pandas-dev/pandas · error · NotImplementedError

Parallel apply is not supported when raw=False and…

Error message

Parallel apply is not supported when raw=False and engine='numba'

What it means

Raised by apply_series_numba when engine='numba' is used with raw=False (the default, where the function receives a Series per row) and engine_kwargs contains {'parallel': True}. The numba backend's Series-based apply path has no parallel implementation, so pandas refuses to run rather than silently ignoring the parallel flag.

Solutions

Remove 'parallel' from engine_kwargs (default engine_kwargs={}, single-threaded numba).
If parallel execution is required, switch to engine='numba' with raw=True so the function receives numpy values and the parallel numba path applies.
Drop numba and use the default python engine, which supports the existing Series func unchanged.

Example fix

# before
df.apply(func, engine='numba', engine_kwargs={'parallel': True})
# after
df.apply(func, engine='numba')

Defensive patterns

Strategy: validation

Validate before calling

def safe_numba_apply(obj, func, **kwargs):
    ek = kwargs.get('engine_kwargs') or {}
    if kwargs.get('engine') == 'numba' and ek.get('parallel', False):
        raise ValueError("parallel=True is not supported with engine='numba' and raw=False; dropping parallel")
    return obj.apply(func, **kwargs)

Try / catch

try:
    df.apply(func, engine='numba', engine_kwargs=ek)
except NotImplementedError as e:
    if 'Parallel apply' in str(e):
        df.apply(func, engine='numba')  # fall back to single-threaded numba
    else:
        raise

Prevention

Default engine_kwargs to {} and only set parallel=True when also setting raw=True.
Document in code review that numba Series-apply is single-threaded only.

When it happens

Trigger: df.apply(func, engine='numba', engine_kwargs={'parallel': True}) where func expects a Series; Series.apply(func, engine='numba', engine_kwargs={'parallel': True}).

Common situations: Developers switch on numba expecting speed and additionally enable parallel=True (copied from a tutorial or numba.jit example) without realizing the Series apply path is single-threaded only.

Related errors

AI-assisted analysis of pandas-dev/pandas@3b7651241d (2026-08-11). Data as JSON: /api/errors/c29c30524a24fd60. Report an issue: GitHub.

Appendix: source

Thrown at pandas/core/apply.py:1307

        assert callable(self.func)

        series_gen = self.series_generator
        res_index = self.result_index

        results = {}

        for i, v in enumerate(series_gen):
            results[i] = self.func(v, *self.args, **self.kwargs)
            if isinstance(results[i], ABCSeries):
                # If we have a view on v, we need to make a copy because
                #  series_generator will swap out the underlying data
                results[i] = results[i].copy(deep=False)

        return results, res_index

    def apply_series_numba(self):
        if self.engine_kwargs.get("parallel", False):
            raise NotImplementedError(
                "Parallel apply is not supported when raw=False and engine='numba'"
            )
        if not self.obj.index.is_unique or not self.columns.is_unique:
            raise NotImplementedError(
                "The index/columns must be unique when raw=False and engine='numba'"
            )
        self.validate_values_for_numba()
        results = self.apply_with_numba()
        return results, self.result_index

    def wrap_results(self, results: ResType, res_index: Index) -> DataFrame | Series:
        from pandas import Series

        # see if we can infer the results
        if len(results) > 0 and 0 in results and is_sequence(results[0]):
            return self.wrap_results_for_axis(results, res_index)

        # dict of scalars

View on GitHub (pinned to 3b7651241d)