Python中實(shí)現(xiàn)參數(shù)類型檢查的簡(jiǎn)單方法

Python是一門弱類型語(yǔ)言，很多從C/C++轉(zhuǎn)過(guò)來(lái)的朋友起初不是很適應(yīng)。比如，在聲明一個(gè)函數(shù)時(shí)，不能指定參數(shù)的類型。用C做類比，那就是所有參數(shù)都是void*類型！void類型強(qiáng)制轉(zhuǎn)換在C++中被廣泛地認(rèn)為是個(gè)壞習(xí)慣，不到萬(wàn)不得已是不會(huì)使用的。
Python自然沒(méi)有類型強(qiáng)制轉(zhuǎn)換一說(shuō)了，因?yàn)樗莿?dòng)態(tài)語(yǔ)言。首先，所有對(duì)象都從Object繼承而來(lái)，其次，它有強(qiáng)大的內(nèi)省，如果調(diào)用某個(gè)不存在的方法會(huì)有異常拋出。大多數(shù)情況，我們都不需要做參數(shù)類型栓查，除了一些特殊情況。例如，某個(gè)函數(shù)接受一個(gè)str類型，結(jié)果在實(shí)際調(diào)用時(shí)傳入的是unicode，測(cè)試過(guò)程中又沒(méi)有代碼覆蓋到，這樣問(wèn)題就比較嚴(yán)重了。解決方法也很簡(jiǎn)單，借助Python的內(nèi)省，很容易就能判斷出參數(shù)的類型。但是每個(gè)地方都寫(xiě)檢查代碼會(huì)很累贅，何況它帶來(lái)的實(shí)際價(jià)值并不高。一個(gè)好的解決方法是使用裝飾器。    
'''
 >>> NONE, MEDIUM, STRONG = 0, 1, 2
 >>>
 >>> @accepts(int, int, int)
 ... def average(x, y, z):
 ...   return (x + y + z) / 2
 ...
 >>> average(5.5, 10, 15.0)
 TypeWarning: 'average' method accepts (int, int, int), but was given
 (float, int, float)
 15.25
'''
def accepts(*types, **kw):
  """ Function decorator. Checks that inputs given to decorated function
  are of the expected type.
 
  Parameters:
  types -- The expected types of the inputs to the decorated function.
       Must specify type for each parameter.
  kw  -- Optional specification of 'debug' level (this is the only valid
       keyword argument, no other should be given).
       debug = ( 0 | 1 | 2 )
 
  """
  if not kw:
    # default level: MEDIUM
    debug = 1
  else:
    debug = kw['debug']
  try:
    def decorator(f):
      def newf(*args):
        if debug == 0:
          return f(*args)
        assert len(args) == len(types)
        argtypes = tuple(map(type, args))
        if argtypes != types:
          msg = info(f.__name__, types, argtypes, 0)
          if debug == 1:
            print >> sys.stderr, 'TypeWarning: ', msg
          elif debug == 2:
            raise TypeError, msg
        return f(*args)
      newf.__name__ = f.__name__
      return newf
    return decorator
  except KeyError, key:
    raise KeyError, key + "is not a valid keyword argument"
  except TypeError, msg:
    raise TypeError, msg
 
def info(fname, expected, actual, flag):
  """ Convenience function returns nicely formatted error/warning msg. """
  format = lambda types: ', '.join([str(t).split("'")[1] for t in types])
  expected, actual = format(expected), format(actual)
  msg = "'%s' method " % fname \
     + ("accepts", "returns")[flag] + " (%s), but " % expected\
     + ("was given", "result is")[flag] + " (%s)" % actual
  return msg
本質(zhì)上講，這也是一種運(yùn)行時(shí)檢查，但效果已經(jīng)不錯(cuò)了