Following  `from __future__ import unicode_literals`   the expression `eval(" 'foo' ")` will return a `unicode` instance.  However, using the same input, `ast.literal_eval(" 'foo' ")` will return a `str` instance. The caller's preference, that those plain single-quotes should a denote unicode literal, is respected by `eval()` but not by `ast.literal_eval()`.   I propose that `ast.literal_eval()` be made sensitive to this preference, to bring it in line with `eval()`.