Mailman 3 December 2009 - NumPy-Discussion

objects are not aligned. Matrix and Array
by Wayne Watson 19 Dec '09

19 Dec '09

This program gives me the message following it: ================Program========== import numpy as np from numpy import matrix import math def sinD(D): # given in degrees, convert to radians return math.sin(math.radians(D)) def cosD(D): return math.cos(math.radians(D)) r = math.sqrt(2*2+5*5) print r m1 = matrix([[2], [5]]) print "m1: ", m1 theta = 5.0 # degrees #CW 2x2 clockwise matrix rotCW = matrix([ [cosD(theta), sinD(theta)], [-sinD(theta), cosD(theta)] ]) print rotCW m2 = rotCW*m1 print "m2: ", m2 print "aaaaaaaa: ", type(m1), type(m2) m1=np.array(m1) m2=np.array(m2) print "zzzzzzzz: ", type(m1), type(m2) print"dot, ..." dotres = np.dot(m1,m2) print "dotres", dotres ==============end========== ==========Output msgs======== 5.38516480713 m1: [[2] [5]] [[ 0.9961947 0.08715574] [-0.08715574 0.9961947 ]] m2: [[ 2.42816811] [ 4.806662 ]] aaaaaaaa: <class 'numpy.core.defmatrix.matrix'> <class 'numpy.core.defmatrix.matrix'> zzzzzzzz: <type 'numpy.ndarray'> <type 'numpy.ndarray'> dot, ... Traceback (most recent call last): File "C:/Sponsor_Meteors/Sentinel_Development/Development_Sentuser+Utilities/Playground/junk.py", line 30, in <module> dotres = np.dot(m1,m2) ValueError: objects are not aligned ================end msgs=========== Why the msg? The types look alike and each array/matrix contains two elements.. -- Wayne Watson (Watson Adventures, Prop., Nevada City, CA) (121.015 Deg. W, 39.262 Deg. N) GMT-8 hr std. time) Obz Site: 39° 15' 7" N, 121° 2' 32" W, 2700 feet "... humans'innate skills with numbers isn't much better than that of rats and dolphins." -- Stanislas Dehaene, neurosurgeon Web Page: <www.speckledwithstars.net/>

3 4

Re: [Numpy-discussion] [AstroPy] Rotating and Transforming Vectors--Flight Path of a Celestial Body
by Wayne Watson 19 Dec '09

