Implement set_many function about cashews

[AIGeneratedUsername]

I am +1 on this.
I think it makes sense to add such a function, because now in my code I need to inherit from cashews and define missing functionality on my side:

   async def multi_set(self, pairs: Mapping[AnyKeyT, EncodableT], *, ttl: int | None = None) -> None:
        async with self._client.pipeline(transaction=True) as pipe:
            pairs = {self._build_key(k): v for k, v in pairs.items()}  # type: ignore[misc]
            await pipe.mset(pairs)  # type: ignore[arg-type]
            if ttl:
                for key in pairs.keys():
                    await pipe.expire(key, ttl)
            await pipe.execute()

[Krukov]

Hi there.

Yep, totally agree with set many, but I have some doubts about the decorator.
A first call it will set multi keys. It is easy, but what will happens if I call decorated function many times? Will It return cached results?

[yashh]

Thats right. The first call with set the multi keys while the subsequent would return from cache. The decorator would compute which of the keys are a hit, then compute the missing keys and then call the inner fn with the rest, set the cache and return.

Anyway set_many would be sufficient.

[Krukov]

Got it.
You are welcome to make a PR even with the decorator. I will be glad

[yashh]

master...yashh:cashews:master

Could you guide me a bit here. I have a WIP set_many working. Since I am not using the self._with_middlewares function set_many is not created backend._client attribute yet. If I call get / get_many that will initiate the connection everything works as expected. Am I on the right path?

[Krukov]

You are right about with_middlewares, but you are not on a right way. The problem that Wrapper - it a place to work with many backends and execute middewares - it is a some kid of proxy class for others "backends". And there are a few Backends: redis, memory and disk - they all are implementation of Backend interface. So I think that you should define a new method set_many as a part of the backend interface . And all magic with a pipeline should be only in the Redis backend

[yashh]

Take #2.
master...yashh:cashews:master

So I implemented set_many in redis, disk & memory backends. I see that redis-cli monitor is getting the right commands. However when I read the keys back with get_many right after set_many I get (None, None). I know redis is properly returning the right data. However

            _values = await self._with_middlewares("get_many", _keys[0])(*_keys, default=default)
            result.update(dict(zip(_keys, _values)))```

_values is (None, None). Any idea what it could be?

[Krukov]

@yashh
You on right way, LGFM. I think that the problem with Nones related to a pickle serialization.To solve it you need to implement set_many in PickleSerializerMixin and PickleSerializerMixin shouldn't know about Redis ). And to be honest I don't now how exactly you can implement it.

[yashh]

    async def set_many(self, pairs: Dict[str, Any], *args: Any, **kwargs: Any):
        data = {}
        for key, value in pairs.items():
            transformed_value = None
            if isinstance(value, int) and not isinstance(value, bool):
                transformed_value = pairs[key]
            else:
                transformed_value = self._pickler.dumps(value)
            data[key] = self._prepend_sign_to_value(key, transformed_value)
        return await super().set_many(data, *args, **kwargs)

Great. I implemented the set_many on the PickleSerializer and now get_many works! I was able to store strings, objects and other types.

[Krukov]

Great ! Waiting for your PR )