API reference¶

def is_discrete(obj: object, /) -> TypeIs[DiscreteDistribution]:
    """Whether ``obj`` is a discrete distribution.

    Narrowing guard: on the true branch a type checker sees ``obj`` as a
    ``DiscreteDistribution``, so ``pmf`` / ``log_pmf`` are reachable without a cast, and on the
    false branch it drops ``DiscreteDistribution`` from the type.

    Arguments:
        obj: Any object.

    Returns:
        ``True`` if ``obj`` is an instance of ``DiscreteDistribution``.

    Examples:
        >>> import polars_stats as ps
        >>> ps.is_discrete(ps.Binomial(10, 0.5))
        True
        >>> ps.is_discrete(ps.Normal())
        False

        Dispatching on the kind picks the density method that exists:

        >>> import polars as pl
        >>> def density(dist: ps.DiscreteDistribution | ps.ContinuousDistribution, value: str):
        ...     return dist.pmf(value) if ps.is_discrete(dist) else dist.pdf(value)
        >>> pl.DataFrame({"x": [0.0, 1.0]}).select(density(ps.Binomial(10, 0.5), "x"))
        shape: (2, 1)
        ┌──────────┐
        │ x        │
        │ ---      │
        │ f64      │
        ╞══════════╡
        │ 0.000977 │
        │ 0.009766 │
        └──────────┘
    """
    return isinstance(obj, DiscreteDistribution)

def is_continuous(obj: object, /) -> TypeIs[ContinuousDistribution]:
    """Whether ``obj`` is a continuous distribution.

    The counterpart of [`is_discrete`][polars_stats.is_discrete]: the two kinds are disjoint, so
    exactly one of the guards holds for any distribution in this package.

    Arguments:
        obj: Any object.

    Returns:
        ``True`` if ``obj`` is an instance of ``ContinuousDistribution``.

    Examples:
        >>> import polars_stats as ps
        >>> ps.is_continuous(ps.Normal())
        True
        >>> ps.is_continuous(ps.Binomial(10, 0.5))
        False

        The guard narrows, so ``pdf`` type checks inside the branch:

        >>> import polars as pl
        >>> dist = ps.Normal(mu=0.0, sigma=1.0)
        >>> if ps.is_continuous(dist):
        ...     print(round(pl.DataFrame({"x": [0.0]}).select(dist.pdf("x")).item(), 6))
        0.398942
    """
    return isinstance(obj, ContinuousDistribution)