Author skrah
Recipients jcon, kristjan.jonsson, mark.dickinson, ncoghlan, paul.moore, petri.lehtinen, pitrou, pv, rupole, skrah, teoliphant, vstinner
Date 2012-01-18.13:34:53
SpamBayes Score 5.50573e-05
Marked as misclassified No
Message-id <1326893694.59.0.983687342873.issue10181@psf.upfronthosting.co.za>
In-reply-to
Content
Revisiting memoryview.size: I foresee problems for NumPy users, since array.size
has a different meaning there:

>>> x = array([[1,2,3], [4,5,6]], dtype='q')
>>> x.shape
(2, 3)
>>> x.itemsize
8
>>> len(x)
2
>>> x.size
6
>>> x.nbytes
48

So here we have:

x.nbytes == product(shape) * itemsize == Py_buffer.len  == (virtual!) byte length
x.size   == product(shape) == number of elements


My suggestion is to use memoryview.nbytes as well. memoryview.size would have
the additional problem that Py_buffer.len is always the byte size of the logical 
structure (e.g. after slicing) and not necessarily the byte size of the physical
memory area.
History
Date User Action Args
2012-01-18 13:34:54skrahsetrecipients: + skrah, teoliphant, paul.moore, mark.dickinson, ncoghlan, rupole, pitrou, kristjan.jonsson, vstinner, pv, jcon, petri.lehtinen
2012-01-18 13:34:54skrahsetmessageid: <1326893694.59.0.983687342873.issue10181@psf.upfronthosting.co.za>
2012-01-18 13:34:54skrahlinkissue10181 messages
2012-01-18 13:34:53skrahcreate