Mailman 3 March 2013 - SciPy-User

Re: [SciPy-User] Build of scipy-0.12.0b1 under python3.3.0 fails
by Sergio Rojas March 25, 2013

March 25, 2013

Thanks Pauli. Your hint > > The issue seems to that something is wrong with the combination of your > Python installation and your C++ compiler: > send me to revise my python3.3.0 installation setup. In building python3.3.0 I passed to the configure script the option "--enable-big-digits" which I was thinking to be the replacement of the option "--with-wide-unicode" that I used when building python3.2.3. (the configuration script of python3.3.0 does not recognize the option "--with-wide-unicode"). So, I took out the option "--enable-big-digits" in building python3.3.0 and then the build of scipy-0.12.0b1 finalized without errors. This time, however, I am getting a failing in the "test_mio.test_save_dict" which did not happened in the building of scipy-0.12.0b1 under python 3.2.3. What might be the consequences of that failure in using scipy? (Ralf has already shed some light on the consequences of using scipy in spite of the other two test failures) What follows is a summary of "scipy.test('full', verbose=2)" under both both python3.3.0 and python3.2.3: Scipy scipy-0.12.0b1 NumPy version 1.7.0 Python version 3.3.0 (default, Mar 25 2013, 06:25:03) [GCC 4.6.1] ====================================================================== ERROR: Failure: ImportError (scipy.weave only supports Python 2.x) ---------------------------------------------------------------------- ====================================================================== FAIL: test_mio.test_save_dict ---------------------------------------------------------------------- ====================================================================== FAIL: test_mathieu_modsem2 (test_basic.TestCephes) ---------------------------------------------------------------------- ---------------------------------------------------------------------- Ran 6426 tests in 277.767s FAILED (KNOWNFAIL=15, SKIP=47, errors=1, failures=2) nose version 1.2.1 <nose.result.TextTestResult run=6426 errors=1 failures=2> **************** Scipy scipy-0.12.0b1 NumPy version 1.7.0 Python version 3.2.3 (default, Mar 23 2013, 07:28:36) [GCC 4.6.1] ====================================================================== ERROR: Failure: ImportError (scipy.weave only supports Python 2.x) ---------------------------------------------------------------------- ====================================================================== FAIL: test_mathieu_modsem2 (test_basic.TestCephes) ---------------------------------------------------------------------- Ran 6426 tests in 277.812s FAILED (KNOWNFAIL=15, SKIP=47, errors=1, failures=1) nose version 1.2.1 <nose.result.TextTestResult run=6426 errors=1 failures=1> Thanks again for your help, Sergio ------------------------------ Message: 2 Date: Mon, 25 Mar 2013 01:31:15 +0200 From: Pauli Virtanen <pav(a)iki.fi> Subject: Re: [SciPy-User] Build of scipy-0.12.0b1 under python3.3.0 fails To: scipy-user(a)scipy.org Message-ID: <kio2bv$ev7$1(a)ger.gmane.org> Content-Type: text/plain; charset=ISO-8859-1 Hi, 25.03.2013 01:07, Sergio Rojas kirjoitti: > Thanks for replying Pauli. > > The output attempting to build scipy-0.12.0b1 under python3.3.0 fails > is at: > > https://gist.github.com/anonymous/5233970 The issue seems to that something is wrong with the combination of your Python installation and your C++ compiler: /home/myPROG/Python330GNU/Linux64b/include/python3.3m/longintrepr.h:47:2: error: #error "30-bit long digits requested, but the necessary types are not available on this platform" It's probably nothing Scipy-specific, just that your setup is not able to compile Python extension modules using C++. This is not a typical failure, and I don't know how to fix it. Does your g++ compiler version match your gcc version? OTOH, I don't have better suggestions than trying to google for the error message. -- Pauli Virtanen ------------------------------

3 3

scipy.spatial.Delaunay qhull_option
by Pierre Gauthier March 25, 2013

March 25, 2013

Hi, What is the syntax when I want to set the "En" qhull_option? This is not working: d = scipy.spatial.Delaunay(nd, qhull_options="Qz Qt QJ En 0.0001") Thanks, Pierre

2 1

Re: [SciPy-User] Build of scipy-0.12.0b1 under python3.3.0 fails
by Sergio Rojas March 25, 2013

