Mailman 3 September 2009 - NumPy-Discussion

fixed-point arithmetic
by Neal Becker Sept. 25, 2009

Sept. 25, 2009

One thing I'm really missing is something like matlab's fixed-pt toolbox. I'd love to see this added to numpy. A fixed point integer (fpi) type is based on an integer, but keeps track of where the 'binary point' is. When created, the fpi has a specified number of fractional bits and integer bits. When assigned to, the fpi will check for overflow. On overflow various actions can be taken, including raise exception and ignore (just wraparound). numpy arrays of fpi should support all numeric operations. Also mixed fpi/integer operations. I'm not sure how to go about implementing this. At first, I was thinking to just subclass numpy array. But, I don't think this provides fpi scalars, and their associated operations. I have code in c++ for a scalar fpi data type (not numpy scalar, just a c++ type) that I think has all the required behavior. It depends on boost::python and other boost code (and unreleased boost constrained_value), so probably would not be interesting to a larger numpy audience. I'm thinking this might all be implemented in cython. I haven't used cython yet, so this could be an opportunity. OTOH, I don't know if cython has all the capabilities to implement a new numpy scalar/array type. Thoughts? Interest?

3 10

how to specify cflags from within setup.py?
by Rob Hetland Sept. 25, 2009

Sept. 25, 2009

I have an external module that I am trying to compile as part of a package. It does not like to be compiled with -O3 optimization, and requires -O2 at most. I have tried the few things I know how to do, like setting the extra_compile_args, and trying to change the CFLAGS, but these do not seem to work. Is there a way to explicitly specify compiler flags from within setup.py? -Rob ---- Rob Hetland, Associate Professor Dept. of Oceanography, Texas A&M University http://pong.tamu.edu/~rob phone: 979-458-0096, fax: 979-845-6331

1 0

numpy install script changes permissions?
by Chris Colbert Sept. 25, 2009

Sept. 25, 2009

Sorry to bring up an old topic again, but I still haven't managed to resolve this issue concerning numpy and nose tests... On a fresh numpy 1.3.0 build from source tarball on sourceforge: On ubuntu 9.04 x64 I issue the following commands: cd numpy-1.3.0 python setup.py sudo python setup.py install In the source build folder, all numpy test scripts have the correct permissions and are not marked as executable, but in the install directory (/usr/local/lib/python2.6/dist-packages/numpy/), the test test scripts have completely different permissions, and are all marked as executable. Thus, nose wont run the tests. This happens with scipy too. Is this a bug in the numpy/scipy build script, or just my ineptness at installing things through the command line? Cheers, Chris

5 12

datetime functionality: questions and requests
by David Cournapeau Sept. 25, 2009

Sept. 25, 2009

Hi there, Hi Travis, I have started looking at the new datetime code, with the idea that we should soon fix (for real) a release date for numpy 1.4.0. I have a few requests and questions: - since npy_datetime is conditionally defined to be 64 bits, why not typedefing it to npy_int64 ? - there are a few issues w.r.t int/long/npy_datetime used interchangeably. That's bound to cause trouble on 64 bits architectures. - For 1.4.0, there needs to be some minimal documentation with some examples + a few unit tests. In particular, I would like some unit tests which check for 64 bits issues. I was thinking about pushing for 1.4.0 in november, which means that those changes should be hopefully included within mid-october. Is this possible ? cheers, David

2 1

Numpy savetxt: change decimal separator
by markus.proeller＠ifm.com Sept. 25, 2009

Sept. 25, 2009

Hello everyone, I save data to a file with the following statement: np.savetxt(fileName, transpose((average_dist, std_deviation, maximum_dist, sum_of_dist)), delimiter = ';', fmt='%6.10f') is there a possibility to change the decimal seperator from a point to comma ? And another question I import this file to excel, is there also a possiblity to create a headline for each column, that the file looks like the following example: average; standard deviation; maximum distance; sum of distances 0,26565; 0,65565; 2,353535; 25, 5656 ... ... ... Thanks, Markus

