Mailman 3 November 2007 - NumPy-Discussion

Should array iterate over a set?
by Michael McNeil Forbes 17 Nov '07

17 Nov '07

My expectation was that array would iterate over a set. This is incorrect: >>> array(set([1,2,3])) array(set([1, 2, 3]), dtype=object) Is this the intended behaviour? A trivial work-around that does what I need is >>> array(list(set([1,2,3]))) array([1, 2, 3]) but I was wondering if this was by design or just a forgotten corner. (Maybe a vestige of the tuple special case for record arrays?) Michael. P.S. I just found that this was brought up by Ed Schofield on 2006-05-03, but there were no replies in that thread.

2 1

Bug in arange dtype ">f" was: Using arr.dtype.type to check byteorder-independed dtype fails for bool
by Sebastian Haase 17 Nov '07

17 Nov '07

On Nov 13, 2007 2:18 PM, Stefan van der Walt <stefan(a)sun.ac.za> wrote: > Hi Sebastian > > On Tue, Nov 13, 2007 at 01:11:33PM +0100, Sebastian Haase wrote: > > Hi, > > I need to check the array dtype in a way that it is ignoring > > differences coming only from big-endian vs. little-endian. > > Does > > N.issubdtype(first_dtype, second_dtype) > > work? Hi Stéfan, trying to anwer your question with a quick "arange" test, I ran into more confusion: >> a = N.arange(.5, dtype=">f") >>> `a.dtype` 'dtype('float32')' >>> a = N.arange(.5, dtype="<f") >>> `a.dtype` 'dtype('float32')' Both equal positively with N.float32 now ! >>> N.__version__ '1.0.1' This is (was?) apparently a bug in arange It works as expected with empty(): >>> N.empty(5, dtype=">f").dtype==N.float32 False >>> N.empty(5, dtype="<f").dtype==N.float32 True Cheers, Sebastian

2 2

segfault with dot
by Jesus Torrecilla Pinero 17 Nov '07

17 Nov '07

I am using Python 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) under Windows XP and converting a program from Numeric to Numpy. If I have two arrays, say K and T and do dot(K,T) or K*T everything goes well, but if I have a vector b and try dot(K,b) or K*b I get a segfault. I have tried to run the test in tes_numeric.py and get the same result. I have tried too to convert b in a matrix with two columns, the second one being all zeroes, and numpy makes the multiplication correctly, so I think the problem is just with vectors Does this happens with other versions or do you have any idea to fix this problem? Thanks in advance -- Jesús Torrecilla Pinero Universidad de Extremadura jtorrecilla at utcsl.com jtorreci at unex.es

2 3

bug in numpy.apply_along_axis: numpy.__version__ '1.0.3.1'
by Sean Ross-Ross 17 Nov '07

17 Nov '07

Hi, I think I have found a bug in the function "apply_along_axis". recall that the function definition if apply_along_axis(func1d, axis, arr, *args) This bug occurs if func1d returns a python object without a __len__ attribute and is not a scalar, determined by numpy.core.numeric.isscalar. eg. a self contained example of the bug is: >>> from numpy import array, apply_along_axis >>> >>> # works >>> arr = array( [1,1] ) >>> apply_along_axis( lambda arr:arr[0], 0, arr ) >>> >>> # however this raises a type error >>> class Foo(object): pass >>> >>> arr = array( [ Foo(), Foo() ] ) >>> apply_along_axis( lambda arr:arr[0], 0, arr ) ------------------------------------------------------------------------ --- exceptions.TypeError Traceback (most recent call last) /sw/lib/python2.4/site-packages/numpy/lib/shape_base.py in apply_along_axis(func1d, axis, arr, *args) 52 holdshape = outshape 53 outshape = list(arr.shape) ---> 54 outshape[axis] = len(res) 55 outarr = zeros(outshape,asarray(res).dtype) 56 outarr[tuple(i.tolist())] = res TypeError: len() of unsized object >>> Thanks ~Sean

1 0

beginner question: rank-1 arrays
by Robin 16 Nov '07

16 Nov '07

Hi, I am trying to implement a project in scipy. I think I am getting somewhere finally. However in my code (I am converting from MATLAB) it is important to maintain 2d arrays, and keep the difference between row and column vectors. After working through some initial problems I think I am getting more of a picture of how things work in numpy. However I am finding my code littered with things like: np.array(np.r_[0:nterms],ndmin=2) (for a row vector) or np.array(np.r_[0:nterms],ndmin=2).T (for a column vector) Coming from matlab and being use to 0:10 for row or (0:10)' for column this seems a bit messy. Is there a better way of constructing row/column 2d arrays from a slice type range? Thanks, Robin

6 9

Problems building numpy using setupegg
by Chris 16 Nov '07

16 Nov '07

I am trying to build some installers of numpy using setuptools and bdist_mpkg on Leopard using the following command: python setupegg.py config_fc --fcompiler gnu95 config -L../staticlibs build bdist_mpkg --open However, I get the following error towards the end of the build: Copying numpy.egg-info to build/bdist.macosx-10.3-i386/mpkg/ platlib/Library/Python/2.5/site-packages/numpy-1.0.5.dev4464-py2.5.egg-info running install_scripts error: cannot copy tree 'build/scripts.macosx-10.3-i386-2.5': not a directory This is a confusing error, since there is a "build/scripts.macosx-10.3-i386-2.5" and it is certainly a directory.

