tylerjereddy / hausdorff_scipy Goto Github PK

At the moment attempting to run nosetests test_hausdorff.py (at path: scipy/spatial/tests) produces the traceback shown below. This is probably related to my recent switch to Cython from the original python module.

======================================================================
ERROR: Failure: ImportError (No module named hausdorff)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/treddy/anaconda/lib/python2.7/site-packages/nose/loader.py", line 418, in loadTestsFromName
    addr.filename, addr.module)
  File "/Users/treddy/anaconda/lib/python2.7/site-packages/nose/importer.py", line 47, in importFromPath
    return self.importFromDir(dir_path, fqname)
  File "/Users/treddy/anaconda/lib/python2.7/site-packages/nose/importer.py", line 94, in importFromDir
    mod = load_module(part_fqname, fh, filename, desc)
  File "/Users/treddy/github_projects/scipy/scipy/spatial/tests/test_hausdorff.py", line 5, in <module>
    import scipy.spatial.hausdorff
ImportError: No module named hausdorff

The docstring for directed_hausdorff in scipy/spatial/hausdorff.pyx should be brought up to numpy docstring standards, probably to a format similar to those for distance measures in the distance.py spatial module.

We'll want to have some nice working examples, including one for the (more common) symmetric Hausdorff distance (which involves both forward and reverse directed Hausdorff calculations).

It will almost certainly be more reasonable to propose implementation of scipy.spatial.distance.directed_hausdorff rather than the current scipy.spatial.directed_hausdorff, so we should think about trying to make the import mechanics work that way.

I think this makes sense because there are already things like scipy.spatial.distance.euclidean and so on.