Pandas Pinning methods to NaTType - Why not to _NaT?

In tslib.pyx there is an inheritance chain (rearranged here for python-friendliness)

cdef class _Timestamp(datetime):
[...]

cdef class _NaT(_Timestamp)
[...]

class NaTType(_NaT):
[...]
[...stuff I'm about to ask about...]

NaT = NaTType()

A bunch of attributes are pinned to NaTType that look like they could be pinned to _NaT just as easily. What's the reason for this? My understanding is that pinning things to _NaT should be more performant in general.

# inject the Timestamp field properties
# these by definition return np.nan
fields = ['year', 'quarter', 'month', 'day', 'hour',
          'minute', 'second', 'millisecond', 'microsecond', 'nanosecond',
          'week', 'dayofyear', 'weekofyear', 'days_in_month', 'daysinmonth',
          'dayofweek', 'weekday_name', 'days', 'seconds', 'microseconds',
          'nanoseconds', 'qyear', 'quarter']
for field in fields:
    prop = property(fget=lambda self: np.nan)
    setattr(NaTType, field, prop)

Comment From: jbrockmendel

Parking an actual bug here, will get around to it before long.

A few lines further down:

# GH9513 NaT methods (except to_datetime64) to raise, return np.nan, or
# return NaT create functions that raise, for binding to NaTType
for _maybe_method_name in dir(NaTType):
    _maybe_method = getattr(NaTType, _maybe_method_name)
    if (callable(_maybe_method)
        and not _maybe_method_name.startswith("_")
        and _maybe_method_name not in _implemented_methods):

        def _make_error_func(func_name):
            def f(*args, **kwargs):
                raise ValueError("NaTType does not support " + func_name)
            f.__name__ = func_name
            f.__doc__ = _get_docstring(func_name)
            return f

        setattr(NaTType, _maybe_method_name,
                _make_error_func(_maybe_method_name))

_get_docstring(_method_name) should be _get_docstring(_maybe_method_name). Demonstration:

>>> print(pd.NaT.total_seconds.__doc__)

        Total duration of timedelta in seconds (to ns precision)

>>> print(pd.NaT.ctime.__doc__)

        Total duration of timedelta in seconds (to ns precision)

Comment From: jreback

_NaT is a private cython class that is not exposed.

I guess you could define this for .ctime. All of the methods that are not explicitly defined are errors.

Comment From: jreback

if you want to make a specific issue of things to do on NaT that is fine. but pls don't mix issues.

Comment From: jbrockmendel

I guess you could define this for .ctime. All of the methods that are not explicitly defined are errors.

The code that defines this error-raising methods is also pinning docstrings to these methods. The methods are right, but it's mixing up the docstrings. Not a big deal, but easily fixed.

Comment From: jbrockmendel

if you want to make a specific issue of things to do on NaT that is fine. but pls don't mix issues.

Sure. Any preference over a new issue to list questions/todos as I find them vs renaming this one? I don't want to inundate you.