polars-python 0.52.0

use polars::lazy::dsl;
use polars::prelude::*;
use polars_plan::plans::DynLiteralValue;
use polars_plan::prelude::UnionArgs;
use polars_utils::python_function::PythonObject;
use pyo3::exceptions::{PyTypeError, PyValueError};
use pyo3::prelude::*;
use pyo3::types::{PyBool, PyBytes, PyFloat, PyInt, PyString};

use crate::conversion::any_value::py_object_to_any_value;
use crate::conversion::{Wrap, get_lf};
use crate::error::PyPolarsErr;
use crate::expr::ToExprs;
use crate::expr::datatype::PyDataTypeExpr;
use crate::lazyframe::PyOptFlags;
use crate::utils::EnterPolarsExt;
use crate::{PyDataFrame, PyExpr, PyLazyFrame, PySeries, map};

macro_rules! set_unwrapped_or_0 {
    ($($var:ident),+ $(,)?) => {
        $(let $var = $var.map(|e| e.inner).unwrap_or(dsl::lit(0));)+
    };
}

#[pyfunction]
pub fn rolling_corr(
    x: PyExpr,
    y: PyExpr,
    window_size: IdxSize,
    min_periods: IdxSize,
    ddof: u8,
) -> PyExpr {
    dsl::rolling_corr(
        x.inner,
        y.inner,
        RollingCovOptions {
            min_periods,
            window_size,
            ddof,
        },
    )
    .into()
}

#[pyfunction]
pub fn rolling_cov(
    x: PyExpr,
    y: PyExpr,
    window_size: IdxSize,
    min_periods: IdxSize,
    ddof: u8,
) -> PyExpr {
    dsl::rolling_cov(
        x.inner,
        y.inner,
        RollingCovOptions {
            min_periods,
            window_size,
            ddof,
        },
    )
    .into()
}

#[pyfunction]
pub fn arg_sort_by(
    by: Vec<PyExpr>,
    descending: Vec<bool>,
    nulls_last: Vec<bool>,
    multithreaded: bool,
    maintain_order: bool,
) -> PyExpr {
    let by = by.into_iter().map(|e| e.inner).collect::<Vec<Expr>>();
    dsl::arg_sort_by(
        by,
        SortMultipleOptions {
            descending,
            nulls_last,
            multithreaded,
            maintain_order,
            limit: None,
        },
    )
    .into()
}
#[pyfunction]
pub fn arg_where(condition: PyExpr) -> PyExpr {
    dsl::arg_where(condition.inner).into()
}

#[pyfunction]
pub fn as_struct(exprs: Vec<PyExpr>) -> PyResult<PyExpr> {
    let exprs = exprs.to_exprs();
    if exprs.is_empty() {
        return Err(PyValueError::new_err(
            "expected at least 1 expression in 'as_struct'",
        ));
    }
    Ok(dsl::as_struct(exprs).into())
}

#[pyfunction]
pub fn field(names: Vec<String>) -> PyExpr {
    dsl::Expr::Field(names.into_iter().map(|x| x.into()).collect()).into()
}

#[pyfunction]
pub fn coalesce(exprs: Vec<PyExpr>) -> PyExpr {
    let exprs = exprs.to_exprs();
    dsl::coalesce(&exprs).into()
}

#[pyfunction]
pub fn col(name: &str) -> PyExpr {
    dsl::col(name).into()
}

#[pyfunction]
pub fn element() -> PyExpr {
    dsl::element().into()
}

fn lfs_to_plans(lfs: Vec<PyLazyFrame>) -> Vec<DslPlan> {
    lfs.into_iter()
        .map(|lf| lf.ldf.into_inner().logical_plan)
        .collect()
}

#[pyfunction]
pub fn collect_all(
    lfs: Vec<PyLazyFrame>,
    engine: Wrap<Engine>,
    optflags: PyOptFlags,
    py: Python<'_>,
) -> PyResult<Vec<PyDataFrame>> {
    let plans = lfs_to_plans(lfs);
    let dfs = py.enter_polars(|| {
        LazyFrame::collect_all_with_engine(plans, engine.0, optflags.inner.into_inner())
    })?;
    Ok(dfs.into_iter().map(Into::into).collect())
}

#[pyfunction]
pub fn explain_all(lfs: Vec<PyLazyFrame>, optflags: PyOptFlags, py: Python) -> PyResult<String> {
    let plans = lfs_to_plans(lfs);
    let explained =
        py.enter_polars(|| LazyFrame::explain_all(plans, optflags.inner.into_inner()))?;
    Ok(explained)
}

