pandas-dev/pandas · error · TypeError

datetime64 type does not support operation

Error message

datetime64 type does not support operation '{how}'

What it means

Raised in _groupby_op (the groupby/reduction dispatcher) when dtype.kind == 'M' (datetime64) and `how` is one of {'sum','prod','cumsum','cumprod','var','skew','kurt'}. Summing or multiplying timestamps is numerically meaningless, so pandas refuses with a per-operation message naming `how`.

Solutions

  1. Use .min()/.max()/.median()/.mean() (or .first()/.last()) for datetime reductions in groupby.
  2. Exclude datetime columns before numeric aggregations: df.select_dtypes(exclude='datetime').
  3. If you need a sum of elapsed time, subtract a reference first: (df[date] - df[date].min()).sum() to aggregate as a Timedelta.

Example fix

// before
df.groupby('k')['timestamp'].sum()  # TypeError

// after
df.groupby('k')['timestamp'].min()                                   # valid reduction
# or, sum of elapsed time:
df['elapsed'] = df['timestamp'] - df['timestamp'].min()
df.groupby('k')['elapsed'].sum()
Defensive patterns

Strategy: validation

Validate before calling

import pandas as pd

DISALLOWED_DT_OPS = {'sum', 'prod', 'cumsum', 'cumprod', 'var', 'skew', 'kurt'}

def groupby_dt_safe(df, key, col, how):
    if df[col].dtype.kind == 'M' and how in DISALLOWED_DT_OPS:
        raise TypeError(f'datetime64 does not support {how}; use min/max/median/mean or subtract a reference first')
    return getattr(df.groupby(key)[col], how)()

Type guard

import pandas as pd

def is_datetime64_series(s) -> bool:
    return getattr(s, 'dtype', None) is not None and s.dtype.kind == 'M'

Try / catch

try:
    return getattr(df.groupby('k')[date_col], how)()
except TypeError as e:
    if 'datetime64 type does not support' in str(e):
        return df.groupby('k')[date_col].min()
    raise

Prevention

When it happens

Trigger: df.groupby('key')[date_col].sum(); .prod(), .cumsum(), .cumprod(), .var(), .skew(), .kurt() on a datetime64 column; rolling/expanding windows dispatching the same ops.

Common situations: Generic .sum() across all numeric+datetime columns; migrating legacy aggregation pipelines that used to silently drop datetime columns.

Related errors

AI-assisted analysis of pandas-dev/pandas@3b7651241d (2026-08-11). Data as JSON: /api/errors/8998255087089b9b. Report an issue: GitHub.

Appendix: source

Thrown at pandas/core/arrays/datetimelike.py:1633

    # ------------------------------------------------------------------
    # GroupBy Methods

    def _groupby_op(
        self,
        *,
        how: str,
        has_dropped_na: bool,
        min_count: int,
        ngroups: int,
        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

View on GitHub (pinned to 3b7651241d)