From 1f2f9fcfef92e26ca918ab761d08fff043774d83 Mon Sep 17 00:00:00 2001
From: nicolas-zimmermann <nicolaszimmermann.snv@gmail.com>
Date: Wed, 18 Sep 2019 17:09:32 +0200
Subject: [PATCH] added Conformations.small_kmedoids method to compute kmedoids

---
 src/projet8.py | 42 +++++++++++++++++++++++++++++++++++++-----
 1 file changed, 37 insertions(+), 5 deletions(-)

diff --git a/src/projet8.py b/src/projet8.py
index 8f1e702..c9464ee 100644
--- a/src/projet8.py
+++ b/src/projet8.py
@@ -36,8 +36,8 @@ class Conformations:
         """
         self.df = pd.DataFrame()
         self.filename = filename
-        self.identity_dist = None
-        self.dissimilarity_dist = None
+        self.identity_dist = np.array(None)
+        self.dissimilarity_dist = np.array(None)
         for chain_name, chain in pbx.chains_from_files([filename]): 
                 dihedrals = chain.get_phi_psi_angles() 
                 pb_seq = pbx.assign(dihedrals) 
@@ -104,9 +104,41 @@ class Conformations:
 
         return dist
     
-    #def small_kmedoid(self, ncluster):
-    #    pass
-    
+    def small_kmedoids(self, matrix, ncluster):
+        """
+        Returns clusters and medoids computed with kmedoids on a distance matrix
+        Arguments :
+            - matrix : str, ('identity' or 'dissimilarity')
+                       corresponding to the desired distance matrix to
+                       computed
+            - ncluster number of clusters to be computed
+        """
+        if matrix == 'identity':
+
+            matrix = self.identity_dist
+            if matrix.all() == None:
+                print("Error : distance matrix from identity hasn't been " \
+                      "computed yet")
+                return
+        elif matrix == 'dissimilarity':
+            matrix = self.dissimilarity_dist
+            if matrix.all() == None:
+                print("Error : distance matrix from dissimilarity hasn't " \
+                      "been computed yet")
+                return
+        
+        if ncluster > matrix.shape[0]:
+            print("Error : number of desired clusters > number of objects")
+            return
+
+        initial_medoids = np.random.randint(matrix.shape[0], size=ncluster)
+        kmed1 = kmedoids(matrix, initial_medoids, data_type='distance_matrix')
+        kmed1.process()
+
+        clusters = kmed1.get_clusters()
+        medoids = kmed1.get_medoids()
+
+        return (clusters, medoids)
 
 if __name__ == "__main__":
     if len(sys.argv) < 2: