performance over alternatives?

[ibigbug]

I saw there's a go version of same idea https://github.com/openacid/succinct

which has some kind of "index" to speed up the read and they have some numbers in terms of perf.

maybe this lib can have something similar too?

[johnhurt]

I think that's a good idea. I included the results in the nodes to avoid lookups, but it might be worth it to use an index to make the stored graph smaller. gperf does that too.