Mailman 3 January 2007 - NumPy-Discussion

asarray() behaviour.
by Tim Leslie Jan. 8, 2007

Jan. 8, 2007

I get different behaviour from asarray depending on whether I pass in a list or an array, Additionally, neither of them give me the expected results, which is a TypeError complaining about being unable to use complex values as floats. Can someone shed some light on this. Cheers, Tim In [27]: N.asarray([1j,2,3], dtype=N.float64) --------------------------------------------------------------------------- exceptions.ValueError Traceback (most recent call last) /home/timl/src/scipy/Lib/linsolve/<ipython console> /usr/lib/python2.4/site-packages/numpy/core/numeric.py in asarray(a, dtype, order) 130 are converted to base class ndarray. 131 """ --> 132 return array(a, dtype, copy=False, order=order) 133 134 def asanyarray(a, dtype=None, order=None): ValueError: setting an array element with a sequence. In [28]: N.asarray(N.array([1j,2,3]), dtype=N.float64) Out[28]: array([ 0., 2., 3.])

2 1

Indexing with weave
by Keith Goodman Jan. 8, 2007

Jan. 8, 2007

My first weave attempt weaves something to be desired. test() works, but test2() doesn't. It complains about the line "y = x(0,i);" In function 'PyObject* compiled_func(PyObject*, PyObject*)': blah.cpp:667: error: no match for call to '(py::object) (int, int&)' I am no match for that error message. What am I doing wrong? from scipy import weave from scipy.weave import converters import numpy.matlib as M def test(i): x = M.matrix([[1.0, 2.0, 3.0]]); code = """ return_val = x(0,i); """ y = weave.inline(code, ['x', 'i'], type_converters=converters.blitz) return y def test2(i): x = M.matrix([[1.0, 2.0, 3.0]]); code = """ double y; y = x(0,i); return_val = y; """ y = weave.inline(code, ['x', 'i'], type_converters=converters.blitz) return y

2 3

segfault in numpy.float64(z) for complex z
by Tim Leslie Jan. 8, 2007

Jan. 8, 2007

Hi All, While debugging a scipy crash I came across the problem outlined in http://projects.scipy.org/scipy/numpy/ticket/412 Could someone shed some light on why this is happening, I got a bit lost in the numpy internals when I tried to track it down. Cheers, Tim

3 4

linalg.qr versus scilab qr
by R. David Jan. 8, 2007

Jan. 8, 2007

Hello, In scilab, the QR decomposition can produce "sorted" Q and R matrices, where diagonal elements of R are decreasing. Thus, a permutation vector E is returned, so that X*E = Q*R. Is there a way to compute such a permutation QR in numpy ? It seems that numpy.linalg.qr does not sort the diagonal R elements. Do I have to wrap around lapack's zgeqpf or zgeqp3 ? Romaric

1 0

Ticket #132
by Tim Leslie Jan. 7, 2007

Jan. 7, 2007

Hi all, I've been looking through some open tickets and found ticket #132 which suggests a number of ways to improve numpy's web presence. http://projects.scipy.org/scipy/numpy/ticket/132 The only outstanding item on this is point #2, "Pointing www.numpy.org to numeric.scipy.org instead of the SF page". This seems like a good idea, but hasn't been done yet. Are there plans to implement this at some point? Cheers, Tim

1 0

Missing Py_DECREF?
by Charles R Harris Jan. 7, 2007

Jan. 7, 2007

In the following lines from the initmultiarray function in multiarraymodule.c there seems to be a missing Py_DECREF MultiArrayError = PyString_FromString ("multiarray.error"); PyDict_SetItemString (d, "error", MultiArrayError); Should, I think, be followed by Py_DECREF(MultiArrayError); Chuck

1 0

Installing numpy from rpm
by Satyendra Tomar Jan. 7, 2007

Jan. 7, 2007

Hi, I'm entirely new to Python. I just used "rpm -Uvh numpy-1.0.1-1.src.rpm" to install numpy from the rpm package, running under SuSE 10.1, 2.6.16.21, python 2.4.2, GCC 4.1.0 but python can't still find the numpy module. What I only find is /usr/src/packages/SPECS/numpy.spec /usr/src/packages/SOURCES/numpy-1.0.1.tar.gz What else I'm supposed to do? Thanks in advance. S. K. Tomar