19 Dec '09

It's starting to come back to me. I found a few old graphics books that get into transformation matrices and such. Yes, earth centered. I ground out some code with geometry and trig that at least gets the first point on the path right. I think I can probably apply a rotation around the x-axis several times with a 1/2 degree rotation for each step towards the north to get the job done. I'm still fairly fresh with Python, and found a little bit of info in a Tentative numpy tutorial. I found this on getting started with matrices: from numpy import matrix Apparently matrix is a class in numpy, and there are many others, linalg I think is one. How does one find all the classes, and are there rules that keep them apart. It was tempting to add import numpy in addition to the above, but how is what is in numpy different that the classes? Is numpy solely class driven? That is, if one does a from as above to get all the classes, does it give all the capabilities that just using import numpy does? Anne Archibald wrote: > 2009/12/17 Wayne Watson <sierra_mtnview(a)sbcglobal.net>: > >> I'm just getting used to the math and numpy library, and have begun >> working on a problem of the following sort. >> >> Two observing stations are equidistant, 1/2 degree, on either side of a >> line of longitude of 90 deg west, and both are at 45 deg latitude. Given >> the earth's circumference as 25020 miles, a meteor is 60 miles above the >> point between the two sites. That is, if you were standing at long >> 90deg and lat 45 deg, the meteor would be that high above you. 70 miles >> along the long line is 1 degree, so the stations are 70 miles apart. I >> want to know the az and el of the meteor from each station. With some >> geometry and trig, I've managed to get that first point; however, I can >> see moving the meteor say, 1/2 deg, along its circular path towards the >> north pole is going to require more pen and pencil work to get the az/el >> for it. >> >> Long ago in a faraway time, I used to do this stuff. It should be easy >> to rotate the vector to the first point 1/2 deg northward, and find the >> vector there, then compute the new az and el from each station. Maybe. >> I'm just beginning to look at the matrix and vector facilities in numpy. >> Maybe someone can comment on how this should be done, and steer me >> towards what I need to know in numpy. >> > > You may find that the problem grows drastically easier if you work as > much as possible in so-called earth-centered earth-fixed coordinates > (sometimes called XYZ) coordinates. These are a rectilinear coordinate > system that rotates with the earth, with the Z axis through the north > pole and the X axis through the equator at the Greenwich meridian. > It's kind of horrible for getting altitudes, since the Earth is sort > of pear-shaped, but it makes the 3D geometry much simpler. > > If you don't go this route, I'd recommend picking one station and > defining a rectilinear coordinate system based on its north and > vertical vectors. The north and vertical vectors of the other station > will be at somewhat funny angles (unless you can get away with > treating the Earth as flat between the two), but whatever rectilinear > coordinates you choose, a dot product lets you calculate vector > lengths and angles between them, and a cross product lets you build > vectors orthogonal to a given pair. So, for example, if your station > has north vector N and up vector U, you can get its east vector as > E=cross(N,U) (then normalize it); if you want to convert an absolute > north to a local north (i.e. one that is horizontal) you can do > N=cross(E,U) (and normalize it). Then you can get the azimuth of a > vector V using dot(N,V)/sqrt(dot(V,V)) and dot(E,V)/sqrt(dot(V,V)). > > > Anne > > -- Wayne Watson (Watson Adventures, Prop., Nevada City, CA) (121.015 Deg. W, 39.262 Deg. N) GMT-8 hr std. time) Obz Site: 39° 15' 7" N, 121° 2' 32" W, 2700 feet "... humans'innate skills with numbers isn't much better than that of rats and dolphins." -- Stanislas Dehaene, neurosurgeon Web Page: <www.speckledwithstars.net/>

4 3

ValueError for numpy when importing KDTree from scipy.spatial
by Howard Chong 19 Dec '09

19 Dec '09

I'm getting an odd behavior when I try to load KDTree. In the interactive interpreter: the first time I try to load it, it gives me an error. Second time works fine. When trying to run from command line, same error. To reproduce, just type the 4 commands below. If people can't reproduce it, I might do a re-install. Here's my session. *** Python 2.6.4 (r264:75708, Oct 26 2009, 08:23:19) [MSC v.1500 32 bit (Intel)] on win32. *** >>> from scipy.spatial import KDTree Traceback (most recent call last): File "<interactive input>", line 1, in <module> File "C:\Python26\lib\site-packages\scipy\spatial\__init__.py", line 7, in <module> from ckdtree import * File "numpy.pxd", line 30, in scipy.spatial.ckdtree (scipy\spatial\ckdtree.c:6087) ValueError: numpy.dtype does not appear to be the correct type object >>> KDTree Traceback (most recent call last): File "<interactive input>", line 1, in <module> NameError: name 'KDTree' is not defined >>> from scipy.spatial import KDTree >>> KDTree <class 'scipy.spatial.kdtree.KDTree'> >>> -- Howard Chong Dept. of Agricultural and Resource Economics and Energy Institute @ Haas Business School UC Berkeley

