pandas-dev/pandas · error · ValueError

Value must be 1-D array-like or scalar, {type(value).__name_

Error message

Value must be 1-D array-like or scalar, {type(value).__name__} is not supported

What it means

Raised by IndexOpsMixin.searchsorted (pandas/core/base.py:1653) when the `value` argument is a 2-D object such as a DataFrame. searchsorted performs a binary search on a sorted 1-D array-like, so a rectangular DataFrame has no well-defined insertion point and is explicitly rejected before any lookup is attempted. Only 1-D array-likes (list, Series, Index, 1-D ndarray) or scalars are accepted.

Source

Thrown at pandas/core/base.py:1653

        If the values are not monotonically sorted, wrong locations
        may be returned:

        >>> ser = pd.Series([2, 1, 3])
        >>> ser
        0    2
        1    1
        2    3
        dtype: int64

        >>> ser.searchsorted(1)  # doctest: +SKIP
        0  # wrong result, correct would be 1
        """
        if isinstance(value, ABCDataFrame):
            msg = (
                "Value must be 1-D array-like or scalar, "
                f"{type(value).__name__} is not supported"
            )
            raise ValueError(msg)

        values = self._values
        if not isinstance(values, np.ndarray):
            # Going through EA.searchsorted directly improves performance GH#38083
            return values.searchsorted(value, side=side, sorter=sorter)

        return algorithms.searchsorted(
            values,
            value,
            side=side,
            sorter=sorter,
        )

    def drop_duplicates(self, *, keep: DropKeep = "first") -> Self:
        duplicated = self._duplicated(keep=keep)
        # error: Value of type "IndexOpsMixin" is not indexable
        return self[~duplicated]  # type: ignore[index]

View on GitHub (pinned to 71959b8cb9)

Solutions

1. Change the passed value to a 1-D structure: use `df['col']` (Series) or `df['col'].values` (1-D ndarray) instead of `df[['col']]`.
2. If you genuinely have multiple keys to locate, call searchsorted once per column in a loop or list comprehension, or use `Index.get_indexer` for vectorized lookups.
3. If you have a single value per row, squeeze the frame first: `df['col']` or `df.squeeze('columns')`.

Example fix

# before
idx.searchsorted(df[['date']])

# after
idx.searchsorted(df['date'])

import pandas as pd

def safe_searchsorted(index, value):
    if isinstance(value, pd.DataFrame):
        raise TypeError("searchsorted requires 1-D array-like or scalar, got DataFrame")
    return index.searchsorted(value)

import pandas as pd

def is_searchsortable(value) -> bool:
    return not isinstance(value, pd.DataFrame) and (
        pd.api.types.is_scalar(value) or getattr(value, 'ndim', 1) == 1
    )

try:
    pos = idx.searchsorted(value)
except ValueError as e:
    if 'not supported' in str(e):
        value = value.squeeze() if hasattr(value, 'squeeze') else value
        pos = idx.searchsorted(value)
    else:
        raise

When it happens

Trigger: Calling `idx.searchsorted(df)` or `ser.searchsorted(df)` where `df` is a pandas DataFrame. Also triggered indirectly when a function computes a searchsorted value and accidentally passes a DataFrame column-selection that returns a DataFrame (e.g. `df[['col']]` instead of `df['col']`).

Common situations: Selecting with double brackets `df[['col']]` (returns DataFrame) instead of single `df['col']` (returns Series) and feeding it into searchsorted. Constructing a target from `pd.concat(..., axis=1)` and forgetting to squeeze to one dimension.

Related errors

• Named aggregation is not supported when {axis=}.
• by_row={by_row} not allowed
• Cannot apply ufunc {ufunc} to mixed DataFrame and Series inp
• key must be an int or slice, got {type(key).__name__}
• name_or_index must be an int, str, bytes, pyarrow.compute.Ex

AI-assisted analysis of pandas-dev/pandas@71959b8cb9 (2026-08-07). Data as JSON: /api/errors/e47d285dd33e809f. Report an issue: GitHub.