Merge pull request #1197 from langner/master

Spruce up deterministic_cache, ecosystem and eigen

Author: meatballs

axelrod/ecosystem.py

@@ -1,15 +1,47 @@
+"""Tools for simulating population dynamics of immutable players.
+
+An ecosystem runs in the context of a previous tournament, and takes the
+results as input. That means no matches are run by the ecosystem, and a
+tournament needs to happen before it is created. For example:
+
+players = [axelrod.Cooperator(), axlerod.Defector()]
+tournament = axelrod.Tournament(players=players)
+results = tournament.play()
+ecosystem = axelrod.Ecosystem(results)
+ecosystem.reproduce(100)
+"""
+
 import random
-from axelrod.result_set import ResultSet
+
 from typing import List, Callable
 
+from axelrod.result_set import ResultSet
+
 
 class Ecosystem(object):
-    """Create an ecosystem based on the payoff matrix from an Axelrod
-    tournament."""
+    """An ecosystem based on the payoff matrix from a tournament.
+
+    Attributes
+    ----------
+    num_players: int
+        The number of players
+    """
 
     def __init__(self, results: ResultSet,
                  fitness: Callable[[float], float] = None,
                  population: List[int] = None) -> None:
+        """Create a new ecosystem.
+        
+        Parameters
+        ----------
+        results: ResultSet
+            The results of the tournament run beforehand to use.
+        fitness: List of callables
+            The reproduction rate at which populations reproduce.
+        population: List of ints.
+            The initial populations of the players, corresponding to the
+            payoff matrix in results.
+        """
 
         self.results = results
         self.num_players = self.results.num_players
@@ -45,7 +77,13 @@ def __init__(self, results: ResultSet,
             self.fitness = lambda p: p
 
     def reproduce(self, turns: int):
-
+        """Reproduce populations according to the payoff matrix.
+        
+        Parameters
+        ----------
+        turns: int
+            The number of turns to run.
+        """
         for iturn in range(turns):
             plist = list(range(self.num_players))
             pops = self.population_sizes[-1]

axelrod/eigen.py

@@ -9,21 +9,21 @@
 from typing import Tuple
 
 
-def normalise(nvec: numpy.ndarray) -> numpy.ndarray:
+def _normalise(nvec: numpy.ndarray) -> numpy.ndarray:
     """Normalises the given numpy array."""
     with numpy.errstate(invalid='ignore'):
         result = nvec / numpy.sqrt(numpy.dot(nvec, nvec))
     return result
 
 
-def squared_error(vector_1: numpy.ndarray, vector_2: numpy.ndarray) -> float:
+def _squared_error(vector_1: numpy.ndarray, vector_2: numpy.ndarray) -> float:
     """Computes the squared error between two numpy arrays."""
     diff = vector_1 - vector_2
     s = numpy.dot(diff, diff)
     return numpy.sqrt(s)
 
 
-def power_iteration(mat: numpy.matrix, initial: numpy.ndarray) -> numpy.ndarray:
+def _power_iteration(mat: numpy.matrix, initial: numpy.ndarray) -> numpy.ndarray:
     """
     Generator of successive approximations.
 
@@ -41,7 +41,7 @@ def power_iteration(mat: numpy.matrix, initial: numpy.ndarray) -> numpy.ndarray:
 
     vec = initial
     while True:
-        vec = normalise(numpy.dot(mat, vec))
+        vec = _normalise(numpy.dot(mat, vec))
         yield vec
 
 
@@ -60,6 +60,13 @@ def principal_eigenvector(mat: numpy.matrix, maximum_iterations=1000,
         The maximum number of iterations of the approximation
     max_error: float, 1e-8
         Exit criterion -- error threshold of the difference of successive steps
+
+    Returns
+    -------
+    ndarray
+        Eigenvector estimate for the input matrix
+    float
+        Eigenvalue corresonding to the returned eigenvector
     """
 
     mat_ = numpy.matrix(mat)
@@ -70,10 +77,10 @@ def principal_eigenvector(mat: numpy.matrix, maximum_iterations=1000,
     if not maximum_iterations:
         maximum_iterations = float('inf')
     last = initial
-    for i, vector in enumerate(power_iteration(mat, initial=initial)):
+    for i, vector in enumerate(_power_iteration(mat, initial=initial)):
         if i > maximum_iterations:
             break
-        if squared_error(vector, last) < max_error:
+        if _squared_error(vector, last) < max_error:
             break
         last = vector
     # Compute the eigenvalue (Rayleigh quotient)

