added support for numpy <1.8 (released 2013-10-28) which don't support

the axis argument to numpy.linalg.norm. this commit monkeypatches
a custom version when necessary which uses numpy.apply_along_axis()
to achieve the same effect.

Author: cdellin

src/prpy/ik_ranking.py

@@ -28,8 +28,20 @@
 # ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 # POSSIBILITY OF SUCH DAMAGE.
 
+from distutils.version import LooseVersion
 import numpy
 
+# handle older versions of numpy without norm axis argument
+if LooseVersion(numpy.__version__) < LooseVersion('1.8.0'):
+    oldnorm = numpy.linalg.norm
+    def newnorm(x, ord=None, axis=None):
+        if axis is None:
+            return oldnorm(x, ord=ord)
+        elif isinstance(axis, (int,long)):
+            return numpy.apply_along_axis(oldnorm, axis, x)
+        else:
+            raise NotImplementedError('older numpy without norm axis')
+    numpy.linalg.norm = newnorm
 
 def NoRanking(robot, ik_solutions):
     """