`user_type` scalars have array-like functionality, and it's weird

[ofloveandhate]

In the EigenPy user_type that's compiled and tested as part of eigenpy, the scalar types end up having numpy-array-like functions, which is weird.

For example,

import user_type
a = user_type.CustomDouble(1) # makes a scalar.  

dir(a) # reveals a ton of functionality that I wouldn't expect for a scalar

For example, we get .flatten, .prod and .cumsum. This seems weird to me. Like, I don't think a Scalar should have these properties. Furthermore, many of these functions produce errors when called.

a.std() # error!

produces the following error message:

---------------------------------------------------------------------------
SystemError                               Traceback (most recent call last)
Cell In[12], line 1
----> 1 a.std()

File ~/env/b2/lib/python3.12/site-packages/numpy/_core/_methods.py:227, in _std(a, axis, dtype, out, ddof, keepdims, where, mean)
    225 def _std(a, axis=None, dtype=None, out=None, ddof=0, keepdims=False, *,
    226          where=True, mean=None):
--> 227     ret = _var(a, axis=axis, dtype=dtype, out=out, ddof=ddof,
    228                keepdims=keepdims, where=where, mean=mean)
    230     if isinstance(ret, mu.ndarray):
    231         ret = um.sqrt(ret, out=ret)

File ~/env/b2/lib/python3.12/site-packages/numpy/_core/_methods.py:173, in _var(a, axis, dtype, out, ddof, keepdims, where, mean)
    168     arrmean = mean
    169 else:
    170     # Compute the mean.
    171     # Note that if dtype is not of inexact type then arraymean will
    172     # not be either.
--> 173     arrmean = umr_sum(arr, axis, dtype, keepdims=True, where=where)
    174     # The shape of rcount has to match arrmean to not change the shape of
    175     # out in broadcasting. Otherwise, it cannot be stored back to arrmean.
    176     if rcount.ndim == 0:
    177         # fast-path for default case when where is True

SystemError: <built-in method reduce of numpy.ufunc object at 0x10768ac40> returned NULL without setting an exception

What am I not understanding? Thanks for your help!

[jcarpent]

I have followed the same strategy as the example provided by NumPy here.
We can further investigate if we really need to inherit from the GenericType of NumPy