Open
Description
Right now the compiler will always load and deserialize all query result cache entries that are still valid and then serialize and write them to back to disk again, similar to how a copying garbage collector works. The main reason for implementing the cache this way is that it is very simple.
There might be value in making the cache file updateable in-place in order to save serialization time for unchanged entries. First though, we need some profiling evidence that this would actually save a noticeable amount of time.