One last refactoring of the import handling:

* Create an import resolving stage. Currently this involves a copy of the ast,
  we can probably revisit this once we revisit doing full copies of the ast in
  general.

* Don't repeat work. Once we resolve a import, put it on a hash table and use
  it next time we are asked for it. This solves a O(n^2) behaviour in
  degenerated cases.

* Once import resolution is done, the target of an import is stored on the
  import itself.

0 files changed, 0 insertions, 0 deletions