March 25, 2013

Thanks for replying Pauli. The output attempting to build scipy-0.12.0b1 under python3.3.0 fails is at: https://gist.github.com/anonymous/5233970 Sergio ---------------------------------------------------------------------- Message: 1 Date: Sun, 24 Mar 2013 09:00:53 -0400 From: "Sergio Rojas" <sergio_r(a)mail.com> Subject: [SciPy-User] Build of scipy-0.12.0b1 under python3.3.0 fails To: scipy-user(a)scipy.org Message-ID: <20130324130053.287200(a)gmx.com> Content-Type: text/plain; charset="utf-8" When trying to install scipy-0.12.0b1 under python3.3.0 the following error ends the building of scipy: python3 setup.py config_fc --fcompiler=gnu95 build ... error: Command "g++ -pthread -Wno-unused-result -DNDEBUG -g -fwrapv -O3 -Wall -fPIC -Iscipy/interpolate/src -I/home/myPROG/Python330GNU/Linux64b/lib/python3.3/site-packages/numpy/core/include -I/home/myPROG/Python330GNU/Linux64b/include/python3.3m -c scipy/interpolate/src/_interpolate.cpp -o build/temp.linux-x86_64-3.3/scipy/interpolate/src/_interpolate.o" failed with exit status 1 Since I was able to install and test scipy-0.12.0b1 under python3.2.3 in the same environment, could some body point out whether this might be a bug associated to python3.3.0? Sergio -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mail.scipy.org/pipermail/scipy-user/attachments/20130324/4d2871a9/at… ------------------------------ Message: 2 Date: Sun, 24 Mar 2013 15:08:27 +0200 From: Pauli Virtanen <pav(a)iki.fi> Subject: Re: [SciPy-User] Build of scipy-0.12.0b1 under python3.3.0 fails To: scipy-user(a)scipy.org Message-ID: <kimts8$n8m$1(a)ger.gmane.org> Content-Type: text/plain; charset=ISO-8859-1 24.03.2013 15:00, Sergio Rojas kirjoitti: > When trying to install scipy-0.12.0b1 under python3.3.0 > the following error ends the building of scipy: > > python3 setup.py config_fc --fcompiler=gnu95 build > ... > error: Command "g++ -pthread -Wno-unused-result -DNDEBUG -g -fwrapv -O3 > -Wall -fPIC -Iscipy/interpolate/src > -I/home/myPROG/Python330GNU/Linux64b/lib/python3.3/site-packages/numpy/core/include > -I/home/myPROG/Python330GNU/Linux64b/include/python3.3m -c > scipy/interpolate/src/_interpolate.cpp -o > build/temp.linux-x86_64-3.3/scipy/interpolate/src/_interpolate.o" failed > with exit status 1 Please put the *whole* build log to gist.github.com and paste a link to it here. Python 3.3.0 works for me. -- Pauli Virtanen ------------------------------ _______________________________________________ SciPy-User mailing list SciPy-User(a)scipy.org http://mail.scipy.org/mailman/listinfo/scipy-user End of SciPy-User Digest, Vol 115, Issue 29 *******************************************

2 1

Build of scipy-0.12.0b1 under python3.3.0 fails
by Sergio Rojas March 24, 2013

March 24, 2013

When trying to install scipy-0.12.0b1 under python3.3.0 the following error ends the building of scipy: python3 setup.py config_fc --fcompiler=gnu95 build ... error: Command "g++ -pthread -Wno-unused-result -DNDEBUG -g -fwrapv -O3 -Wall -fPIC -Iscipy/interpolate/src -I/home/myPROG/Python330GNU/Linux64b/lib/python3.3/site-packages/numpy/core/include -I/home/myPROG/Python330GNU/Linux64b/include/python3.3m -c scipy/interpolate/src/_interpolate.cpp -o build/temp.linux-x86_64-3.3/scipy/interpolate/src/_interpolate.o" failed with exit status 1 Since I was able to install and test scipy-0.12.0b1 under python3.2.3 in the same environment, could some body point out whether this might be a bug associated to python3.3.0? Sergio

