Fixed in one of PRs #100 xor #98 - BIP32 subkey derivation raises exception for negative indexes (contrary to documentation) about pycoin HOT 1 CLOSED

I was dimly aware of the problems of using negative integers to indicate hardened subkeys when I wrote the test that I submitted with PR #98, but I didn't realize the implications until daydreaming about it just now.¹

Allowing negative numbers to represent hardened children with in a BIP32Node may be confusing, since there's no way to represent a hardened subkey at index "-0". In python:

>>> -0 == 0
True
>>> -0 < 0
False

This means that BIP32Node.subkey(i=-0) will always yield the _un_hardened subkey (just like BIP32Node.subkey(i=0)). Given this oddity (and that 0 is probably a very popular index), it's probably best to abandon the convention of using negative numbers for hardened subkeys, and instead rely on explicit parameters or the H or p notation in BIP32Node.subkey_for_path.

As an aside, the '-' notation also conflicts with the current implementation of the hyphen range notation for BIP32Node.subkeys.

¹ The fact that the test wouldn't pass for i = 0 should have tipped me off, but hey, better late than never, right? 😁

from pycoin.