QTable ignores dtype, tries converting string column with unit to Quantity

Description

I have a table that (ab)uses the “unit” metadata for some additional info on non-numeric strings, mostly by putting the String string there:

[…]
TTYPE10 = 'TEMPLATE'                                                            
TFORM10 = '30A     '                                                            
TUNIT10 = 'String  '                                                            
[…]

astropy.table.QTable.read() raises a type error when trying to read this file, so that I need to revert to Table, which gives other unexpected problems (#11521).

/usr/lib/python3/dist-packages/astropy/table/table.py in _convert_data_to_col(self, data, copy, default_name, dtype, name)
   1081             raise ValueError('unable to convert data to Column for Table')
   1082 
-> 1083         col = self._convert_col_for_table(col)
   1084 
   1085         return col

/usr/lib/python3/dist-packages/astropy/table/table.py in _convert_col_for_table(self, col)
   3649             # identified in the unit (such as u.mag()).
   3650             q_cls = getattr(col.unit, '_quantity_class', Quantity)
-> 3651             qcol = q_cls(col.data, col.unit, copy=False)
   3652             qcol.info = col.info
   3653             qcol.info.indices = col.info.indices

/usr/lib/python3/dist-packages/astropy/units/quantity.py in __new__(cls, value, unit, dtype, copy, order, subok, ndmin)
    380             not (value.dtype.kind == 'O' and
    381                  isinstance(value.item(0), numbers.Number))):
--> 382             raise TypeError("The value must be a valid Python or "
    383                             "Numpy numeric type.")
    384 

TypeError: The value must be a valid Python or Numpy numeric type.

It would be good if these columns could just kept as-is in case of non-numeric columns (where units make no sense anyway), at least with an additioal flag in QTable.read().

[olebole]

The following code shows it:

from astropy.table import Table, QTable

t = Table({'foo': list('*'*100)})  # just a sample string column
t['foo'].unit = 'String'
qt = QTable(t)                     # convert to QTable

which gives the exception shown above.

[mhvk]

See #11585 for a possible fix.