pandas-dev/pandas · error · NotImplementedError

the 'numba' engine doesn't support result_type='broadcast'

Error message

the 'numba' engine doesn't support result_type='broadcast'

What it means

Raised in `FrameApply.apply` when `result_type='broadcast'` is combined with `engine='numba'`. Broadcasting requires pandas to invoke the func per column and reshape the output to the original frame shape; this orchestration is not implemented for the numba engine, so pandas rejects it up front with NotImplementedError.

Solutions

Drop `engine='numba'` when you need `result_type='broadcast'`.
Drop `result_type='broadcast'` when you need numba, and reshape the output manually afterwards.
Rewrite the UDF to return a same-length Series natively, eliminating the need for broadcast.

Example fix

// before
df.apply(func, result_type='broadcast', engine='numba')
// after
df.apply(func, result_type='broadcast')  # python engine

Defensive patterns

Strategy: validation

Validate before calling

def frame_apply_broadcast(df, func, result_type=None, engine='python'):
    if result_type == 'broadcast' and engine == 'numba':
        raise ValueError('numba engine does not support result_type=broadcast; drop one')
    return df.apply(func, result_type=result_type, engine=engine)

Type guard

def numba_engine_supports_result_type(result_type) -> bool:
    return result_type != 'broadcast'

Try / catch

try:
    out = df.apply(func, result_type='broadcast', engine='numba')
except NotImplementedError as e:
    if 'numba' in str(e).lower() and 'broadcast' in str(e).lower():
        out = df.apply(func, result_type='broadcast')  # python engine
    else:
        raise

Prevention

Pick one: either numba or result_type='broadcast', never both.
Rewrite UDFs to return same-length Series natively to avoid broadcast entirely.
Add a project-level lint that flags the combination.

When it happens

Trigger: `df.apply(func, result_type='broadcast', engine='numba')`. The check `self.result_type == 'broadcast'` triggers the guard.

Common situations: Users enable numba for speed and also request broadcasting for a UDF that returns scalars per row; combining two performance-oriented options without realizing they are incompatible.

Related errors

AI-assisted analysis of pandas-dev/pandas@3b7651241d (2026-08-11). Data as JSON: /api/errors/4915527e12bfd874. Report an issue: GitHub.

Appendix: source

Thrown at pandas/core/apply.py:1048

                )
            return self.apply_str()

        # ufunc
        elif isinstance(self.func, np.ufunc):
            if self.engine == "numba":
                raise NotImplementedError(
                    "the 'numba' engine doesn't support "
                    "using a numpy ufunc as the callable function"
                )
            with np.errstate(all="ignore"):
                results = self.obj._mgr.apply("apply", func=self.func)
            # _constructor will retain self.index and self.columns
            return self.obj._constructor_from_mgr(results, axes=results.axes)

        # broadcasting
        if self.result_type == "broadcast":
            if self.engine == "numba":
                raise NotImplementedError(
                    "the 'numba' engine doesn't support result_type='broadcast'"
                )
            return self.apply_broadcast(self.obj)

        # one axis empty
        elif not all(self.obj.shape):
            return self.apply_empty_result()

        # raw
        elif self.raw:
            return self.apply_raw(engine=self.engine, engine_kwargs=self.engine_kwargs)

        return self.apply_standard()

    def agg(self):
        obj = self.obj
        axis = self.axis

View on GitHub (pinned to 3b7651241d)