UnicodeDecodeError occurs when there is pickle binary data with decode_response=True

[xbynet]

the thing is that:

>>> from redis import StrictRedis
>>> r=StrictRedis(decode_responses=True)
>>> r.set('data1',{'aa':11})
>>> r.get('data1')
throw UnicodeDecodeError: 'utf-8' codec can't decode byte...

I know it can r.set('data1',json.dumps({'aa':11})) or use r.hmset('data1',{'aa':11}) to slove, but that is example to describe it. there is more complex scene like session pickle to redis or other meanwhile also want to use decode_responses=True.
but if i don't use decode_responses=True. i have to do manual decode() everytime.
So,what can i do if there is pickle data when use decode_responses=True.

My solve is change the connection.py about line 279：

       if isinstance(response, bytes) and self.encoding:
            import pickle
            from pickle import UnpicklingError
            try:
                response = pickle.loads(response)
            except UnpicklingError as e:
                response = response.decode(self.encoding)
        return response

so,can this merge into master or there is better resolution?

[andymccurdy]

decode_responses is designed to be used when you want to return data in a specific character set. It is not designed to automate type casting when dealing with dicts, lists, or your own data structures.

If you store both pickled data and encoded character set data (let's say UTF8) in your Redis instance, then I'd suggest creating two separate clients and connection pools to the same server:

redis_native_types = StrictRedis(decode_responses=True)
redis_pickled = StrictRedis(decode_responses=False)

Then use the appropriate client based on what data you're fetching/retrieving.

[xbynet]

thinks @andymccurdy . I see it. Your solution is very useful for me. thinks.