Mailman 3 April 2008 - NumPy-Discussion

f2py: optional parameters and present() bug?
by Garry Willgoose April 23, 2008

April 23, 2008

in a F90 routine called from python the present() test (for optional parameters) always returns true whether the parameter is present in the function call or not. My F90 test code looks like this subroutine test(test1,test2,test3) integer,intent(in) :: test1 integer,intent(in),optional :: test2 integer,intent(inout),optional :: test3 write (*,*) test1,present(test2),present(test3) if (present(test2)) write (*,*) 'test2',test2 if (present(test3)) then write (*,*) 'test3',test3 test3=5 end if end subroutine test The output from python calls to this routine (its inside a module tsdtm.tsg ... just to explain why the function looks like it does). I have just created this with f2py automatically generating the interface with no hand modification of the interfaces. What this test shows is that present() always returns true no matter (1) the intent (2) whether the parameter is present in the call, or (3) the type of the parameter (real, integer ... though I haven't tested arrays etc) as well. Secondly I can't see how to get the returned value of 'test3'. It doesn't seem to be returned as a result of the function? The docs say f2py handles optional parameters and other than present () my testing suggests that optional parameters seem otherwise OK. I'm on Mac OSX and using numpy 1.0.3. Any thoughts? >>> print tsimdtm.tsg.test.__doc__ test - Function signature: test(test1,[test2,test3]) Required arguments: test1 : input int Optional arguments: test2 : input int test3 : in/output rank-0 array(int,'i') >>> tsimdtm.tsg.test(1) 1 T T test2 0 test3 0 >>> tsimdtm.tsg.test(1,2) 1 T T test2 2 test3 0 >>> tsimdtm.tsg.test(1,2,3) 1 T T test2 2 test3 3 >>> tsimdtm.tsg.test(1,test2=2) 1 T T test2 2 test3 0 >>> tsimdtm.tsg.test(1,test3=3) 1 T T test2 0 test3 3 >>> tsimdtm.tsg.test(1,test2=2,test3=3) 1 T T test2 2 test3 3 ==================================================================== Prof Garry Willgoose, Australian Professorial Fellow in Environmental Engineering, Director, Centre for Climate Impact Management (C2IM), School of Engineering, The University of Newcastle, Callaghan, 2308 Australia. Centre webpage: www.c2im.org.au Phone: (International) +61 2 4921 6050 (Tues-Fri AM); +61 2 6545 9574 (Fri PM-Mon) FAX: (International) +61 2 4921 6991 (Uni); +61 2 6545 9574 (personal and Telluric) Env. Engg. Secretary: (International) +61 2 4921 6042 email: garry.willgoose(a)newcastle.edu.au; g.willgoose(a)telluricresearch.com email-for-life: garry.willgoose(a)alum.mit.edu personal webpage: www.telluricresearch.com/garry ==================================================================== "Do not go where the path may lead, go instead where there is no path and leave a trail" Ralph Waldo Emerson ====================================================================

2 1

different behaviour in asfarray(None)
by lorenzo bolla April 23, 2008

April 23, 2008

I noticed a change in the behaviour of numpy.asfarray between numpy version 1.0.5 and 1.1.0: 1.0.5 ==== In [3]: numpy.asfarray(None) Out[3]: array(nan) In [4]: numpy.__version__ Out[4]: '1.0.5.dev4455' 1.1.0 ==== In [16]: numpy.asfarray(None) --------------------------------------------------------------------------- <type 'exceptions.TypeError'> Traceback (most recent call last) c:\Documents and Settings\bollalo001\My Documents\archive\python\openopt\<ipython console> in <module>() c:\Python25\Lib\site-packages\numpy\lib\type_check.py in asfarray(a, dtype) 47 if not issubclass(dtype, _nx.inexact): 48 dtype = _nx.float_ ---> 49 return asarray(a,dtype=dtype) 50 51 def real(val): c:\Python25\Lib\site-packages\numpy\core\numeric.py in asarray(a, dtype, order) 133 are converted to base class ndarray. 134 """ --> 135 return array(a, dtype, copy=False, order=order) 136 137 def asanyarray(a, dtype=None, order=None): <type 'exceptions.TypeError'>: float() argument must be a string or a number In [17]: numpy.__version__ Out[17]: '1.1.0.dev5061' Is this intended? why? L.

4 5

Fromfunction generalization
by Gnata Xavier April 22, 2008

April 22, 2008

Hi, fromfunction is fine but I have like to be able to create 2Darrays using a function of i,j but also of one (or more) parameters. Something like that : def f(i,j,a): return (i+j)*a #replace that by another non trivial computation M=fromfunction(f( hum well something like i,j,a),(1000,1000)) this only way I know is to use global : a=5 def f(i,j): global a return (i+j)*a M=fromfunction(f,(1000,1000)) a=7 P=fromfunction(f,(1000,1000)) but it is quite ugly :( Is there a clean and *fast* way to do that ? Xavier

3 3

Adding nested loops to template processor.
by Charles R Harris April 22, 2008

April 22, 2008

Hi All, I want to use nested loops in our c template processor. The least intrusive way I see to do this is to change the repeat header slightly, i.e. /**begin repeat #var1= a, b, c# #var2= d, e, f# #nest=1# #var3= w, x# #var4= y, z# */ Where "nest" becomes a keyword. The variables then take values in the order var1=a, var2=d, var3=w, var4=y var1=a, var2=d, var3=x, var4=z .... Does anyone object to this change or have a better suggestion? Chuck

1 0

Numpy 1.1.0 and matplotlib 0.90.1
by lorenzo bolla April 22, 2008

April 22, 2008

Hello, the latest svn numpy version 1.1.0.dev5061 does not work with matplotlib 0.90.1 (version shipped with enthought distribution), unless a change in Python25/Lib/site-packages/matplotlib-0.90.1.0003-py2.5-win32.egg/matplotlib/numerix/ma/__init__.py is done: $ diff __init__.py.orig __init__.py 12c12 < from numpy.core.ma import * --- > from numpy.ma import * Maybe this should be forwarded to the pylab mailing list, but I'm not subscribed there... L.

