can anyone explain why HAMT and derivates are not used in main memory database systems, but rather B+/Bw tree indexes or for instance tries based on different node sizes as for instance the ART (Adaptive Radix Tree) ot HOT (Height Optimized Trie)?
- https://db.in.tum.de/~leis/papers/ART.pdf
- https://dbis-informatik.uibk.ac.at/sites/default/files/2018-06/hot-height-optimized.pdf
I mean the HAMT also focuses on the same problem domain (BitSet + population count operation), that is reducing storage space consumption and being cache oblivious in comparison with ART and HOT. Binary comparable keys can be constructed as well.You could even combine the trie variants.
kind regards
Johannes