字典大小變化，爲什麼

Question

這裏的錯誤消息：

RuntimeError: dictionary changed size during iteration

這裏是我的代碼段（< =標誌着錯誤行）：

# Probability Distribution from a sequence of tuple tokens 
def probdist_from_tokens (tokens, N, V = 0, addone = False): 
    cfd = ConditionalFreqDist (tokens) 
    pdist = {} 

    for a in cfd: # <= line with the error 
     pdist[a] = {} 
     S = 1 + sum (1 for b in cfd[a] if cfd[a][b] == 1) 
     A = sum (cfd[a][b] for b in cfd[a]) 

     # Add the log probs. 
     for b in cfd[a]: 
      B = sum (cfd[b][c] for c in cfd[b]) 
      boff = ((B + 1)/(N + V)) if addone else (B/N) 
      pdist[a][b] = math.log ((cfd[a][b] + (S * boff))/(A + S)) 

     # Add OOV for tag if relevant 
     if addone: 
      boff = 1/(N + V) 
      pdist[a]["<OOV>"] = math.log ((S * boff)/(A + S)) 

    return pdist 

我基本上只是使用cfd作爲參考，將正確的值放在pdist中。我不想改變cfd，我只是想遍歷它的鍵和它的子字典的鍵。

我認爲問題是由設置變量A和B的行所引起的，當我在這些行上使用不同的代碼時遇到了同樣的錯誤，但是當我用常量替換它們時沒有得到錯誤值。

Answer 1

nltk.probability.ConditionalFreqDist繼承defaultdict，這意味着如果你讀一個不存在的條目cfd[b]，一個新的條目(b, FreqDist())將被插入到字典中，從而改變它的大小。問題的演示：

import collections 
d = collections.defaultdict(int, {'a': 1}) 
for k in d: 
    print(d['b']) 

輸出：

0 
Traceback (most recent call last): 
    File "1.py", line 4, in <module> 
    for k in d: 
RuntimeError: dictionary changed size during iteration 

---

所以，你應該檢查這行：

for b in cfd[a]: 
     B = sum (cfd[b][c] for c in cfd[b]) 

你確定的b關鍵確實存在於cfd？您可能需要將其更改爲

 B = sum(cfd[b].values()) if b in cfd else 0 
#        ^~~~~~~~~~~