[Numpy-tickets] [NumPy] #315: Inconsistencies with sequence parameters

[NumPy]

#315: Inconsistencies with sequence parameters
------------------------+---------------------------------------------------
 Reporter:  faltet      |       Owner:  somebody   
     Type:  defect      |      Status:  new        
 Priority:  normal      |   Milestone:  1.0 Release
Component:  numpy.core  |     Version:  devel      
 Severity:  normal      |    Keywords:             
------------------------+---------------------------------------------------
 Hi,

 Is the next behavior intended?

 {{{
 In [18]:numpy.array([("aaa","x")], dtype="S3,S1")
 Out[18]:
 array([('aaa', 'x')],    # Good
       dtype=[('f0', '|S3'), ('f1', '|S1')])
 In [19]:numpy.array(["aaa","x"], dtype="S3,S1")
 Out[19]:
 array([('aaa', '\xb7'), ('x[,', '\x08')],  # uh?
       dtype=[('f0', '|S3'), ('f1', '|S1')])
 In [20]:numpy.array(["aaa","x",24], dtype="S3,S1,i4")
 ---------------------------------------------------------------------------
 exceptions.TypeError                                 Traceback (most
 recent call last)

 /home/faltet/python.nobackup/numpy/<ipython console>

 TypeError: expected a readable buffer object

 }}}

 i.e. numpy doesn't complain when dtype only contains string types (but,
 still, gives bad results).

 Furthermore, the above seems to work with rec.array:

 {{{
 In [22]:numpy.rec.array(["aaa","x"], dtype="S3,S1")
 Out[22]:
 recarray(('aaa', 'x'),
       dtype=[('f0', '|S3'), ('f1', '|S1')])
 In [23]:numpy.rec.array(["aaa","x",24], dtype="S3,S1,i4")
 Out[23]:
 recarray(('aaa', 'x', 24),
       dtype=[('f0', '|S3'), ('f1', '|S1'), ('f2', '<i4')])

 }}}

 Should I understand that rec.array is preferred for getting recarrays from
 lists?

 Thanks

-- 
Ticket URL: <http://projects.scipy.org/scipy/numpy/ticket/315>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.