Jean-Marie Comets

I had a look at switching out the underlying map from `IndexMap` to a standard `HashMap`. There are a few issues. Can be re-implemented, but there are performance considerations: -...

@jonimake I've rebased this PR over master. There were quite a few changes, so you might want to reset your branch on [my branch's head](https://github.com/jmcomets/petgraph/tree/c87ac3a3d28be88e07f76cffef85151ea4392c63). Since it's been a while...

Yeah I just had a look & it seems to be the `--no-default-features` that I forgot to pass on its own. The `.travis.yml` has the full command-line that is run....

While the changes here could still use some refactoring, I think we only need to flip `no_std` to an `std` flag to keep things additive (should be enabled by default),...

You aren't the first to mention this. There's been a few issues / PRs on this subject and I'm on the fence on whether pathfinding algorithms should actually be in...

I've done some thinking on this issue and think we could benefit from some form of strongly-typed builder API. Something in the lines of: ```rust // compute all paths from...

I'm looking at the performance impact of this change right now. I'll submit a PR in a bit.

Another issue raised by @m4b: > Some more examples would be helpful, especially on how to use certain iterators, e.g.: https://docs.rs/petgraph/0.4.9/petgraph/graph/struct.Graph.html#method.node_indices > > would be nice to show stuff like....

I'm satisfied with this version, but I'd like a few opinions before this is merged.

@daiheitan Haven't had a look at this in a while, I'll check out your suggestion and see what I can do.