#[pyfunction]
pub fn collect_all_with_callback(
    lfs: Vec<PyLazyFrame>,
    engine: Wrap<Engine>,
    optflags: PyOptFlags,
    lambda: Py<PyAny>,
    py: Python<'_>,
) {
    let plans = lfs
        .into_iter()
        .map(|lf| lf.ldf.into_inner().logical_plan)
        .collect();
    let result = py
        .enter_polars(|| {
            LazyFrame::collect_all_with_engine(plans, engine.0, optflags.inner.into_inner())
        })
        .map(|dfs| {
            dfs.into_iter()
                .map(Into::into)
                .collect::<Vec<PyDataFrame>>()
        });

    Python::attach(|py| match result {
        Ok(dfs) => {
            lambda.call1(py, (dfs,)).map_err(|err| err.restore(py)).ok();
        },
        Err(err) => {
            lambda
                .call1(py, (PyErr::from(err),))
                .map_err(|err| err.restore(py))
                .ok();
        },
    })
}

#[pyfunction]
pub fn concat_lf(
    seq: &Bound<'_, PyAny>,
    rechunk: bool,
    parallel: bool,
    to_supertypes: bool,
    maintain_order: bool,
) -> PyResult<PyLazyFrame> {
    let len = seq.len()?;
    let mut lfs = Vec::with_capacity(len);

    for res in seq.try_iter()? {
        let item = res?;
        let lf = get_lf(&item)?;
        lfs.push(lf);
    }

    let lf = dsl::concat(
        lfs,
        UnionArgs {
            rechunk,
            parallel,
            to_supertypes,
            maintain_order,
            ..Default::default()
        },
    )
    .map_err(PyPolarsErr::from)?;
    Ok(lf.into())
}

#[pyfunction]
pub fn concat_list(s: Vec<PyExpr>) -> PyResult<PyExpr> {
    let s = s.into_iter().map(|e| e.inner).collect::<Vec<_>>();
    let expr = dsl::concat_list(s).map_err(PyPolarsErr::from)?;
    Ok(expr.into())
}

#[pyfunction]
pub fn concat_arr(s: Vec<PyExpr>) -> PyResult<PyExpr> {
    let s = s.into_iter().map(|e| e.inner).collect::<Vec<_>>();
    let expr = dsl::concat_arr(s).map_err(PyPolarsErr::from)?;
    Ok(expr.into())
}

#[pyfunction]
pub fn concat_str(s: Vec<PyExpr>, separator: &str, ignore_nulls: bool) -> PyExpr {
    let s = s.into_iter().map(|e| e.inner).collect::<Vec<_>>();
    dsl::concat_str(s, separator, ignore_nulls).into()
}

#[pyfunction]
pub fn len() -> PyExpr {
    dsl::len().into()
}

#[pyfunction]
pub fn cov(a: PyExpr, b: PyExpr, ddof: u8) -> PyExpr {
    dsl::cov(a.inner, b.inner, ddof).into()
}

#[pyfunction]
#[cfg(feature = "trigonometry")]
pub fn arctan2(y: PyExpr, x: PyExpr) -> PyExpr {
    y.inner.arctan2(x.inner).into()
}

#[pyfunction]
pub fn cum_fold(
    acc: PyExpr,
    lambda: Py<PyAny>,
    exprs: Vec<PyExpr>,
    returns_scalar: bool,
    return_dtype: Option<PyDataTypeExpr>,
    include_init: bool,
) -> PyExpr {
    let exprs = exprs.to_exprs();
    let func = PlanCallback::new_python(PythonObject(lambda));
    dsl::cum_fold_exprs(
        acc.inner,
        func,
        exprs,
        returns_scalar,
        return_dtype.map(|v| v.inner),
        include_init,
    )
    .into()
}

#[pyfunction]
pub fn cum_reduce(
    lambda: Py<PyAny>,
    exprs: Vec<PyExpr>,
    returns_scalar: bool,
    return_dtype: Option<PyDataTypeExpr>,
) -> PyExpr {
    let exprs = exprs.to_exprs();

    let func = PlanCallback::new_python(PythonObject(lambda));
    dsl::cum_reduce_exprs(func, exprs, returns_scalar, return_dtype.map(|v| v.inner)).into()
}