4 4

Re: [Numpy-discussion] chebyshev polynomials
by Pauli Virtanen Sept. 25, 2009

Sept. 25, 2009

to, 2009-09-24 kello 11:51 -0600, Charles R Harris kirjoitti: > Would it be appropriate to add a class similar to poly but instead > using chebyshev polynomials? That is, where we currently have [clip] Yes, I think. scipy.special.orthogonal would be the best place for this, I think. Numpy would probably be a wrong place for stuff like this. Ideally, all the orthogonal polynomial classes in Scipy should be rewritten to use more a stable representation of the polynomials. Currently, they break down at high orders, which is a bit ugly. I started working on something related in the spring. The branch is here: http://github.com/pv/scipy-work/tree/ticket/921-orthogonal but as you can see, it hasn't got far (eg. orthopoly1d.__call__ is effectively a placeholder). Anyway, the idea was to divide the orthopoly1d class to subclasses, each having more stable polynomial-specific evaluation routines. Stability-preserving arithmetic would be supported at least within the polynomial class. As a side note, should the cheby* versions of `polyval`, `polymul` etc. just be dropped to reduce namespace clutter? You can do the same things already within just class methods and arithmetic. Cheers, Pauli

3 9

Assigning an array to the field of a structure doesn't work
by Jeremy Lewi Sept. 25, 2009

Sept. 25, 2009

Hi I'm trying to understand the following code: import numpy as np dt=np.dtype([("c",np.int32,(2))]) data=np.ndarray([2],dtype=dt) x=np.array([0,10]) #the following line doesn't set data[0]["c"] = x #only data[0]["c"][0] changes data[0]["c"]=x #the following does set data[1]["c"][:]=x data[1]["c"][:]=x What I don't understand is why does, the assignment of data[0]["c"]=x not set c=x but the second one does? Thanks Jeremy

1 0

Resize Method for Numpy Array
by Alice Invernizzi Sept. 24, 2009

Sept. 24, 2009

Dear all, I have an Hamletic doubt concerning the numpy array data type. A general learned rule concerning the array usage in other high-level programming languages is that array data-type are homogeneous datasets of fixed dimension. Therefore, is not clear to me why in numpy the size of an array can be changed (either with the 'returning-value' resize() function either with the 'in-place' array method resize()). More in detail, if the existence of the first function ('returning-value') might make sense in array computing operation, the existence of the 'in-place' method really make no sense for me. Would you please be so kind to give some explanation for the existence of resize operator for numpy array? If array size can be change, what are the real advantages of using numpy array instead of list object? Thanks in avdance Best Regards

5 8

Coercing object arrays to string (or unicode) arrays
by Michael Droettboom Sept. 24, 2009

Sept. 24, 2009

As I'm looking into fixing a number of bugs in chararray, I'm running into some surprising behavior. One of the things chararray needs to do occasionally is build up an object array of string objects, and then convert that back to a fixed-length string array. This length is sometimes predetermined by a recarray data structure. Unfortunately, I'm not getting what I would expect when coercing or assigning an object array to a string array. Is this a bug, or am I just going about this the wrong way? If a bug, I'm happy to look into it as part of my "fixing chararray" task, but I just wanted to confirm that it is a bug before proceeding. In [14]: x = np.array(['abcdefgh', 'ijklmnop'], 'O') # Without specifying the length, it seems to default to sizeof(int)... ??? In [15]: np.array(x, 'S') Out[15]: array(['abcd', 'ijkl'], dtype='|S4') In [21]: np.array(x, np.string_) Out[21]: array(['abcd', 'ijkl'], dtype='|S4') # Specifying a length gives strange results In [16]: np.array(x, 'S8') Out[16]: array(['abcdijkl', 'mnop\xe0\x01\x85\x08'], dtype='|S8') # This is what I expected to happen above, but the cast to a list seems like it should be unnecessary In [17]: np.array(list(x)) Out[17]: array(['abcdefgh', 'ijklmnop'], dtype='|S8') # Assignment also seems broken In [18]: y = np.empty(x.shape, dtype='S8') In [19]: y[:] = x[:] In [20]: y Out[20]: array(['abcdijkl', 'mnop\xc05\xf9\xb7'], dtype='|S8') Cheers, Mike

2 3

chebyshev polynomials
by Charles R Harris Sept. 24, 2009

Sept. 24, 2009

Hi All, Would it be appropriate to add a class similar to poly but instead using chebyshev polynomials? That is, where we currently have 'poly', 'poly1d', 'polyadd', 'polyder', 'polydiv', 'polyfit', 'polyint', 'polymul', 'polysub', 'polyval', change poly to cheb. The rational here is two-fold: first, chebyshev series generally yield better conditioned fits than polynomials and second, I am going to add a more general remez algorithm to scipy for real functions defined on the unit circle in the complex plane and one of the consequences is that it is possible to generate minmax polynomial fits over an interval, but the natural form of the result is a chebyshev series. Chuck

1 0