* Fixed bug #1727558

* Some more tests
* ValueError changed to ClusteringError where appropriate


git-svn-id: https://python-cluster.svn.sourceforge.net/svnroot/python-cluster/trunk@3 57eab859-f816-0410-af72-e61ffa1cc713

Author: exhuma

cluster.py

@@ -17,6 +17,9 @@
 
 from types import TupleType
 
+class ClusteringError(Exception):
+   pass
+
 def flatten(L):
    """
    Flattens a list.
@@ -632,6 +635,8 @@ def __init__(self, data, distance=None):
                              euclidian-distance algorithm on them.
       """
       self.__data = data
+      self.distance = distance
+      self.__initial_length = len(data)
 
       # test if each item is of same dimensions
       if len(data) > 1 and isinstance(data[0], TupleType):
@@ -654,14 +659,21 @@ def getclusters(self, n):
          n - The amount of clusters that should be generated.
              n must be greater than 1
       """
+
       # only proceed if we got sensible input
       if n <= 1:
-         raise ValueError("When clustering, you need to ask for at least two clusters! You asked for %d" % n)
+         raise ClusteringError("When clustering, you need to ask for at least two clusters! You asked for %d" % n)
 
       # return the data straight away if there is nothing to cluster
-      if self.__data == [] or len(self.__data) == 1:
+      if self.__data == [] or len(self.__data) == 1 or n == self.__initial_length:
          return self.__data
 
+      # It makes no sense to ask for more clusters than data-items available
+      if n > self.__initial_length:
+         raise ClusteringError( """Unable to generate more clusters than items 
+available. You supplied %d items, and asked for %d clusters.""" %
+               (self.__initial_length, n) )
+
       self.initialiseClusters(self.__data, n)
 
       items_moved = True     # tells us if any item moved between the clusters,

clusterTests.py

@@ -130,8 +130,13 @@ class KCluster2DTestCase(unittest.TestCase):
    def testClusterCount(self):
       "Test that asking for less than 2 clusters raises an error"
       cl = KMeansClustering([876, 123, 344, 676], distance=lambda x,y: abs(x-y))
-      self.assertRaises(ValueError, cl.getclusters, 0)
-      self.assertRaises(ValueError, cl.getclusters, 1)
+      self.assertRaises(ClusteringError, cl.getclusters, 0)
+      self.assertRaises(ClusteringError, cl.getclusters, 1)
+
+   def testNonsenseCluster(self):
+      "Test that asking for more clusters than data-items available raises an error"
+      cl = KMeansClustering([876, 123], distance=lambda x,y: abs(x-y))
+      self.assertRaises(ClusteringError, cl.getclusters, 5)
 
    def testUniformLength(self):
       "Test if there is an item in the cluster that has a different cardinality"
@@ -161,12 +166,25 @@ def testClustering(self):
             [[(8, 2), (8, 1), (8, 3), (7, 3), (9, 2), (9, 3)],
              [(3, 5), (1, 5), (3, 4), (2, 6), (2, 5), (3, 6)]])
 
+class KClusterSFBugs(unittest.TestCase):
+
+   def testLostFunctionReference(self):
+      "test for bug #1727558"
+      cl = KMeansClustering([(1,1), (20,40), (20,41)], lambda x,y:x+y)
+      clusters = cl.getclusters(3)
+      expected = [(1,1), (20,40), (20,41)]
+      self.assertTrue( compare_list(
+            clusters,
+            expected ),
+            "Elements differ!\n%s\n%s" % (clusters, expected))
+
 unittest.TextTestRunner(verbosity=2).run(
       unittest.TestSuite((
             unittest.makeSuite(HClusterSmallListTestCase),
             unittest.makeSuite(HClusterIntegerTestCase),
             unittest.makeSuite(HClusterStringTestCase),
             unittest.makeSuite(KClusterSmallListTestCase),
             unittest.makeSuite(KCluster2DTestCase),
+            unittest.makeSuite(KClusterSFBugs),
          ))
       )