feat(polars): add bitwise operations (bit_and, bit_or, bit_xor)

[IndexSeek]

Description of changes

Add support for [bit_and], [bit_or], and [bit_xor] for the Polars backend. This uses bitwise_and, bitwise_or, and bitwise_xor Polars expression methods.

The is_in tests are failing still:
FAILED ibis/backends/tests/test_aggregation.py::test_reduction_ops[polars-is_in-bit_xor] - AssertionError:
FAILED ibis/backends/tests/test_aggregation.py::test_reduction_ops[polars-is_in-bit_and] - AssertionError:
FAILED ibis/backends/tests/test_aggregation.py::test_reduction_ops[polars-is_in-bit_or] - AssertionError:

I've tried to repro to see why they are failing, but I may be overlooking something. Here is how I repro'd:

In [1]: from ibis.interactive import *
In [2]: import numpy as np
In [3]: import pandas as pd
In [4]: ibis.set_backend("polars")
In [5]: alltypes = ibis.read_parquet("ci/ibis-testing-data/parquet/functional_alltypes.parquet")
In [6]: alltypes = alltypes.filter(_.id < 1550, _.string_col.isin(["1", "7"]))
In [7]: alltypes.agg(tmp=_.bigint_col.bit_or())
Out[7]: 
┏━━━━━━━┓
┃ tmp   ┃
┡━━━━━━━┩
│ int64 │
├───────┤
│    78 │
└───────┘
In [8]: np.bitwise_or.reduce(alltypes.execute().bigint_col).squeeze()
Out[8]: np.int64(78)

Here is what the test is reporting:

E           AssertionError: 
E           Not equal to tolerance rtol=0.001, atol=0
E           
E           Mismatched elements: 1 / 1 (100%)
E           Max absolute difference among violations: 48
E           Max relative difference among violations: 0.61538462
E            ACTUAL: array(126)
E            DESIRED: array(78)

ibis/backends/tests/test_aggregation.py:568: AssertionError
============================================== short test summary info ==============================================
FAILED ibis/backends/tests/test_aggregation.py::test_reduction_ops[polars-is_in-bit_or] - AssertionError: 

I'm wondering maybe if rows aren't being properly filtered or something.

Issues closed

[cpcloud]

@IndexSeek Thanks!

The issue was that you didn't thread the where argument through the rule.

I was able to solve the problem by reusing the _reductions mapping, which we use to construct reduction dispatch rules that have this bit of code which handles the filtering:

       if op.where is not None:
           arg = arg.filter(translate(op.where, **kw))

[IndexSeek]

@IndexSeek Thanks!

The issue was that you didn't thread the where argument through the rule.

I was able to solve the problem by reusing the _reductions mapping, which we use to construct reduction dispatch rules that have this bit of code which handles the filtering:

       if op.where is not None:
           arg = arg.filter(translate(op.where, **kw))

You're welcome! I suspected it had something to do with the filter, thanks for sharing that snippet. That will probably be useful in future ops.

Adding it to the _reductions dictionary is a much cleaner approach. ✨