axelrod/tests/unit/test_deterministic_cache.py

@@ -32,17 +32,15 @@ def setUp(self):
         self.cache = DeterministicCache()
 
     def test_basic_init(self):
-        cache = DeterministicCache()
-        self.assertTrue(cache.mutable)
+        self.assertTrue(self.cache.mutable)
 
     def test_init_from_file(self):
-        cache = DeterministicCache(file_name=self.test_load_file)
-        self.assertEqual(cache[self.test_key], self.test_value)
+        loaded_cache = DeterministicCache(file_name=self.test_load_file)
+        self.assertEqual(loaded_cache[self.test_key], self.test_value)
 
     def test_setitem(self):
-        cache = DeterministicCache()
-        cache[self.test_key] = self.test_value
-        self.assertEqual(cache[self.test_key], self.test_value)
+        self.cache[self.test_key] = self.test_value
+        self.assertEqual(self.cache[self.test_key], self.test_value)
 
     def test_setitem_invalid_key_not_tuple(self):
         invalid_key = 'test'
@@ -87,41 +85,35 @@ def test_setitem_invalid_key_stochastic_player(self):
             self.cache[invalid_key] = self.test_value        
 
     def test_setitem_invalid_value_not_list(self):
-        cache = DeterministicCache()
         with self.assertRaises(ValueError):
-            cache[self.test_key] = 5
+            self.cache[self.test_key] = 5
 
     def test_setitem_with_immutable_cache(self):
-        cache = DeterministicCache()
-        cache.mutable = False
+        self.cache.mutable = False
         with self.assertRaises(ValueError):
-            cache[self.test_key] = self.test_value
+            self.cache[self.test_key] = self.test_value
 
     def test_save(self):
-        cache = DeterministicCache()
-        cache[self.test_key] = self.test_value
-        cache.save(self.test_save_file)
+        self.cache[self.test_key] = self.test_value
+        self.cache.save(self.test_save_file)
         with open(self.test_save_file, 'rb') as f:
             text = f.read()
         self.assertEqual(text, self.test_pickle)
 
     def test_load(self):
-        cache = DeterministicCache()
-        cache.load(self.test_load_file)
-        self.assertEqual(cache[self.test_key], self.test_value)
+        self.cache.load(self.test_load_file)
+        self.assertEqual(self.cache[self.test_key], self.test_value)
 
     def test_load_error_for_inccorect_format(self):
         filename = "test_outputs/test.cache"
         with open(filename, 'wb') as io:
             pickle.dump(range(5), io)
 
         with self.assertRaises(ValueError):
-            cache = DeterministicCache()
-            cache.load(filename)
+            self.cache.load(filename)
 
     def test_del_item(self):
-        cache = DeterministicCache()
-        cache[self.test_key] = self.test_value
-        self.assertTrue(self.test_key in cache)
-        del cache[self.test_key]
-        self.assertFalse(self.test_key in cache)
+        self.cache[self.test_key] = self.test_value
+        self.assertTrue(self.test_key in self.cache)
+        del self.cache[self.test_key]
+        self.assertFalse(self.test_key in self.cache)

axelrod/tests/unit/test_ecosystem.py

@@ -24,10 +24,7 @@ def setUpClass(cls):
         cls.res_cooperators = cooperators.play()
         cls.res_defector_wins = defector_wins.play()
 
-    def test_init(self):
-        """Are the populations created correctly?"""
-
-        # By default create populations of equal size
+    def test_default_population_sizes(self):
         eco = axelrod.Ecosystem(self.res_cooperators)
         pops = eco.population_sizes
         self.assertEqual(eco.num_players, 4)
@@ -36,7 +33,7 @@ def test_init(self):
         self.assertAlmostEqual(sum(pops[0]), 1.0)
         self.assertEqual(list(set(pops[0])), [0.25])
 
-        # Can pass list of initial population distributions
+    def test_non_default_population_sizes(self):
         eco = axelrod.Ecosystem(self.res_cooperators, population=[.7, .25, .03, .02])
         pops = eco.population_sizes
         self.assertEqual(eco.num_players, 4)
@@ -45,7 +42,7 @@ def test_init(self):
         self.assertAlmostEqual(sum(pops[0]), 1.0)
         self.assertEqual(pops[0], [.7, .25, .03, .02])
 
