Make pickleable and allow data_array to be retrieved and set

[stevesimmons]

Here's my first go at making in-memory bloom filters pickleable.

• Adds a BloomFilter.data_array property to retrieve the bf array data.
• Adds an optional data_array input to the BloomFilter constructor. This is only valid if no filename is given. If used, the hash_seeds must also be provided.
• Add a reduce function so that the hash_seeds and data_array are saved when pickling.

I also added cases to add() and contains() for bytes. Without this, the bloomfilter doesn't work for byte inputs when saved/reopened in a new Python process.

Please can you take a look and provide feedback?
I should add tests :), have run out of time today.

[vmizg]

Hi @stevesimmons, thanks a lot for your contribution! I will see if I have a bit of time during this week to have a look at this.

[vmizg]

Could you move this to below the if (!array) check?

[prashnts]

If I'm not mistaken, this basically means:

class KlassX:
    pass

class KlassHashable:
    def __hash__(self):
        return 123

bf.add(KlassX()) # this won't work.
bf.add(KlassHashable()) # this will "work" but not in a way we want?

right?

[prashnts]

No problem copying the tests from your numba implementation at
https://github.com/stevesimmons/bloomfilter/blob/main/tests/test_bloomfilter.py#L8 ;)

[kylebd99]

Hey, I just wanted to second the merge on this pull request. I use pybloom with the multiprocessing library and this branch worked like a charm!

[123epsilon]

Any updates on getting this merged?

[vmizg]

I've addressed comments above and resolved conflicts in a separate PR here: #56

[prashnts]

Merged in #56, thanks everyone.

[prashnts]

I've just released this as v0.6.0 on pypi. 🎉