MAINT: Don't use `__name__` to inspect type information

It makes much more sense to get this information directly from the typeinfo / dtype, rather than roundtripping it though a string.

The following code produces the same output (on 64-bit windows, python 3) before and after this change:

```
In [4]: {
   ...:     name: np.core.numerictypes.bitname(info.type)
   ...:     for name, info in sorted(np.core.numerictypes._concrete_typeinfo.items())
   ...: }
Out[4]:
{'bool': ('bool', 8, 'b1'),
 'byte': ('int', 8, 'i1'),
 'cdouble': ('complex', 128, 'c16'),
 'cfloat': ('complex', 64, 'c8'),
 'clongdouble': ('complex', 128, 'c16'),
 'datetime': ('datetime', 64, 'M8'),
 'double': ('float', 64, 'f8'),
 'float': ('float', 32, 'f4'),
 'half': ('float', 16, 'f2'),
 'int': ('int', 32, 'i4'),
 'intp': ('int', 64, 'i8'),
 'long': ('int', 32, 'i4'),
 'longdouble': ('float', 64, 'f8'),
 'longlong': ('int', 64, 'i8'),
 'object': ('object', 0, 'O'),
 'short': ('int', 16, 'i2'),
 'string': ('bytes', 0, 'S'),
 'timedelta': ('timedelta', 64, 'm8'),
 'ubyte': ('uint', 8, 'u1'),
 'uint': ('uint', 32, 'u4'),
 'uintp': ('uint', 64, 'u8'),
 'ulong': ('uint', 32, 'u4'),
 'ulonglong': ('uint', 64, 'u8'),
 'unicode': ('str', 0, 'U'),
 'ushort': ('uint', 16, 'u2'),
 'void': ('void', 0, 'V')}
```

Before this change, it was possible to call `bitname` on the abstract types, and get garbage results:
```
In [7]: {
   ...: name: np.core.numerictypes.bitname(t)
   ...: for name, t in sorted(np.core.numerictypes._abstract_types.items())
   ...: }
Out[7]:
{'character': ('character', 0, 'c'),
 'complexfloating': ('complexfloating', 0, 'c'),
 'flexible': ('flexible', 0, 'f'),
 'floating': ('floating', 0, 'f'),
 'generic': ('generic', 0, 'g'),
 'inexact': ('inexact', 0, 'i'),
 'integer': ('integer', 0, 'i'),
 'number': ('number', 0, 'n'),
 'signedinteger': ('signedinteger', 0, 's'),
 'unsignedinteger': ('unsignedinteger', 0, 'u')}
```

Now it raises a ValueError for all abstract types

0 files changed, 0 insertions, 0 deletions