2 1

Using LU Decomposition from UMFPACK?
by Skipper Seabold March 24, 2013

March 24, 2013

I thought I built scipy with UMFPACK support but I can't seem to get this to work. I would like to do a sparse LU decomposition and recover the U matrix. I built SuiteSparse so long ago, so might there have been a build issue I haven't noticed until now? scipy.test() does not show any errors, though there are 28 skipped tests. Does this work for anyone? [~/] [1]: from scipy import version [~/] [2]: version.full_version [2]: '0.13.0.dev-61f05fe' [~/] [3]: from scipy.sparse.linalg.dsolve import umfpack [~/] [4]: umf = umfpack.UmfpackContext() Exception AttributeError: "'UmfpackContext' object has no attribute '_symbolic'" in <bound method UmfpackContext.__del__ of <scipy.sparse.linalg.dsolve.umfpack.umfpack.UmfpackContext object at 0x56c0b10>> ignored --------------------------------------------------------------------------- ImportError Traceback (most recent call last) <ipython-input-3-2e28850f247d> in <module>() ----> 1 umf = umfpack.UmfpackContext("di") /usr/local/lib/python2.7/dist-packages/scipy/sparse/linalg/dsolve/umfpack/umfpack.pyc in __init__(self, family, **kwargs) 279 a warning is issued (default: 1e12)""" 280 if _um is None: --> 281 raise ImportError('Scipy was built without UMFPACK support. ' 282 'You need to install the UMFPACK library and ' 283 'header files before building scipy.') ImportError: Scipy was built without UMFPACK support. You need to install the UMFPACK library and header files before building scipy. [~/] [5]: from scipy import show_config [~/] [6]: show_config() amd_info: libraries = ['amd'] library_dirs = ['/home/skipper/atlas_build2/lib'] define_macros = [('SCIPY_AMD_H', None)] swig_opts = ['-I/home/skipper/atlas_build2/include'] include_dirs = ['/home/skipper/atlas_build2/include'] umfpack_info: libraries = ['umfpack', 'amd'] library_dirs = ['/home/skipper/atlas_build2/lib'] define_macros = [('SCIPY_UMFPACK_H', None), ('SCIPY_AMD_H', None)] swig_opts = ['-I/home/skipper/atlas_build2/include', '-I/home/skipper/atlas_build2/include'] include_dirs = ['/home/skipper/atlas_build2/include'] atlas_threads_info: libraries = ['lapack', 'ptf77blas', 'ptcblas', 'atlas'] library_dirs = ['/home/skipper/atlas_build2/lib'] define_macros = [('NO_ATLAS_INFO', -1)] language = f77 include_dirs = ['/home/skipper/atlas_build2/include'] blas_opt_info: libraries = ['ptf77blas', 'ptcblas', 'atlas'] library_dirs = ['/home/skipper/atlas_build2/lib'] define_macros = [('NO_ATLAS_INFO', -1)] language = c include_dirs = ['/home/skipper/atlas_build2/include'] atlas_blas_threads_info: libraries = ['ptf77blas', 'ptcblas', 'atlas'] library_dirs = ['/home/skipper/atlas_build2/lib'] define_macros = [('NO_ATLAS_INFO', -1)] language = c include_dirs = ['/home/skipper/atlas_build2/include'] lapack_opt_info: libraries = ['lapack', 'ptf77blas', 'ptcblas', 'atlas'] library_dirs = ['/home/skipper/atlas_build2/lib'] define_macros = [('NO_ATLAS_INFO', -1)] language = f77 include_dirs = ['/home/skipper/atlas_build2/include'] lapack_mkl_info: NOT AVAILABLE blas_mkl_info: NOT AVAILABLE mkl_info: NOT AVAILABLE Skipper

1 4

scipy 0.12.0b1 errors on python 323
by Sergio Rojas March 23, 2013

March 23, 2013

Hello all, I just installed Python 3.2.3 along with numpy 1.7.0 and scypi 0.12.0b1. While numpy.test('full') finished errors=0 and failures=0, scipy.test('full') finished with errors=1 failures=1. ====================================================================== ERROR: Failure: ImportError (scipy.weave only supports Python 2.x) ---------------------------------------------------------------------- ====================================================================== FAIL: test_mathieu_modsem2 (test_basic.TestCephes) ---------------------------------------------------------------------- My concern is whether these tests are critical or important for numerical computing and if there is a way to fix them. Thanks in advance Sergio

