Graph server edgeCache invalidate strategy

[haohao0103]

According to the current logic, the edgeCache invalidation mechanism is triggered by any vertex addition, update, or deletion, as well as any edge addition, update, or deletion, resulting in the clearing of the entire edge cache. As shown in the image, my expectation is to handle edgeCache invalidation differently based on the type of vertex change. From what I understand, the edgeCache's cache key is the queryId, such as:

Query * from EDGE where [OWNER_VERTEX == 2, DIRECTION == OUT OR DIRECTION == IN] Query * from EDGE where [OWNER_VERTEX == 1, DIRECTION == OUT OR DIRECTION == IN]

This data is sourced from CachedGraphTransactionTest#testEdgeCacheClearWhenUpdateVertex.

The queryId in edgeCache is related to the vertex ID and not to the vertex's properties. Therefore, vertex addition or update operations should not trigger edgeCache invalidation. Is this understanding correct?

For vertex deletion operations, only the edgeCache entries corresponding to the relevant queryId should be cleared, rather than clearing the entire edge cache. This requires designing a data structure to maintain the mapping between vertices and their associated queryIds.

For edge change operations, I do not yet have a clear approach and will temporarily maintain the original design.

I welcome any guidance, suggestions, or feedback on this approach. Thank you in advance for your input!