Working up the stack to fill in some of the API methods and remove some dead commented-out code.
Fixed the last remaining benchmark, which required the testing grpc services, now working again.
Out of the box, there is less cache so it's slower than the old implementation. With a larger cache (20k nodes), it outperforms the old code (6s for 1M rows vs 84s).
Also fixed propose so it doesn't require a mutable DB.
Working up the stack to fill in some of the API methods and remove some dead commented-out code.
Fixed the last remaining benchmark, which required the testing grpc services, now working again.
Out of the box, there is less cache so it's slower than the old implementation. With a larger cache (20k nodes), it outperforms the old code (6s for 1M rows vs 84s).
Also fixed propose so it doesn't require a mutable DB.