BTreeMap vs HashMap about ramhorns

[maciejhirsz]

The performance gain you have is not from replacing HashMap with BTreeMap, it's from removing the hash from a Field and forcing string compares everywhere. Why exactly is that faster isn't perfectly clear to me, but it's probably just leaving an extra register free.

The benchmarks won't reflect it since they only use two fiends and the names have different lengths, but this will be a performance degradation one you have more fields, and especially when you have two or more fields with names of the same length. I've explained the purpose of the hash in a blog post.