To be honest, this doesn't seem justifiable.

Where it got me is interfacing with c-code that expected a 1d array, and I was calling it with arrays of varying length. I was using this to ensure the proper typing; however, when the array was length 1, the program crashed...

Should I file a bug report?

--Hoyt

On Mon, Apr 28, 2008 at 11:51 PM, Charles R Harris charlesr.harris@gmail.com wrote:

On Tue, Apr 29, 2008 at 12:28 AM, Hoyt Koepke hoytak@gmail.com wrote:

Hello,

I have a quick question that I'm hoping will improve my numpy understanding. I noticed some behavior when using float64 to convert a matrix type that I didn't expect:

In [35]: b1 = array([1.0])

In [36]: float64(b1) Out[36]: 1.0

In [37]: b2 = array([1.0, 2.0])

In [38]: float64(b2) Out[38]: array([ 1., 2.])

I didn't expect calling float64 would convert b1 to a scalar. Seems like an inconsistency. I assume this is intentional, as someone would have noticed it a long time ago if not, so could someone explain the reasoning behind it? (or point me to a source that will help?)

It's inconsistent and looks like a bug:

In [4]: float32(array([[[1]]])) Out[4]: array([[[ 1.]]], dtype=float32)

In [5]: float64(array([[[1]]])) Out[5]: 1.0

Float64 is a bit special because it starts as the python float. Maybe Travis can say what the differences are.

Chuck

Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion