Bbtree: Fastest Ordered Map

The discussion focuses on whether Bbtree’s performance claims generalize beyond the workloads used to promote it. One commenter argues that trie-based and stratified structures are serious alternatives to B-trees, especially for ordered primitive keys. Another tested Bbtree in a mesh kernel and found it slower in smaller operations and mixed at million-entry scale. In follow-up microbenchmarks, Bbtree’s automatic key handling improved insert and lookup times, while the custom InlineKey path slowed narrow keys; iteration was much faster, but was not the kernel’s bottleneck. The author says Bbtree works best with InlineKey and invites more workload reports. A separate test found higher memory use for string keys, while noting that the project’s lower-memory claim concerned integers. A developer plans to try it for bulk-write deduplication.

Opinions divide between appreciation for the implementation and skepticism about broad “fastest” claims: commenters stress that map performance depends on key type, workload, and tradeoffs, and suggest comparing against other specialized structures. Some criticize the project’s promotional tone, its claim that LLMs outperform human coders, and perceived code quality; others praise the work or focus on possible low-level optimizations. Commenters also question what “model checking” means in the project’s verification claims and debate whether the named Rust tools qualify. Another finds the article’s dynamic presentation distracting.