重定义森林火灾模拟

在前面的例子中,我们定义了一个 BurnableForest,实现了一个循序渐进的生长和燃烧过程。

假设我们现在想要定义一个立即燃烧的过程(每次着火之后燃烧到不能燃烧为止,之后再生长,而不是每次只燃烧周围的一圈树木),由于燃烧过程不同,我们需要从 BurnableForest 中派生出两个新的子类 SlowBurnForest(原来的燃烧过程) 和 InsantBurnForest,为此

  • BurnableForest 中的 burn_trees() 方法改写,不做任何操作,直接 pass(因为在 advance_one_step() 中调用了它,所以不能直接去掉)
  • 在两个子类中定义新的 burn_trees() 方法。

In [1]:

  1. import numpy as np
  2. from scipy.ndimage.measurements import label
  3.  
  4. class Forest(object):
  5. """ Forest can grow trees which eventually die."""
  6. def __init__(self, size=(150,150), p_sapling=0.0025):
  7. self.size = size
  8. self.trees = np.zeros(self.size, dtype=bool)
  9. self.p_sapling = p_sapling
  10.  
  11. def __repr__(self):
  12. my_repr = "{}(size={})".format(self.__class__.__name__, self.size)
  13. return my_repr
  14.  
  15. def __str__(self):
  16. return self.__class__.__name__
  17.  
  18. @property
  19. def num_cells(self):
  20. """Number of cells available for growing trees"""
  21. return np.prod(self.size)
  22.  
  23. @property
  24. def tree_fraction(self):
  25. """
  26. Fraction of trees
  27. """
  28. num_trees = self.trees.sum()
  29. return float(num_trees) / self.num_cells
  30.  
  31. def _rand_bool(self, p):
  32. """
  33. Random boolean distributed according to p, less than p will be True
  34. """
  35. return np.random.uniform(size=self.trees.shape) < p
  36.  
  37. def grow_trees(self):
  38. """
  39. Growing trees.
  40. """
  41. growth_sites = self._rand_bool(self.p_sapling)
  42. self.trees[growth_sites] = True
  43.  
  44. def advance_one_step(self):
  45. """
  46. Advance one step
  47. """
  48. self.grow_trees()
  49.  
  50. class BurnableForest(Forest):
  51. """
  52. Burnable forest support fires
  53. """
  54. def __init__(self, p_lightning=5.0e-6, **kwargs):
  55. super(BurnableForest, self).__init__(**kwargs)
  56. self.p_lightning = p_lightning
  57. self.fires = np.zeros((self.size), dtype=bool)
  58.  
  59. def advance_one_step(self):
  60. """
  61. Advance one step
  62. """
  63. super(BurnableForest, self).advance_one_step()
  64. self.start_fires()
  65. self.burn_trees()
  66.  
  67. @property
  68. def fire_fraction(self):
  69. """
  70. Fraction of fires
  71. """
  72. num_fires = self.fires.sum()
  73. return float(num_fires) / self.num_cells
  74.  
  75. def start_fires(self):
  76. """
  77. Start of fire.
  78. """
  79. lightning_strikes = (self._rand_bool(self.p_lightning) &
  80. self.trees)
  81. self.fires[lightning_strikes] = True
  82.  
  83. def burn_trees(self):
  84. pass
  85.  
  86. class SlowBurnForest(BurnableForest):
  87. def burn_trees(self):
  88. """
  89. Burn trees.
  90. """
  91. fires = np.zeros((self.size[0] + 2, self.size[1] + 2), dtype=bool)
  92. fires[1:-1, 1:-1] = self.fires
  93. north = fires[:-2, 1:-1]
  94. south = fires[2:, 1:-1]
  95. east = fires[1:-1, :-2]
  96. west = fires[1:-1, 2:]
  97. new_fires = (north | south | east | west) & self.trees
  98. self.trees[self.fires] = False
  99. self.fires = new_fires
  100.  
  101. class InstantBurnForest(BurnableForest):
  102. def burn_trees(self):
  103. # 起火点
  104. strikes = self.fires
  105. # 找到连通区域
  106. groves, num_groves = label(self.trees)
  107. fires = set(groves[strikes])
  108. self.fires.fill(False)
  109. # 将与着火点相连的区域都烧掉
  110. for fire in fires:
  111. self.fires[groves == fire] = True
  112. self.trees[self.fires] = False
  113. self.fires.fill(False)

测试:

In [2]:

  1. forest = Forest()
  2. sb_forest = SlowBurnForest()
  3. ib_forest = InstantBurnForest()
  4.  
  5. forests = [forest, sb_forest, ib_forest]
  6.  
  7. tree_history = []
  8.  
  9. for i in xrange(1500):
  10. for fst in forests:
  11. fst.advance_one_step()
  12. tree_history.append(tuple(fst.tree_fraction for fst in forests))

显示结果:

In [3]:

  1. import matplotlib.pyplot as plt
  2. %matplotlib inline
  3.  
  4. plt.figure(figsize=(10,6))
  5.  
  6. plt.plot(tree_history)
  7. plt.legend([f.__str__() for f in forests])
  8.  
  9. plt.show()

08.10 重定义森林火灾模拟 - 图1

原文: https://nbviewer.jupyter.org/github/lijin-THU/notes-python/blob/master/08-object-oriented-programming/08.10-refactoring-the-forest-fire-simutation.ipynb