3 2

fromfile can segfault if data is corrupted
by Michael Droettboom 18 Dec '09

18 Dec '09

I just discovered a bug in fromfile where it can segfault if the file data is corrupted in such a way that the array size is insanely large. (It was a byte-swapping problem in my own code, but it would be preferable to get an exception rather than a crash). It's a simple fix to propagate the "array too large" exception before trying to dereference the NULL array pointer (ret) in PyArray_FromFile (see attached patch). But my question is: is this an appropriate fix for 1.4 (it seems pretty straightforward), or should I only make this to the trunk? Mike -- Michael Droettboom Science Software Branch Operations and Engineering Division Space Telescope Science Institute Operated by AURA for NASA Index: ctors.c =================================================================== --- ctors.c (revision 7992) +++ ctors.c (working copy) @@ -3042,6 +3055,10 @@ (next_element) fromfile_next_element, (skip_separator) fromfile_skip_separator, NULL); } + if (ret == NULL) { + Py_DECREF(dtype); + return NULL; + } if (((intp) nread) < num) { /* Realloc memory for smaller number of elements */ const size_t nsize = NPY_MAX(nread,1)*ret->descr->elsize;

3 3

Segmentation fault with argsort
by Keith Goodman 18 Dec '09

18 Dec '09

I am using the numpy 1.3 binary from Ubuntu 9.10. Is this already known, fixed, reproducible? >> np.array(121).argsort(0).argsort(0) Segmentation fault The expected result: AttributeError: 'np.int64' object has no attribute 'argsort'

6 13

test_multiarray.TestIO.test_ascii segmentation fault with Python2.7
by Bruce Southey 18 Dec '09

18 Dec '09

Hi, After installing Python2.7, a patched nose (http://bitbucket.org/kumar303/nose-2_7_fixes/ because unittest._TextTestResult has been removed) and numpy '1.5.0.dev8011', numpy.test crashes with a segmentation fault with the test for: test_multiarray.TestIO.test_ascii If I understand the test correctly: $ python Python 2.7a1 (r27a1:76674, Dec 14 2009, 13:46:01) [GCC 4.4.1 20090725 (Red Hat 4.4.1-2)] on linux2 Type "help", "copyright", "credits" or "license" for more information. >>> import numpy as np >>> from numpy.compat import asbytes, getexception >>> np.fromstring(asbytes('1 , 2 , 3 , 4'),sep=',') Segmentation fault This code works under Python2.6 and numpy '1.5.0.dev8011'. Bruce

3 7

np.void from 0d array + subclassing
by Pierre GM 18 Dec '09

18 Dec '09

All, * What is the most efficient way to get a np.void object from a 0d structured ndarray ? * Is there any way to subclass np.void ? Thanks a lot in advance ! P.

4 8

segfault in vdot
by Neal Becker 17 Dec '09

17 Dec '09

http://projects.scipy.org/numpy/ticket/1335

1 0

tracking additions to NumPy across versions
by David Warde-Farley 16 Dec '09

16 Dec '09

Hi all, Is there currently anything in the docstring standard about tracking when functions get added to NumPy? The recent discussion of matrix_rank got me thinking about this. Regards, David

1 1

Slicing slower than matrix multiplication?
by Jasper van de Gronde 16 Dec '09

16 Dec '09

(Resending without attachment as I don't think my previous message arrived.) I just started using numpy and am very, very pleased with the functionality and cleanness so far. However, I tried what I though would be a simple optimization and found that the opposite was true. Specifically, I had a loop where something like this was done: w += Xi[mu,:] E = np.dot(Xi,w) Instead of repeatedly doing the matrix product I thought I'd do the matrix product just once, before the loop, compute the product np.dot(Xi,Xi.T) and then do: w += Xi[mu,:] E += Xi2[mu,:] Seems like a clear winner, instead of doing a matrix multiplication it simply has to sum two vectors (in-place). However, it turned out to be 1.5 times SLOWER... I've attached a test file which shows the problem. It also tries adding columns instead of rows (in case the memory layout is playing tricks), but this seems to make no difference. This is the output I got: Dot product: 5.188786 Add a row: 8.032767 Add a column: 8.070953 Any ideas on why adding a row (or column) of a matrix is slower than computing a matrix product with a similarly sized matrix... (Xi has less columns than Xi2, but just as many rows.)

10 22