Class SetIndex

dask/dataframe/dask_expr/_shuffle.py:806–972

Abstract ``set_index`` class. Simplifies (later lowers) either to Blockwise ops if we are already sorted or to ``SetPartition`` which handles shuffling. Parameters ---------- frame: Expr Frame-like expression where the index is set. _other: Expr | Scalar Eit

class SetIndex(BaseSetIndexSortValues):
    """Abstract ``set_index`` class.

    Simplifies (later lowers) either to Blockwise ops if we are already sorted
    or to ``SetPartition`` which handles shuffling.

    Parameters
    ----------
    frame: Expr
        Frame-like expression where the index is set.
    _other: Expr | Scalar
        Either a Series-like expression to use as Index or a scalar defining the column.
    drop: bool
        Whether we drop the old column.
    sorted: str
        No need for shuffling if we are already sorted.
    user_divisions: int
        Divisions as passed by the user.
    upsample: float
        Used to increase the number of samples for quantiles.
    """

    _parameters = [
        "frame",
        "_other",
        "drop",
        "user_divisions",
        "partition_size",
        "ascending",
        "npartitions",
        "upsample",
        "shuffle_method",
        "append",
        "options",  # Options for the chosen shuffle method
    ]
    _defaults = {
        "drop": True,
        "user_divisions": None,
        "partition_size": 128e6,
        "ascending": True,
        "npartitions": None,
        "upsample": 1.0,
        "shuffle_method": None,
        "options": None,
        "append": False,
    }
    _filter_passthrough = True

    @property
    def _projection_columns(self):
        return self.columns + (
            [self._other] if not isinstance(self._other, Expr) else []
        )

    @functools.cached_property
    def _meta(self):
        if isinstance(self._other, Expr):
            other = self._other._meta