[NumPy-Tickets] [NumPy] #1539: MSVC specific TypeError when using double, longdouble in numpy.dot

[NumPy Trac]

#1539: MSVC specific TypeError when using double, longdouble in numpy.dot
---------------------+------------------------------------------------------
 Reporter:  cgohlke  |       Owner:  somebody
     Type:  defect   |      Status:  new     
 Priority:  normal   |   Milestone:  2.0.0   
Component:  Other    |     Version:          
 Keywords:           |  
---------------------+------------------------------------------------------
 This issue has been discussed at [http://mail.scipy.org/pipermail/numpy-
 discussion/2010-July/051352.html] and is the cause of the scipy.sparse
 TypeError test failures reported at [http://mail.scipy.org/pipermail
 /scipy-user/2010-July/025961.html].

 Using Python 2.6.5 32 bit on Windows and numpy 1.4.1 built with Visual
 Studio 2008 (msvc9), I get the following:

 {{{
 >>> a = numpy.array([[1.0]])
 >>> numpy.dot(a.astype('double'), a.astype('longdouble'))
 array([[ 1.]])

 >>> a = numpy.array([[1]])
 >>> numpy.dot(a.astype('double'), a.astype('longdouble'))
 Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
 TypeError: array cannot be safely cast to required type

 >>> a = numpy.array([[1]])
 >>> numpy.dot(a.astype('single'), a.astype('longdouble'))
 array([[1.0]], dtype=float64)

 >>> a = numpy.array([[1]])
 >>> numpy.dot(a.astype('longdouble'), a.astype('double'))
 array([[1.0]], dtype=float64)

 >>> numpy.array([1]).astype('longdouble').dtype.num
 13
 >>> numpy.array([1.0]).astype('longdouble').dtype.num
 12
 }}}


 The TypeError is unexpected. The same results are obtained when using a
 msvc7 build of numpy 1.4.1 on Python 2.5, or a msvc9 64 bit build of numpy
 2.0.dev on Python 3.1.

 Using numpy 1.4.1 built with mingw does not raise a TypeError:

 {{{
 >>> a = numpy.array([[1]])
 >>> numpy.dot(a.astype('double'), a.astype('longdouble'))
 array([[1.0]], dtype=float96)
 }}}

 Note that for MSVC compilers sizeof(longdouble) = sizeof(double) = 8

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1539>
NumPy <http://projects.scipy.org/numpy>
My example project