Compressing tree data is not trivial. E.g. using per-node heap allocation could take many times the space of the actual data (memory used by the allocator, meta

Compressing tree data is not trivial. E.g. using per-node heap allocation could take many times the space of the actual data (memory used by the allocator, metainformation, alignment padding, etc.). Using a custom allocator can reduce the memory usage dramatically.