This maybe is just a doc warning - but caught me by surprise. I would prefer the read_csv behavior, but it would be backwards incompatible.

In [57]: pd.read_csv('path/doesn_exist')
---------------------------------------------------------------------------

IOError: File path/doesn_exist does not exist

In [56]: pd.read_msgpack('path/doesn_exist')
Out[56]: [112, 97, 116, 104, 47, 100, 111, 101, 115, 110, 95, 101, 120, 105, 115, 116]

In [59]: pd.read_msgpack(u'path/doesn_exist')
---------------------------------------------------------------------------
ValueError: path_or_buf needs to be a string file path or file-like

Comment From: jreback

xref https://github.com/pydata/pandas/issues/5924

Comment From: chris-b1

Closed in favor of #15296