The two most notable changes in this PR are switching to sha256 for the data
keys and the addition of ref counting on data in the data map so that we can
delete it once it is no longer needed.
The key generation could probably use some profiling-based optimization. Right
now, we are serializing the data using the same serializer used in the data
section and then taking the sha256 of that. There is likely a faster way to
generate a unique key.
The two most notable changes in this PR are switching to sha256 for the data keys and the addition of ref counting on data in the data map so that we can delete it once it is no longer needed.
The key generation could probably use some profiling-based optimization. Right now, we are serializing the data using the same serializer used in the data section and then taking the sha256 of that. There is likely a faster way to generate a unique key.