You are asked to implement a data storage system that supports basic operations, filtering, and "point-in-time" recovery (backups).

Level 1: Basic Storage (CRUD)
The goal is to manage records with multiple fields.

SET_OR_INC(key, field, value): If the field is a string, update it. If it's an integer, increment it.

GET(key, field): Basic retrieval.

DELETE(key, field): Remove a specific field or the entire key if no fields remain.

Level 2: Scanning & Filtering (The "Fetch" Level)
This level tests how you iterate over your data.

FETCH(field, value): Scan all records and return those that match.

The Catch: Usually, the output must be formatted as a string (e.g., "field1(val1), field2(val2)") and the results must be sorted alphabetically by the key.

Level 3: Advanced Filtering or TTL (Time-To-Live)
The final level usually adds one of two "Industry" constraints:

Option A (TTL): Every SET operation can include a ttl (Time-To-Live). If you GET or FETCH after the time has expired, the record must behave as if it was deleted.

Option B (Multi-field Filtering): Instead of matching one field, you might be asked to FETCH based on a prefix or a range of values (e.g., value > 100), requiring more complex iteration logic.

Level 4: Point-in-Time Snapshots (The "Backup" Level)
This is where the state management becomes tricky.

BACKUP(): Capture the state and return a backupId.

RESTORE(backupId): The database must instantly revert.

Implementation Tip: If you use a simple deepCopy(), you might fail the hidden performance tests for large datasets. High-performing solutions often use versioned maps or a Copy-on-Write strategy.

This description is from gemini but the problem was something like this only

I could not pass all test cases for level 4

scored 510/600

Platform - codesignal