1 0

test failure: Ticket 396
by Robin 16 Nov '07

16 Nov '07

Hello, Just thought I'd report this test failure with current svn (rev 4464) on Mac OS X 10.5 (could be a leopard issue?) since I hadn't seen it mentioned before. Oddly it only seems to fail in ipython, importing and running from the normal python interpreter all the tests pass. Actually, it only seems to fail when I start ipython with ipython -p scipy I haven't been able to recreate the problem with any other combination of startup and imports. Here is my ipythonrc-scipy: # load our basic configuration with generic options include ipythonrc # import ... # Load Numpy an SciPy by themselves so that 'help' works on them import_mod numpy scipy # from ... import ... import_some # Now we load all of SciPy # from ... import * import_all numpy import_all scipy ------------------------------------------- Here is the error: ====================================================================== ERROR: Ticket #396 ---------------------------------------------------------------------- Traceback (most recent call last): File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/numpy/core/tests/test_regression.py", line 600, in check_poly1d_nan_roots self.failUnlessRaises(N.linalg.LinAlgError,getattr,p,"r") File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/unittest.py", line 320, in failUnlessRaises callableObj(*args, **kwargs) File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/numpy/lib/polynomial.py", line 623, in __getattr__ return roots(self.coeffs) File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/numpy/lib/polynomial.py", line 124, in roots roots = _eigvals(A) File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/numpy/lib/polynomial.py", line 40, in _eigvals return eigvals(arg) File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/scipy/linalg/decomp.py", line 378, in eigvals return eig(a,b=b,left=0,right=0,overwrite_a=overwrite_a) File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/scipy/linalg/decomp.py", line 128, in eig a1 = asarray_chkfinite(a) File "/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/numpy/lib/function_base.py", line 398, in asarray_chkfinite raise ValueError, "array must not contain infs or NaNs" ValueError: array must not contain infs or NaNs ---------------------------------------------------------------------- Ran 692 tests in 0.456s FAILED (errors=1) Out[2]: <unittest._TextTestResult run=692 errors=1 failures=0> Thanks, Robin

1 0

Re: [Numpy-discussion] NumPy trunk is frozen for upcoming 1.0.4 release
by Jarrod Millman 16 Nov '07

16 Nov '07

On Nov 3, 2007 11:06 AM, Albert Strasheim <fullung(a)gmail.com> wrote: > Things mostly look good as far as the buildslaves are concerned, except on > Windows x64, where one of the regression tests is failing: > > http://buildbot.scipy.org/Windows%20XP%20x86_64%20MSVC/builds/163/step-shel… Hello, Can someone take a look at the problem with NumPy on Windows XP? Thanks, -- Jarrod Millman Computational Infrastructure for Research Labs 10 Giannini Hall, UC Berkeley phone: 510.643.4014 http://cirl.berkeley.edu/

3 9

array allocation using tuples gives views of same array
by George Nurser 15 Nov '07

15 Nov '07

I tried the (as I thought) nice compact form In [60]: a,b = (zeros((2,)),)*2 But... In [61]: b[0] = 2 In [62]: a Out[62]: array([ 2., 0.]) a and b are the _same_ array.... But In [68]: a,b = (zeros((2,)),zeros((2,))) In [69]: b[0] = 2 In [70]: a Out[70]: array([ 0., 0.]) is OK. a & b are independent in this case. I'm puzzled by this behaviour, I suspect because of my ignorance of how tuples work. It looks to me like a,b = (zeros((2,)),)*2 is equivalent to x= zeros((2,)) a,b=(x,)*2 If this is indeed a feature rather than a bug, is there an alternative compact way to allocate many arrays? Regards, George.

5 5

list as array item and casting problem problem
by Gen-Nan Chen 14 Nov '07

14 Nov '07

Hi! All, I have below strange behavior in numpy. Can anyone shed some light on this: I have an array with each item as a list object. You will get this when you use scipy.sparse In [369]: k Out[369]: array([[0], [1], [2], [3], [4], [5], [6], [7], [8], [9]], dtype=object) Assigning value by using slicing is working In [370]: k[0:3] = [[i+3] for i in range(3)] In [371]: k Out[371]: array([[3], [4], [5], [3], [4], [5], [6], [7], [8], [9]], dtype=object) But assigning value by using a list for slicing is NOT In [374]: k[[0, 1,2]] = [[i+3] for i in range(3)] In [375]: k Out[375]: array([3, 4, 5, [3], [4], [5], [6], [7], [8], [9]], dtype=object) You can see those list objects casted to integer. Is this a bug or a feature? if I want to slice it with random index without casting it, what should I do? -- Gen-Nan Chen, Ph. D. email: gnchen(a)gmail.com

1 0