fix some warnings

momi/compute_sfs.py

@@ -455,7 +455,7 @@ def sum_trailing_antidiagonals(self):
     def convolve_trailing_axes(self, other):
         def within_pop_sfs(a, b):
             return a.sfs * b.liks[
-                [slice(None)] + [0] * len(b.pop_labels)]
+                tuple([slice(None)] + [0] * len(b.pop_labels))]
         self.sfs = within_pop_sfs(
             self, other) + within_pop_sfs(other, self)
 
@@ -506,9 +506,10 @@ def make_last_axis(self, pop):
 
     def add_last_axis_sfs(self, truncated_sfs):
         self.sfs = self.sfs + np.dot(
-            self.liks[[slice(None)] +
-                      [0] * (self.n_pops - 1) +
-                      [slice(None)]],
+            self.liks[tuple(
+                [slice(None)] +
+                [0] * (self.n_pops - 1) +
+                [slice(None)])],
             truncated_sfs)
 
     def get_last_axis_n(self):

momi/data/sfs.py

@@ -152,8 +152,8 @@ def __init__(self, loci, configs, folded, length):
                 self.loc_counts.append(np.array(cnts, dtype=float))
 
         if len(self.loc_idxs) > 1:
-            self._total_freqs = np.array(np.squeeze(np.asarray(
-                self.freqs_matrix.sum(axis=1))), ndmin=1)
+            self._total_freqs = self.freqs_matrix.dot(np.ones(self.n_loci))
+            assert self._total_freqs.shape == (self.freqs_matrix.shape[0],)
         else:
             # avoid costly building of frequency matrix, when there are many
             # Sfs's of a single locus (e.g. in many stochastic minibatches)

momi/demography.py

@@ -127,9 +127,9 @@ def sampled_n(self):
 
     @memoize_instance
     def _n_at_node(self, node):
-        return np.sum(self._G.node[(pop, idx)]['lineages']
-                      for pop, idx in nx.dfs_preorder_nodes(self._G, node)
-                      if idx == 0)
+        return sum(self._G.node[(pop, idx)]['lineages']
+                   for pop, idx in nx.dfs_preorder_nodes(self._G, node)
+                   if idx == 0)
 
     @property
     def _root(self):

momi/moran_model.py

@@ -38,6 +38,6 @@ def rate_matrix(n, sparse_format="csr"):
 
 @memoize
 def moran_eigensystem(n):
-    M = np.asarray(rate_matrix(n).todense())
+    M = rate_matrix(n).toarray()
     d, P = np.linalg.eig(M)
     return P, d, np.linalg.inv(P)