Experimentation with golang.
All memory index for finding similarity within documents.
Less than 100k documents, document is qualified by 1k tags.
go get ./build.sh
In a terminal:
n-1 procs is a good number.
When application starts, 50k documents is randomly generated.
In another terminal:
Looking for document like document 42.
A document got tags. It's represented by a bitset.
Tags can have weight.
Document can have weight.