Mailman 3 April 2011 - NumPy-Discussion

Remove deprecated skiprows
by Charles R Harris 06 Apr '11

06 Apr '11

Should skiprows be removed? if skiprows: warnings.warn(\ "The use of `skiprows` is deprecated, it will be removed in numpy 2.0.\n" \ "Please use `skip_header` instead.", DeprecationWarning) skip_header = skiprows Its been deprecated since 1.4. Personally, I like skiprows better than skip_header ;) Chuck

7 10

atpy and readable buffer error
by Tommy Grav 05 Apr '11

05 Apr '11

I get this error that I do not understand. Does anyone know what is happening and how to get around it >>> import atpy >>> t = atpy.Table() >>> name = ["aa","bb","cc"] >>> t.add_column("name",name) >>> k = [1,2,3] >>> t.add_column("num",k) >>> t <Table name='None' rows=3 fields=2> >>> name = ["aaa","bbb","ccc"] >>> p = atpy.Table() >>> p.add_column("name",name) >>> p.add_column("num",k) >>> t.append(p) Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/atpy/basetable.py", line 301, in append raise Exception("Columns do not match") Exception: Columns do not match >>> nn = t.name.astype("|S3") >>> t.remove_columns("name") >>> t.add_column("name",nn) >>> t.append(p) Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/atpy/basetable.py", line 302, in append self.data = np.hstack((self.data, table.data)) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/numpy/core/shape_base.py", line 258, in hstack return _nx.concatenate(map(atleast_1d,tup),1) TypeError: expected a readable buffer object >>> Cheers Tommy

1 0

convert array to tuples?
by Christopher Barker 05 Apr '11

05 Apr '11

HI folks, What's the fastest way to convert a numpy array to tuples? Unfortunately, not all packages take arbitrary sequences when they are expecting a list of tuples. In this case, I'm using PIL's ImageDraw, and want to do: draw.line(points) (points is an Nx2 numpy array of ints) but if I pass in an array, I get, oddly enough, nothing. NO error, just nothing drawn. if I do: draw.line(points.tolist()) SystemError: new style getargs format but argument is not a tuple So I seem to need tuples. This works: draw.line([tuple(p) for p in points.tolist()]) but that seems like it would be slower than it should be (to be honest, not profiled). Is there a faster way? Maybe numpy should have a ndarray.totuple() method. -Chris -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/OR&R (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception Chris.Barker(a)noaa.gov

4 4

Problem with np.power in numpy 1.5.1-2
by M. 05 Apr '11

05 Apr '11

Hi all, I've been using numpy without problems - version .1.4.1-5 - until this morning, when I upgraded my Debian sid system to numpy 1.5.1-2. A function that was working that uses np.power def get_far_field(self, chi, n): return np.sum(self.an[n,1:self.j] * np.power(chi,-np.arange(1,self.j))) outputs "Warning: overflow encountered in power". I'm using python 2.6.6-12. Is this a bug in the numpy library or (more probably) my code? If it's in my code, what can be happening? How to debug it? Thanks

2 1

Re: [Numpy-discussion] Strange behavior of operator *=
by Alan G Isaac 05 Apr '11

05 Apr '11

On 4/5/2011 9:26 AM, François Steinmetz wrote: > It does not change the dtype, 'int' is just interpreted as 'int64' : So the meaning of 'int' is system specific? >>> import numpy as np; a=np.eye(2,dtype='int'); a.dtype dtype('int32') Alan Isaac

4 4

Strange behavior of operator *=
by François Steinmetz 05 Apr '11

05 Apr '11

Hi all, I have encountered the following strangeness : >>> from numpy import * >>> __version__ '1.5.1' >>> a = eye(2, dtype='int') >>> a *= 1.0 >>> a ; a.dtype array([[1, 0], [0, 1]]) dtype('int64') >>> a = a * 1.0 >>> a ; a.dtype array([[ 1., 0.], [ 0., 1.]]) dtype('float64') So, in this case "a *= x" is not equivalent to "a = a*x" ? François

4 4

How to create structured char arrays?
by Bruce Southey 05 Apr '11

05 Apr '11

Hi, I think that I have resolved my issue down to creating a structured string array. I am using numpy version '2.0.0.dev-3c338cb'. Without a structured array, it should be a 2 by 2 array: >>> np.array([('a','b'),('c','d')]) array([['a', 'b'], ['c', 'd']], dtype='|S1') >>> np.array([('a','b'),('c','d')], dtype='a2') array([['a', 'b'], ['c', 'd']], dtype='|S2') It does not work as a structured array: >>> np.array([('a','b'),('c','d')], dtype=[('label','a2'), ('Value', 'a4')]) array([('a', 'b'), ('c', 'd')], dtype=[('label', '|S2'), ('Value', '|S4')]) What is the correct dtype argument to use in this case? Thanks Bruce

2 2

bug with numpy nanargmax/nanargmin
by Dmitrey 05 Apr '11

05 Apr '11

>>> from numpy import * >>> nanargmax([nan,nan]) nan # ok >>> nanargmax([nan,nan],0) nan # ok >>> nanargmax([[1,nan],[1,nan]],0) Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/local/lib/python2.6/site-packages/numpy/lib/function_base.py", line 1606, in nanargmax return _nanop(np.argmax, -np.inf, a, axis) File "/usr/local/lib/python2.6/site-packages/numpy/lib/function_base.py", line 1346, in _nanop res[mask_all_along_axis] = np.nan ValueError: cannot convert float NaN to integer >>> __version__ '2.0.0.dev-1fe8136' D.

1 0

python3 setup.py install fails with git maint/1.6.x
by Darren Dale 04 Apr '11

04 Apr '11

I just checkout out the 1.6 branch, attempted to install with python3: RefactoringTool: Line 695: You should use a for loop here Running from numpy source directory.Traceback (most recent call last): File "setup.py", line 196, in <module> setup_package() File "setup.py", line 170, in setup_package write_version_py() File "setup.py", line 117, in write_version_py from numpy.version import git_revision as GIT_REVISION ImportError: cannot import name git_revision Next, I built and installed with python2, with no problems. Then I attempted to install with python3 again, at which point git_revision was importable, presumably because it was provided during the python2 build. Darren

4 4

"Best" dot product along an axis?
by Skipper Seabold 04 Apr '11

04 Apr '11

Going through some of the recent threads on similar problems, I'm trying to discern which is "best." I have X is T x i, theta is T x i x j. I want a T by j array that contains X[t].T.dot(theta[t]) along axis 0. Say, T,i,j = 166,7,3 X = np.random.random((T,i)) theta = np.random.random((T,i,j)) # Using sum F1 = (X[:,:,None] * theta).sum(1) # Using a loop F2 = np.zeros((T,j)) for t in range(T): for j_idx in range(j): for i_idx in range(7): F2[t,j_idx] += X[t,i_idx] * theta[t,i_idx,j_idx] # One way with dot, keeps an extra T index F3 = np.squeeze(np.dot(X[:,None,:],theta)).reshape(-1,3)[::167] # the above is way fast than F3_2 = np.dot(X,theta).reshape(-1,3)[::167] # zipped arrays F4 = np.asarray([np.dot(x,theta_i) for x,theta_i in zip(X,theta)]) It seems that F3 is the fastest way to do this given T,i,j, but it seems inefficient to compute things I don't want. Does it make sense that F3 is faster than F3_2? I'm using generic Ubuntu ATLAS on this machine. Is there a more efficient way to do this? I'm not coming up with one. Skipper

1 0