#[pyfunction]
#[pyo3(signature = (year, month, day, hour=None, minute=None, second=None, microsecond=None, time_unit=Wrap(TimeUnit::Microseconds), time_zone=Wrap(None), ambiguous=PyExpr::from(dsl::lit(String::from("raise")))))]
pub fn datetime(
    year: PyExpr,
    month: PyExpr,
    day: PyExpr,
    hour: Option<PyExpr>,
    minute: Option<PyExpr>,
    second: Option<PyExpr>,
    microsecond: Option<PyExpr>,
    time_unit: Wrap<TimeUnit>,
    time_zone: Wrap<Option<TimeZone>>,
    ambiguous: PyExpr,
) -> PyExpr {
    let year = year.inner;
    let month = month.inner;
    let day = day.inner;
    set_unwrapped_or_0!(hour, minute, second, microsecond);
    let ambiguous = ambiguous.inner;
    let time_unit = time_unit.0;
    let time_zone = time_zone.0;
    let args = DatetimeArgs {
        year,
        month,
        day,
        hour,
        minute,
        second,
        microsecond,
        time_unit,
        time_zone,
        ambiguous,
    };
    dsl::datetime(args).into()
}

#[pyfunction]
pub fn concat_lf_diagonal(
    lfs: &Bound<'_, PyAny>,
    rechunk: bool,
    parallel: bool,
    to_supertypes: bool,
    maintain_order: bool,
) -> PyResult<PyLazyFrame> {
    let iter = lfs.try_iter()?;

    let lfs = iter
        .map(|item| {
            let item = item?;
            get_lf(&item)
        })
        .collect::<PyResult<Vec<_>>>()?;

    let lf = dsl::functions::concat_lf_diagonal(
        lfs,
        UnionArgs {
            rechunk,
            parallel,
            to_supertypes,
            maintain_order,
            ..Default::default()
        },
    )
    .map_err(PyPolarsErr::from)?;
    Ok(lf.into())
}

#[pyfunction]
pub fn concat_lf_horizontal(lfs: &Bound<'_, PyAny>, parallel: bool) -> PyResult<PyLazyFrame> {
    let iter = lfs.try_iter()?;

    let lfs = iter
        .map(|item| {
            let item = item?;
            get_lf(&item)
        })
        .collect::<PyResult<Vec<_>>>()?;

    let args = UnionArgs {
        rechunk: false, // No need to rechunk with horizontal concatenation
        parallel,
        to_supertypes: false,
        ..Default::default()
    };
    let lf = dsl::functions::concat_lf_horizontal(lfs, args).map_err(PyPolarsErr::from)?;
    Ok(lf.into())
}

#[pyfunction]
pub fn concat_expr(e: Vec<PyExpr>, rechunk: bool) -> PyResult<PyExpr> {
    let e = e.to_exprs();
    let e = dsl::functions::concat_expr(e, rechunk).map_err(PyPolarsErr::from)?;
    Ok(e.into())
}

#[pyfunction]
#[pyo3(signature = (weeks, days, hours, minutes, seconds, milliseconds, microseconds, nanoseconds, time_unit))]
pub fn duration(
    weeks: Option<PyExpr>,
    days: Option<PyExpr>,
    hours: Option<PyExpr>,
    minutes: Option<PyExpr>,
    seconds: Option<PyExpr>,
    milliseconds: Option<PyExpr>,
    microseconds: Option<PyExpr>,
    nanoseconds: Option<PyExpr>,
    time_unit: Wrap<TimeUnit>,
) -> PyExpr {
    set_unwrapped_or_0!(
        weeks,
        days,
        hours,
        minutes,
        seconds,
        milliseconds,
        microseconds,
        nanoseconds,
    );
    let args = DurationArgs {
        weeks,
        days,
        hours,
        minutes,
        seconds,
        milliseconds,
        microseconds,
        nanoseconds,
        time_unit: time_unit.0,
    };
    dsl::duration(args).into()
}

#[pyfunction]
pub fn fold(
    acc: PyExpr,
    lambda: Py<PyAny>,
    exprs: Vec<PyExpr>,
    returns_scalar: bool,
    return_dtype: Option<PyDataTypeExpr>,
) -> PyExpr {
    let exprs = exprs.to_exprs();
    let func = PlanCallback::new_python(PythonObject(lambda));
    dsl::fold_exprs(
        acc.inner,
        func,
        exprs,
        returns_scalar,
        return_dtype.map(|w| w.inner),
    )
    .into()
}

