pandas-dev/pandas · error · ValueError

cannot broadcast result

Error message

cannot broadcast result

What it means

Raised inside `apply_broadcast` when a per-column UDF result is 1-D but its length does not equal `target.shape[0]` (the frame's row count). Broadcast mode places the result back into a column of fixed length, so a length mismatch is rejected rather than silently truncated or padded.

Solutions

1. Ensure the UDF returns exactly `len(df)` values per column (do not filter inside the UDF).
2. If filtering is needed, perform it before the apply or reindex the result to the original index inside the UDF.
3. Use `df.transform(func)` instead of `df.apply(func, result_type='broadcast')` if you want index-aligned same-length outputs — transform enforces the contract with a clearer message.

Example fix

// before
df.apply(lambda s: s.dropna(), result_type='broadcast')
// after
df.apply(lambda s: s.dropna().reindex(df.index), result_type='broadcast')

def broadcast_apply_length_safe(df, func):
    n = len(df)
    for col in df.columns:
        res = func(df[col].head(2))
        import numpy as np
        arr = np.asarray(res)
        if arr.ndim == 1 and len(arr) != n:
            raise ValueError(f'{func.__name__} returns length {len(arr)} for column {col!r}; expected {n}')
    return df.apply(func, result_type='broadcast')

def func_preserves_length(func, series) -> bool:
    import numpy as np
    try:
        arr = np.asarray(func(series))
        return arr.ndim <= 1 and len(arr) == len(series)
    except Exception:
        return False

try:
    out = df.apply(func, result_type='broadcast')
except ValueError as e:
    if 'cannot broadcast result' in str(e):
        out = df.transform(func)  # transform enforces same-length contract more clearly
    else:
        raise

When it happens

Trigger: `df.apply(func, result_type='broadcast')` where `func(df[col])` returns a 1-D array/Series whose `len(...)` differs from the number of rows. E.g. `df.apply(lambda s: s.dropna().values, result_type='broadcast')` (drops NaNs, shorter than input), or `lambda s: s.head(5)`.

Common situations: UDFs that filter (`dropna`, `drop_duplicates`, `head`), resample, or otherwise change length inside a broadcast apply; assuming broadcast will pad/align like a transform.

Related errors

• too many dims to broadcast
• by_row= not allowed
• Column is backed by an extension array, which is not…
• Column must have a numeric dtype. Found ' ' instead
• invalid value for result_type, must be one of

AI-assisted analysis of pandas-dev/pandas@3b7651241d (2026-08-11). Data as JSON: /api/errors/b5230d09cdd098e7. Report an issue: GitHub.

Appendix: source

Thrown at pandas/core/apply.py:1269

    def apply_broadcast(self, target: DataFrame) -> DataFrame:
        assert callable(self.func)

        result_values = np.empty_like(target.values)

        # axis which we want to compare compliance
        result_compare = target.shape[0]

        for i, col in enumerate(target.columns):
            res = self.func(target[col], *self.args, **self.kwargs)
            ares = np.asarray(res).ndim

            # must be a scalar or 1d
            if ares > 1:
                raise ValueError("too many dims to broadcast")
            if ares == 1:
                # must match return dim
                if result_compare != len(res):
                    raise ValueError("cannot broadcast result")

            result_values[:, i] = res

        # we *always* preserve the original index / columns
        result = self.obj._constructor(
            result_values, index=target.index, columns=target.columns
        )
        return result

    def apply_standard(self):
        if self.engine == "python":
            results, res_index = self.apply_series_generator()
        else:
            results, res_index = self.apply_series_numba()

        # wrap results
        return self.wrap_results(results, res_index)

View on GitHub (pinned to 3b7651241d)