Static Member Variables vs. Dynamic Property Maps for Entity Persistence

The Source Engine entity system allows for property definition via hard-coded member variables or dynamic property maps. When designing for data persistence and save-game compatibility, these two approaches offer different trade-offs regarding performance and schema evolution.

Static member variables ensure type safety and high-speed access during the game loop but create rigid binary layouts. Conversely, dynamic property maps facilitate runtime schema changes and easier serialization of entity states, though they introduce lookup overhead and require explicit type casting.

A specific uncertainty exists regarding the handling of orphaned data during deserialization. If a property is removed from the schema but remains present in a saved game file, the engine's default behavior is to prioritize stability by ignoring unrecognized keys.

• Does the use of dynamic property maps provide a reliable mechanism to migrate legacy data without risking memory corruption?
• What is the performance impact of dynamic lookups compared to static members in high-entity-count scenes?