-        # Distribution will automatically normalise
+    def test_population_normalization(self):
         eco = axelrod.Ecosystem(self.res_cooperators, population=[70, 25, 3, 2])
         pops = eco.population_sizes
         self.assertEqual(eco.num_players, 4)
@@ -54,22 +51,20 @@ def test_init(self):
         self.assertAlmostEqual(sum(pops[0]), 1.0)
         self.assertEqual(pops[0], [.7, .25, .03, .02])
 
-        # If passed list is of incorrect size get error
+    def test_results_and_population_of_different_sizes(self):
         self.assertRaises(TypeError, axelrod.Ecosystem, self.res_cooperators,
                           population=[.7, .2, .03, .1, .1])
 
-        # If passed list has negative values
+    def test_negative_populations(self):
         self.assertRaises(TypeError, axelrod.Ecosystem, self.res_cooperators,
                           population=[.7, -.2, .03, .2])
 
-    def test_fitness(self):
+    def test_fitness_function(self):
         fitness = lambda p: 2 * p
         eco = axelrod.Ecosystem(self.res_cooperators, fitness=fitness)
         self.assertTrue(eco.fitness(10), 20)
 
-    def test_cooperators(self):
-        """Are cooperators stable over time?"""
-
+    def test_cooperators_are_stable_over_time(self):
         eco = axelrod.Ecosystem(self.res_cooperators)
         eco.reproduce(100)
         pops = eco.population_sizes
@@ -79,9 +74,7 @@ def test_cooperators(self):
             self.assertEqual(sum(p), 1.0)
             self.assertEqual(list(set(p)), [0.25])
 
-    def test_defector_wins(self):
-        """Does one defector win over time?"""
-
+    def test_defector_wins_with_only_cooperators(self):
         eco = axelrod.Ecosystem(self.res_defector_wins)
         eco.reproduce(1000)
         pops = eco.population_sizes

axelrod/tests/unit/test_eigen.py

@@ -5,41 +5,37 @@
 import numpy
 from numpy.testing import assert_array_almost_equal
 
-from axelrod.eigen import normalise, principal_eigenvector
+from axelrod.eigen import _normalise, principal_eigenvector
 
 
 class FunctionCases(unittest.TestCase):
 
-    def test_eigen_1(self):
-        # Test identity matrices
+    def test_identity_matrices(self):
         for size in range(2, 6):
             mat = numpy.identity(size)
             evector, evalue = principal_eigenvector(mat)
             self.assertAlmostEqual(evalue, 1)
-            assert_array_almost_equal(evector, normalise(numpy.ones(size)))
+            assert_array_almost_equal(evector, _normalise(numpy.ones(size)))
 
-    def test_eigen_2(self):
-        # Test a 2x2 matrix
+    def test_2x2_matrix(self):
         mat = [[2, 1], [1, 2]]
         evector, evalue = principal_eigenvector(mat)
         self.assertAlmostEqual(evalue, 3)
         assert_array_almost_equal(evector, numpy.dot(mat, evector) / evalue)
-        assert_array_almost_equal(evector, normalise([1, 1]))
+        assert_array_almost_equal(evector, _normalise([1, 1]))
 
-    def test_eigen_3(self):
-        # Test a 3x3 matrix
+    def test_3x3_matrix(self):
         mat = [[1, 2, 0], [-2, 1, 2], [1, 3, 1]]
         evector, evalue = principal_eigenvector(mat, maximum_iterations=None,
                                                 max_error=1e-10)
         self.assertAlmostEqual(evalue, 3)
         assert_array_almost_equal(evector, numpy.dot(mat, evector) / evalue)
-        assert_array_almost_equal(evector, normalise([0.5, 0.5, 1]))
+        assert_array_almost_equal(evector, _normalise([0.5, 0.5, 1]))
 
-    def test_eigen_4(self):
-        # Test a 4x4 matrix
+    def test_4x4_matrix(self):
         mat = [[2, 0, 0, 0], [1, 2, 0, 0], [0, 1, 3, 0], [0, 0, 1, 3]]
         evector, evalue = principal_eigenvector(mat, maximum_iterations=None,
                                                 max_error=1e-10)
         self.assertAlmostEqual(evalue, 3, places=3)
         assert_array_almost_equal(evector, numpy.dot(mat, evector) / evalue)
-        assert_array_almost_equal(evector, normalise([0, 0, 0, 1]), decimal=4)
+        assert_array_almost_equal(evector, _normalise([0, 0, 0, 1]), decimal=4)