pandas-dev/pandas · error · TypeError

Period type does not support

Error message

Period type does not support {how} operations

What it means

Raised inside DatetimeLikeArrayMixin._groupby_op when the underlying dtype is a PeriodDtype and the requested reduction 'how' is one of sum, prod, cumsum, cumprod, var, skew, or kurt. Periods represent ordinal positions on a fixed calendar grid; summing or multiplying them (or computing moments like variance/skew) has no defined meaning, so pandas refuses rather than producing a misleading numeric result.

Solutions

1. Cast the Period column to a timestamp with .dt.to_timestamp() before the aggregation if you actually want time arithmetic.
2. If you need a count-like reduction, use size() or count() instead of sum().
3. Convert the Period to its ordinal integer via .astype('int64') (view of the underlying ordinal) if you intentionally want raw ordinal arithmetic.
4. Drop the Period dtype (e.g. .astype(str) or .dt.to_timestamp()) before applying sum/prod/etc.

Example fix

// before
s = pd.Series(pd.PeriodIndex(['2020-01','2020-02'], freq='M'), name='p')
s.sum()  # TypeError: Period type does not support sum operations

// after
s.dt.to_timestamp().sum()  # or s.groupby(key).size() for counts

def safe_period_groupby(s, how):
    if pd.api.types.is_period_dtype(s):
        if how in {"sum","prod","cumsum","cumprod","var","skew","kurt"}:
            raise ValueError(f"{how} is unsupported on Period; convert to timestamp first")
    return s.groupby(level=0) if isinstance(s, pd.Series) else s

def is_period_reduction_unsupported(dtype, how) -> bool:
    return isinstance(dtype, pd.PeriodDtype) and how in {"sum","prod","cumsum","cumprod","var","skew","kurt"}

try:
    result = s.groupby(g).sum()
except TypeError as e:
    if "Period type does not support" in str(e):
        result = s.dt.to_timestamp().groupby(g).sum()
    else:
        raise

When it happens

Trigger: Calling groupby(...).sum()/.prod()/.cumsum()/.var()/.skew()/.kurt() (or DataFrame/Series reductions routed through the cython groupby engine) on a Series whose dtype is period[...] or a PeriodIndex; equivalently resample/rolling aggregations that map onto those how values over Period data.

Common situations: Loading monthly/quarterly data as Period columns and then running describe-style aggregations; migrating code from object/integer encodings to Period dtype and forgetting to strip it before summing; aggregating Period columns inside a groupby pipeline.

Related errors

• 'std' and 'sem' are not valid for PeriodDtype
• timedelta64 type does not support
• Cannot perform reduction
• Cannot perform reduction
• dtype is not specified and cannot be inferred

AI-assisted analysis of pandas-dev/pandas@3b7651241d (2026-08-11). Data as JSON: /api/errors/a7663c5c7ce63cd7. Report an issue: GitHub.

Appendix: source

Thrown at pandas/core/arrays/datetimelike.py:1644

        ids: npt.NDArray[np.intp],
        **kwargs,
    ):
        dtype = self.dtype
        if dtype.kind == "M":
            # Adding/multiplying datetimes is not valid
            if how in ["sum", "prod", "cumsum", "cumprod", "var", "skew", "kurt"]:
                raise TypeError(f"datetime64 type does not support operation '{how}'")
            if how in ["any", "all"]:
                # GH#34479
                raise TypeError(
                    f"'{how}' with datetime64 dtypes is no longer supported. "
                    f"Use (obj != pd.Timestamp(0)).{how}() instead."
                )

        elif isinstance(dtype, PeriodDtype):
            # Adding/multiplying Periods is not valid
            if how in ["sum", "prod", "cumsum", "cumprod", "var", "skew", "kurt"]:
                raise TypeError(f"Period type does not support {how} operations")
            if how in ["any", "all"]:
                # GH#34479
                raise TypeError(
                    f"'{how}' with PeriodDtype is no longer supported. "
                    f"Use (obj != pd.Period(ordinal=0, freq=freq)).{how}() instead."
                )
        # timedeltas we can add but not multiply
        elif how in ["prod", "cumprod", "skew", "kurt", "var"]:
            raise TypeError(f"timedelta64 type does not support {how} operations")

        # All of the functions implemented here are ordinal, so we can
        #  operate on the tz-naive equivalents
        npvalues = self._ndarray.view("M8[ns]")

        from pandas.core.groupby.ops import WrappedCythonOp

        kind = WrappedCythonOp.get_kind_from_how(how)
        op = WrappedCythonOp(how=how, kind=kind, has_dropped_na=has_dropped_na)

View on GitHub (pinned to 3b7651241d)