4 3

Code generator bug and fix?
by Charles R Harris April 22, 2008

April 22, 2008

I've gotten my own python class with a logical_not method to work correctly if I goto numpy/core/code_generators/generate_umath.py and change 'logical_not' : Ufunc(1, 1, None, 'returns not x elementwise.', TD(noobj, out='?'), TD(M, f='logical_not', out='?'), ), to 'logical_not' : Ufunc(1, 1, None, 'returns not x elementwise.', TD(noobj, out='?'), TD(M, f='logical_not'), ), There are a whole bunch of functions with this same error and before I start changing them, I would like someone more familiar with this code to tell me if I'm doing the right thing. Chuck

2 8

[numscons] 0.6.3 release: building scipy with MS compilers works
by David Cournapeau April 22, 2008

April 22, 2008

Hi, Sorry for announcing one more numscons release in such a short time: this releases can finally handle the last common platform, Visual Studio on win32. Win32 installers and source tarballs can be found on launchpad, as usual: https://code.launchpad.net/numpy.scons.support/0.6/0.6.3 Python eggs on pypi should follow soon (I cannot upload to pypi from my lab, unfortunately). Except the various fixes necessary to make the build work with MS compilers, there were a few improvements: - some more improvements for f2py scons tool, which simplified a bit some scipy scons scripts - the addition of a silent mode, to get more terse command lines output (not perfect yet), ala kbuild for people familiar with compiling linux (e.g. you get CC foo.c instead of gcc -W -Wall blas blas blas). The goal is to make warning more obvious (and to fix them at some point, of course :) ). Just use python setupscons.py scons --silent=N with N between 1 and 3 to see the result. As this was the last major platform I wanted to support, I can now move on polishing the API as well as starting a real documentation for package developers. cheers, David

1 0

Summing indices of heterogeneous shapes
by Michael Hoffman April 22, 2008

April 22, 2008

In the following example I can sum up lists of column indices: >>> x = numpy.arange(30) >>> x.shape = (6, 5) >>> x array([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 8, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19], [20, 21, 22, 23, 24], [25, 26, 27, 28, 29]]) >>> cols = [[0, 1], [2, 3]] >>> x[:, cols].sum(2) array([[ 1, 5], [11, 15], [21, 25], [31, 35], [41, 45], [51, 55]]) Is there a way to do this if the shapes of the lists are heterogeneous? For example, what if I wanted to set cols = [[0, 1], [2, 3], [1, 2, 3]]? What's the best way to get a sensible result? Thanks! -- Michael Hoffman

1 1

Boolean output for Object arrays with M
by Travis E. Oliphant April 22, 2008

April 22, 2008

Hi Chuck, O.K. I've finally caught up with you, I think (at least as far as the umath generators go). All TD(M,.....) lines should *not* have out='?' The reasoning is that this specifies a ufunc type with object inputs and boolean array outputs where the function is a method of the object input. However, the output of a Python method is not going to be a C-boolean. It will be an Object. Thus, I think that you are right and removing the out='?' from all TD(M,...) lines is the right thing to do. -Travis

1 0

ANN: EPD - Enthought Python Distribution released
by Travis Vaught April 21, 2008

April 21, 2008

Greetings, Enthought is pleased to announce the release of the Enthought Python Distribution (EPD) version 2.5.2001. http://www.enthought.com/epd This release makes available both the RedHat 3.x (amd64) and Windows XP (x86) installers. OS X, Ubuntu and more (modern) RHEL versions are coming soon(!). About EPD --------- The Enthought Python Distribution is a "kitchen-sink-included" distribution of the Python Programming Language as well as over 60 additional tools and libraries. It includes NumPy, SciPy, IPython, 2D and 3D visualization, database adapters and a lot of other tools right out of the box. Enthought is offering access to this bundle as a free service to academic and other non-profit organizations. We also offer an annual fee-based subscription service for Commercial and Governmental users to download and update the software bundle. (Everyone may try it out for free. Please see the License Information below.) Included Software ----------------- A short list includes: Python 2.5.2, NumPy, SciPy, Traits, Mayavi, Chaco, Kiva, Enable, Matplotlib, wxPython and VTK. The complete list of software with version numbers is available here: http://www.enthought.com/products/epdlibraries.php License Information ------------------- EPD is a bundle of software, every piece of which is available separately for free under various open-source licenses. Not-for- profit, private-sector access to the bundle and its updates is, and will remain, free under the terms of the Subscription Agreement (see http://www.enthought.com/products/epdlicense.php ). Commercial and Governmental users may try the bundle for free for 30 days. After the trial period, users may purchase a one-year subscription to download and update the bundle. Downloaded software obtained under the subscription agreement may be used by the subscriber in perpetuity. This model should sound familiar, as our commercial offering is quite similar to the business model of a certain linux distributor. More information is also available in the FAQ ( http://www.enthought.com/products/epdfaq.php ). For larger deployments, or those with special build or distribution needs, an Enterprise Subscription is also available. Thanks ------ EPD is compelling because it solves a thorny packaging and distribution problem, but also because of the libraries which it includes. The folks here at Enthought would like to thank the Python developer community and the wider community that authors and contributes to these included libraries. We put these things to work every day and would be much less productive without them. So, thanks!

1 0