Closed quasicomputational closed 6 years ago
Wow, terrific - this looks much simpler. I'll try and get to this soon (along with your other PRs!)
OK, now that #53 is fixed, I've verified that the test case can round-trip, so I've got some confidence in this PR.
I've had another read, and this both way simpler and seems to make sense. You walk the CondTree and cat all branches at each level into a single branch essentially. Given that all the tests pass, I'm happy with this. Good to merge?
I'm happy with it!
I think that this is a lot clearer about what it is doing, rather than the previous algorithm, which was going wrong and causing #36 and then making my eyes cross as I tried to figure out precisely how.
Fixes #36.
I'm not 100% confident that this is correct since it doesn't round trip (see #53) and the test's output is too big to inspect directly, but I'm not sure where a bug could be hiding in the code.