2 2

Advanced selection, duplicate indices, and augmented assignment
by Sean R. Lynch Jan. 6, 2007

Jan. 6, 2007

I have an array of vertices and an array of indices that make a list of triangles. I'd like to calculate the normals for each vertex as a (pseudo-)average of the normals of each triangle that contains it, so I get smooth lighting in OpenGL. I'd also like to avoid looping in Python because the vertex positions need to be able to be updated interactively. I tried using the following code: def calculate_normals(self): i = self.indices faces = self.vertices[i].reshape((i.shape[0] / 3, 3, 3)) face_normals = cross(faces[:,1] - faces[:,0], faces[:,2] - faces[:,0]) normals = zeros(self.vertices.shape, float32) normals[i] += face_normals.repeat(3, axis=0) normals /= sqrt(sum(normals*normals, axis=1)).reshape(normals.shape[0], 1) self.normals = normals but this appears to only add the face normal of the last face that contains each vertex, so I don't get averaged normals but the normals of a single face each. A test in the interactive interpreter shows why: >>> x = zeros((3,)) >>> x[array([0, 1, 1])] += array([1, 1, 1]) >>> x array([ 1., 1., 0.]) If this worked the way I was hoping, the output would be [1 2 0] because it would add to element 1 twice due to its duplication in the advanced selection array. Is the current behavior intentional or is it an accident of implementation? I think the multiple operations on elements with duplicated indices is more intuitive and it makes it possible to "invert" the expansion due to duplicate indices, whereas dropping the duplicate indices throws away information. Are people relying on the current behavior with duplicate indices and augmented assignment? I'd be happy with another way to do this that doesn't require a loop in Python or lots of temporary arrays, as well. Does anybody know of one? -Sean P.S. Could somebody please update http://numpy.scipy.org/ to point to the new list? I wasted a lot of time subscribing and posting to the wrong list.

2 1

custom accumlators
by Matt Knox Jan. 6, 2007

Jan. 6, 2007

I made a post about this a while ago on the scipy-user mailing list, but I didn't receive much of a response so I'm just throwing it out there again (with more detail) in case it got overlooked. Basically, I'd like to be able to do accumulate operations with custom functions. numpy.vectorize does not seem to provide an accumulate method with the functions it returns. I'm hoping I don't have to write ufuncs in C to accomplish this, but I fear that may the case. Either way, it would be nice to know if it can, or cannot be done in an easy manner. I have lots of examples of where this kind of thing is useful, but I'll just outline two for now. Assume the parameter x in all the functions below is a 1-d array --------------------------------------------- Example 1 - exponential moving average: # naive brute force method... def expmave(x, k): result = numpy.array(x, copy=True) for i in range(1, result.size): result[i] = result[i-1] + k * (result[i] - result[i-1]) return result # slicker method (if it worked, which it doesn't)... def expmave(x, k): def expmave_sub(a, b): return a + k * (b - a) return numpy.vectorize(expmave_sub).accumulate(x) --------------------------------------------- Example 2 - forward fill a masked array: # naive brute force method... def forward_fill(x): result = ma.array(x, copy=True) for i in range(1, result.size): if result[i] is ma.masked: result[i] = result[i-1] return result # slicker method (if it worked, which it doesn't)... def forward_fill(x): def forward_fill_sub(a, b): if b is ma.masked: return a else: return b return numpy.vectorize(forward_fill_sub).accumulate(x) --------------------------------------------- Is their a good way to do these kinds of things without python looping? Or is that going to require writing a ufunc in C? Any help is greatly appreciated. Thanks, - Matt Knox

9 16

recompiling needed for binary module after numpy 1.0
by Sebastian Haase Jan. 5, 2007

Jan. 5, 2007

Hi! After I upgraded from numpy 1.0 to 1.0.1 I get an abort in a C-module: RuntimeError: module compiled against version 1000002 of C-API but this version of numpy is 1000009 Fatal Python error: numpy.core.multiarray failed to import... exiting. /opt/bin/priithonN: line 37: 1409 Aborted $PY $* I thought that within numpy 1.0 there was no recompile for external C-modules needed !? Please explain. Thanks, Sebastian Haase

3 2