  userId  sessionId  Session

deleteSession() only removed the leaf sessionId entry but never pruned the parent userId or appName keys when their child maps became empty:

BEFORE: sessions = { myApp: { user1: { ctx1: Session@123 } } }
AFTER (buggy): sessions = { myApp: { user1: {} } }    ghost empty maps!
AFTER (fixed): sessions = {}                               fully cleaned up

At scale (e.g. 1M unique users each creating and deleting one session), these empty ConcurrentHashMap objects accumulate as strongly-reachable JVM objects, causing linear memory growth and eventually OutOfMemoryError.

Fix

Replace the get() + remove() pattern with nested computeIfPresent() calls. When the remapping function returns null, ConcurrentHashMap atomically removes the key no additional locking needed.

sessions.computeIfPresent(appName, (app, appSessionsMap) -> {
    appSessionsMap.computeIfPresent(userId, (user, userSessionsMap) -> {
        userSessionsMap.remove(sessionId);
        // return null when empty  userId key is auto-removed atomically
        return userSessionsMap.isEmpty() ? null : userSessionsMap;
    });
    // return null when empty  appName key is auto-removed atomically
    return appSessionsMap.isEmpty() ? null : appSessionsMap;
});

Tests Added

Two new tests in InMemorySessionServiceTest using Java reflection to access the private sessions field:

Test	Verifies
deleteSession_cleansUpEmptyParentMaps	After deleting the only session, the internal sessions map is completely empty
deleteSession_doesNotRemoveUserMapWhenOtherSessionsExist	When a user still has remaining sessions, the userId and appName entries are NOT prematurely pruned

All 10 InMemorySessionServiceTest tests pass (0 failures, 0 errors).

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!