sam boyer

(I'm now working on this in #196)

note to self (ish) - getting rid of this should mean that `pcSliceToMap()` can probably also go.

I appreciate the PR, but this is intentional. Assuming the calling code has created the `*SourceMgr` through `NewSourceManager()` and is not messing with `unsafe`, it's not possible to reach that...

If you're willing to do that, please also add a test case for it - that really codifies that it's intentional.

Thanks, I do think this helps frame the problem. > VCS DAGs are different than the dependency resolution problem. So a different random structure will be needed. This type of...

> However, that might actually be easy to solve. The hardest instances will be when there are many "promising" partial solutions but only one solution which is difficult to find...

It certainly makes sense that there would be a way to express the problem here as a weird version of subgraph isomorphism - both are NP-complete. The jump I'm not...

> Since, in practice there are a finite number of projects and every project has a finite number of versions there are only a finite number of nodes created. Thus,...

bump :)

Awesome! I'd love to get rid of the binary dependency, and wouldn't mind avoiding cgo-derived complications from relying on libgit2, either. At the same time, libgit2 offers guarantees about comparable...