.. _privacy:

Privacy options
###############

filtering with threshold

:meth:`.CSeries.with_threshold`

.. code:: python

    import crandas as cd

    """Test setting a threshold on the output size of a filtering"""
    cdf = cd.DataFrame({"a": [5,4,7,8,2,1,0,9,3,6]})

    # Option 1: using CSeries.with_threshold

    assert len(cdf[(cdf["a"]<5).with_threshold(4)])==5
    assert len(cdf[(cdf["a"]<5).with_threshold(5)])==5
    with pytest.raises(cd.errors.ServerError) as exc:
        cdf[(cdf["a"]<5).with_threshold(6)]
    assert "threshold" in str(exc)

    # Option 2: using threshold parameter of CDataFrame.filter

    assert len(cdf.filter(cdf["a"]<5, threshold=4))==5
    assert len(cdf.filter(cdf["a"]<5, threshold=5))==5
    with pytest.raises(cd.errors.ServerError) as exc:
        assert len(cdf.filter(cdf["a"]<5, threshold=6))==5
    assert "threshold" in str(exc)



column-wise aggregation functions accept threshold parameters:

:meth:`.CSeries.sum`
:meth:`.CSeries.sum_squares`
:meth:`.CSeries.mean`
:meth:`.CSeries.var`
:meth:`.CSeries.count`
:meth:`.CSeries.max`
:meth:`.CSeries.min`

.. code:: python

    import crandas as cd
    import pandas as pd
    from crandas.errors import ServerError, ServerErrorCode
    import numpy
    import random
    import pytest
    from pathlib import Path


    """Tests whether the threshold parameter works as expected by attempting to
    perform an invalid ioeration"""
    n = 100
    numpy.random.seed(1)
    table = cd.DataFrame({"key": range(n), "value": numpy.random.randint(0, 100, size=n)})
    filtered = table["key"] == 5

    with pytest.raises(ServerError) as exc:
        table[filtered]["value"].sum(threshold=5)
    assert "threshold" in str(exc)