Decoding data is not nesessary on some Redis configurations

[alex-eri]

What reason for this line here?

aiohttp-session/aiohttp_session/redis_storage.py

Line 53 in 606caca

data = data.decode('utf-8')

I want to store some bytes in session. Changed encoder setup_session(app, RedisStorage(redis_pool, encoder=pickle.dumps, decoder=pickle.loads)). And I have error on getting session:

    data = data.decode('utf-8')
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte

It also breakes msgpack

[alex-eri]

Its clear - python3.5 needs str, not bytes.

[asvetlov]

encoder and decoder are object <-> str converters by design.
There is a possibility to extend the contract but it requires abstract serializers, not just dropping .decode() call.

[NotSqrt]

I hit this bug when plugging the RedisStorage in my project, but my current pool uses: pool = await aioredis.create_redis_pool(redis_url, encoding='utf-8').
The encoding parameter already apply .decode('utf-8'), no need to call .decode('utf-8') again on a str.

[asvetlov]

@NotSqrt it's your configuration problem, not the library bug.

[NotSqrt]

So regroup decoding and unserializing ? decoder=lambda data: json.loads(data.decode('utf-8')), so that everyone can customize the decoder/encoder ?

[alex-eri]

@asvetlov redis eats bytes. Why we cant use bytes?
Python 3.6+ designed to use bytes too.

There is opton designed customizable, but that we cant customize it.

Simple if does not breaks anything. Why not?

Maybe param to init to provide decoder? Maybe split it into method?

[alex-eri]

@asvetlov

object <-> str

Here encoder designed as object -> bytes:

aiohttp-session/aiohttp_session/cookie_storage.py

Line 53 in fa10335

).encode('utf-8')

All *.dumps and *.loads i know supports loading bytes except python3.5`s json

[anyUesr]

I also meet this, please fix it quickly

thanks so much : )

[anyUesr]

Error like this:

File "/home/redis/.pyenv/versions/3.6.3/lib/python3.6/site-packages/aiohttp_session/__init__.py", line 113, in get_session
    session = await storage.load_session(request)
  File "/home/redis/.pyenv/versions/3.6.3/lib/python3.6/site-packages/aiohttp_session/redis_storage.py", line 53, in load_session
    data = data.decode('utf-8')
AttributeError: 'str' object has no attribute 'decode'

after I change it to

 53                 try:
 54                     data = data.decode('utf-8')
 55                 except AttributeError:
 56                     pass

it run OK

[anyUesr]

@asvetlov I agree @NotSqrt, maybe except decode error when data.decode('utf-8') is better

[NotSqrt]

This issue is about multiple problems:

• mine, where it's not necessary to re-decode data if the redis client is already configured to do the decoding
• @alex-eri , who wants to use serializers other than json, and where data must be in bytes, not decoded.

Ignoring the AttributeError will "fix" my problem, but not for @alex-eri

This is why I propose the possibility of customizing decoding + deserializing with a function parameter.
I used a lambda in #294 (comment), but it could be a proper function.