#[pyfunction]
pub fn lit(value: &Bound<'_, PyAny>, allow_object: bool, is_scalar: bool) -> PyResult<PyExpr> {
    let py = value.py();
    if value.is_instance_of::<PyBool>() {
        let val = value.extract::<bool>()?;
        Ok(dsl::lit(val).into())
    } else if let Ok(int) = value.downcast::<PyInt>() {
        let v = int
            .extract::<i128>()
            .map_err(|e| polars_err!(InvalidOperation: "integer too large for Polars: {e}"))
            .map_err(PyPolarsErr::from)?;
        Ok(Expr::Literal(LiteralValue::Dyn(DynLiteralValue::Int(v))).into())
    } else if let Ok(float) = value.downcast::<PyFloat>() {
        let val = float.extract::<f64>()?;
        Ok(Expr::Literal(LiteralValue::Dyn(DynLiteralValue::Float(val))).into())
    } else if let Ok(pystr) = value.downcast::<PyString>() {
        Ok(dsl::lit(pystr.to_string()).into())
    } else if let Ok(series) = value.extract::<PySeries>() {
        let s = series.series.into_inner();
        if is_scalar {
            let av = s
                .get(0)
                .map_err(|_| PyValueError::new_err("expected at least 1 value"))?;
            let av = av.into_static();
            Ok(dsl::lit(Scalar::new(s.dtype().clone(), av)).into())
        } else {
            Ok(dsl::lit(s).into())
        }
    } else if value.is_none() {
        Ok(dsl::lit(Null {}).into())
    } else if let Ok(value) = value.downcast::<PyBytes>() {
        Ok(dsl::lit(value.as_bytes()).into())
    } else {
        let av = py_object_to_any_value(value, true, allow_object).map_err(|_| {
            PyTypeError::new_err(
                format!(
                    "cannot create expression literal for value of type {}.\
                    \n\nHint: Pass `allow_object=True` to accept any value and create a literal of type Object.",
                    value.get_type().qualname().map(|s|s.to_string()).unwrap_or("unknown".to_owned()),
                )
            )
        })?;
        match av {
            #[cfg(feature = "object")]
            AnyValue::ObjectOwned(_) => {
                let s = PySeries::new_object(py, "", vec![value.extract()?], false)
                    .series
                    .into_inner();
                Ok(dsl::lit(s).into())
            },
            _ => Ok(Expr::Literal(LiteralValue::from(av)).into()),
        }
    }
}

#[pyfunction]
#[pyo3(signature = (pyexpr, lambda, output_type, is_elementwise, returns_scalar))]
pub fn map_expr(
    pyexpr: Vec<PyExpr>,
    lambda: Py<PyAny>,
    output_type: Option<PyDataTypeExpr>,
    is_elementwise: bool,
    returns_scalar: bool,
) -> PyExpr {
    map::lazy::map_expr(&pyexpr, lambda, output_type, is_elementwise, returns_scalar)
}

#[pyfunction]
pub fn pearson_corr(a: PyExpr, b: PyExpr) -> PyExpr {
    dsl::pearson_corr(a.inner, b.inner).into()
}

#[pyfunction]
pub fn reduce(
    lambda: Py<PyAny>,
    exprs: Vec<PyExpr>,
    returns_scalar: bool,
    return_dtype: Option<PyDataTypeExpr>,
) -> PyExpr {
    let exprs = exprs.to_exprs();
    let func = PlanCallback::new_python(PythonObject(lambda));
    dsl::reduce_exprs(func, exprs, returns_scalar, return_dtype.map(|v| v.inner)).into()
}

#[pyfunction]
#[pyo3(signature = (value, n, dtype=None))]
pub fn repeat(value: PyExpr, n: PyExpr, dtype: Option<Wrap<DataType>>) -> PyExpr {
    let mut value = value.inner;
    let n = n.inner;

    if let Some(dtype) = dtype {
        value = value.cast(dtype.0);
    }

    dsl::repeat(value, n).into()
}

#[pyfunction]
pub fn spearman_rank_corr(a: PyExpr, b: PyExpr, propagate_nans: bool) -> PyExpr {
    #[cfg(feature = "propagate_nans")]
    {
        dsl::spearman_rank_corr(a.inner, b.inner, propagate_nans).into()
    }
    #[cfg(not(feature = "propagate_nans"))]
    {
        panic!("activate 'propagate_nans'")
    }
}

#[pyfunction]
#[cfg(feature = "sql")]
pub fn sql_expr(sql: &str) -> PyResult<PyExpr> {
    let expr = polars::sql::sql_expr(sql).map_err(PyPolarsErr::from)?;
    Ok(expr.into())
}