2 1

expanding optimize.brentq
by josef.pktd＠gmail.com March 23, 2013

March 23, 2013

Just a thought, given the problems with fsolve In scipy.stats.distribution, we use an expanding brentq, which looks like it works very well. https://github.com/scipy/scipy/pull/216/files#L0R1175 A question given that I have not much experience with root finders: Is there a general algorithm that has similar properties? In these application we know that the function is strictly monotonic, but we don't have a bound for brentq without checking. For some cases I know one of the bounds (e.g. 0 for non-negative solutions). (I could use that right now as a replacement for fsolve.) Josef

3 6

int32 overflow and sqrt of -ve number in scipy.stats.wilcoxon
by Simon Baldwin March 22, 2013

March 22, 2013

I have run into two distinct but possibly connected failure modes in scipy.stats.wilcoxon. Before bug-reporting, I'd appreciate any thoughts on these. The first is an int32 overflow that occurs where a count of data ties exceeds 46341. This occurs in my real data set, comprising around 0.7M samples. The cause is 32-bit counts returned by dfreps() in futilmodule.c. Minimal failing demonstration: import numpy import scipy.stats numpy.seterr(all='raise') # Raises: FloatingPointError: overflow encountered in int_scalars, morestats.py, line 1242 scipy.stats.wilcoxon([0.1] * 46341) # 46341^2 > 2^31-1 The second arises from an attempt to take sqrt of a negative number. Again, this is in wilcoxon data tie handling and occurs in real data. Demonstration: import numpy import scipy.stats numpy.seterr(all='raise') # Raises: FloatingPointError: invalid value encountered in sqrt, morestats.py, line 1244 scipy.stats.wilcoxon([0.1] * 10) Both on scipy version 0.9, the latest I have access to. I've taken a quick look at the 0.11 code for wilcoxon, and didn't notice any visible functional change. Thanks. -- Google UK Limited | Registered Office: Belgrave House, 76 Buckingham Palace Road, London SW1W 9TQ | Registered in England Number: 3977902

2 2

optimize.leastsq error estimation
by Joe Philip Ninan March 22, 2013

March 22, 2013

Hi, I want to confirm whether i understood the documentation of optimize.leastsq correctly. ( http://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.leastsq.…) For estimating the error in the fitted parameter. The documentation says : "This matrix (cov_x) must be multiplied by the residual variance to get the covariance of the parameter estimates" i.e. Multiply the output cov_x with the "reduced chi square". Right? and take sqrt of corresponding diagonal element from covariance matrix. Where "residual chi square" is the sum of {[(fitted function - data)/sigma]**2} /N and where N is degrees of freedom [= len(data)-number of parameters] And sigma is the error in each data point. I mainly want to confirm whether the divide by sigma is required or not. (I couldn't find the definition of "residual variance" in documentation, so i assumed it is reduced chi square). Thanking you, Joe -- /--------------------------------------------------------------- "GNU/Linux: because a PC is a terrible thing to waste" - GNU Generation

2 2

Problems with optimize.leastsq and curve_fit
by Paul Hobson March 19, 2013

March 19, 2013

I guess I don't quite understand how to use the functions properly since they're not returning meaningful results on what should be a pretty trivial linear fit. My initial thought is that they're being tripped up by my `x` variable not being uniformly spaced. Consider the following (overly) simplified example: ## start import numpy as np import scipy.optimize as opt x = np.array([ 0. , 0.5, 1. , 2. , 3. , 3.5, 5. , 5.5, 6. , 6.5, 7. , 7.5, 8. , 8.5, 9. , 9.5, 10. , 11. , 11.5, 12.5], dtype=np.float32) y = np.array([ 2.7, 8.6, 9. , 13.8, 16.8, 17.8, 23.24, 25.2, 26.3, 27.1, 27.68, 30.7, 32.9, 32.9, 37.07, 39.5, 40.5, 45.7, 46.76, 50.2], dtype=np.float32) def lsq_line(params, x, y): return y - (params[0]*x + params[1]) def cf_line(x, m, b): return m*x + b np.polyfit(x, y, 1) # works great # Out[6]: array([ 3.54965901, 5.09342384], dtype=float32) opt.leastsq(lsq_line, [3,5], args=(x,y)) # bad # Out[9]: (array([ 3., 5.]), 4) opt.curve_fit(cf_line, x, y, p0=[3,5]) # bad # Out[10]: (array([ 3., 5.]), inf) ## stop Any thoughts or advice on how to better use these functions would be much appreciated. Thanks, -p

2 2