CODEWITHSUNDEEP
pythonhash
memecs
memecs

Reputation: 7574

Hashing python functions across different sessions

I want to generate a hash based on the parameters of the function (for some quick automated caching). The hash should be consistent between different sessions.

Let's assume I have the following functions:

def generate_hash(dictionary):
  return hashlib.sha224(str(dictionary.items())).hexdigest()

def foo(a,b,c):
 return generate_hash(locals())

Which is perfectly fine as long as the arguments str representation is consistent across sessions. The problem is that often I pass function as arguments and looks like they are not.

For example, the following call will return a different result across sessions.

foo(1,2,np.sum)

Any workaround?

Upvotes: 2

Views: 409

Answers (1)

memecs
memecs

Reputation: 7574

Seems like I figured something out myself: 

 16 def generate_hash(dictionary):
 17   args = []
 18   for key,item in dictionary.iteritems():
 19     if isinstance(item,functools.partial):
 20       args.append((key,item.func.__module__,item.func.__name__,                                                  
 21         generate_hash(item.keywords)))
 22     elif inspect.isfunction(item):
 23       args.append((key,item.__module__,item.__name__))
 24     else:
 25       args.append((key,item))
 26   return hashlib.sha224(str(args)).hexdigest()

which also works consistenly with partial functions:

foo('a','1',np.average)                         -> 1631c5fd0050fd01cb7a7ee9666d366b35c1415cb4181c7220ead043
foo('a',1,functools.partial(np.average,axis=0)) -> 692227d3b52b0cdcd4ed2204650cb207c1ab6f274a09977c711d35d5                                                            
foo('a',1,functools.partial(np.average,axis=1)) -> ba1e0b01f2e12ef1c9ca2e3bf5235aaadcbe4ab29d9c977e1ee6e799

Upvotes: 1

Related Questions