Thanks. I merged and pushed the code.

I don't think so. If something significant goes wrong, run-all-tests should find it.

Thanks! I made the changes an updated the pull request again. Do we need
experiments for this?

I left some further comments on bitbucket.

Thanks! I handled them and updated the pull request.

I left some minor comments.

I created a pull request for this on Bitbucket:
https://bitbucket.org/FlorianPommerening/downward-issues/pull-requests/48

The rationale for TaskID (or whatever the name will be) is that we want to give
public access to an identify for a task proxy (for comparison, maps and
unordered_maps) without publicly exposing the internal AbstractTask *.

We decided to switch the PerTaskInformation introduced in issue774 to use
TaskProxy instead of AbstractTask* in its public interface. To make this
possible, we will give a get_id() method to TaskProxy that returns the pointer
wrapped in a class similar to StateID.