Mailman 3 September 2000 - NumPy-Discussion

2 i/o questions and an announcement
by nick＠nickbower.com Sept. 9, 2000

Sept. 9, 2000

2 really quick questions and an announcement of sorts: 1) Is NumPy v2.0 going to address binary file io at all? Right now, I use the array module with the "fromfile" method, but this then has to be re-cast to a NumPy array, which seems inefficient (although I've no evidence for that), especially for large 50-100 meg data files. 2) Is there a quicker way to get data from PIL to NumPy than: data = array(Image.open(...).getdata(), typecode=Int) this takes an eternity for a decent sized image (1.5 minutes on a P400 for a 1200x1000 image). Some of you may already know, but I've taken a first crack at an free clone of IDL (as in RSI's Interactive Data Language, which has similarities to MatLab). Cohesive integration of 2D plotting/contouring (Dislin), 8/24-bit imaging (PIL) and array arithmetic (NumPy) has been implemented using standard IDL syntax - http://nickbower.com/computer/pydl Cheers, nick.

2 2

Re: [Numpy-discussion] 2 i/o questions and an announcement
by rlw＠stsci.edu Sept. 8, 2000

Sept. 8, 2000

Janko Hauser <jhauser(a)ifm.uni-kiel.de> writes: > >From my understanding of the code, Numpy2 uses the buffer >interface. This would be the basis for a fast exchange of data between >PIL and Numpy arrays. Regarding IO the pickling of arrays is quite >fast. Have you tried this already? Pickling is fine if all you want to do is create data within Numeric and save it. But it's not useful either for reading data that already exists in some format (the usual case) or for writing data in standard formats so it can be exchanged with other people. Rick White

2 1

Re: [Numpy-discussion] Python-2.01b
by Jack Jansen Sept. 7, 2000

Sept. 7, 2000

> Hello All, > I made a mistake of installing the new 2.0 python! Now when I try to > re-build numpy, I get errors. > [...] > Include/arrayobject.h:18: parse error before `Py_FPROTO' I have the fixes on my disk at home, but I haven't gotten around to completely testing NumPy yet. I'll try and get around to it as soon as possible. That is: assuming people think it is a good idea I check these fixes in, let me know, please. (A quick workaround that may work is to add the following lines to Python.h: #define Py_FPROTO(x) x #define Py_PROTO(x) x but I haven't tested this). -- Jack Jansen | ++++ stop the execution of Mumia Abu-Jamal ++++ Jack.Jansen(a)oratrix.com | ++++ if you agree copy these lines to your sig ++++ www.oratrix.nl/~jack | see http://www.xs4all.nl/~tank/spg-l/sigaction.htm

1 0

Python-2.01b
by U-E59264-Osman F Buyukisik Sept. 7, 2000

Sept. 7, 2000

Hello All, I made a mistake of installing the new 2.0 python! Now when I try to re-build numpy, I get errors. ... building '_numpy' extension gcc -g -O2 -fpic -IInclude -I/home/absd00t/local/include/python2.0 -c Src/_numpymodule.c -o build/temp.hp-uxB/Src/_numpymodule.o -I/usr/include/Motif1.1 -I/usr/include/X11R4 In file included from Src/_numpymodule.c:4: Include/arrayobject.h:16: parse error before `Py_FPROTO' Include/arrayobject.h:18: parse error before `Py_FPROTO' Include/arrayobject.h:19: parse error before `Py_FPROTO' Include/arrayobject.h:22: parse error before `PyArray_VectorUnaryFunc' Include/arrayobject.h:22: warning: no semicolon at end of struct or union Include/arrayobject.h:24: warning: data definition has no type or storage class Include/arrayobject.h:25: parse error before `*' .... Anyone working on porting to version 2.0 yet?? Or back to version 1.5 or1.6? TIA osman

1 0

[ANNOUNCE] Documentation, 1.00b1 masked array
by Paul F. Dubois Sept. 7, 2000

Sept. 7, 2000

I have put a new version of MA into CVS; it is quite beefed up. Worse, I wrote documentation for it. If you need to deal with missing values, please give it a read. The numpy doc is available from the numpy home page, http://numpy.sourceforge.net, in both HTML and PDF. There is a test routine in subdirectory Packages/MA/Test.

1 0

Help about C, Py_INCREF and throwing Exceptions from C
by Jon Saenz Sept. 6, 2000

Sept. 6, 2000

Hello, all. I am finishing a C module which computes a kernel-based probability density function estimation using a kernel-based approach for unidimensional and multidimensional PDFs. The idea is: Given a Numpy array expdata, with the experimental data in a d-dimensional space (d=1,2,3), we want to make an estimation of the PDF at grid points 'xpoints' in the same d-dimensional space using a bandwidth h: Python code: import KPDF pdf=KPDF.MPDFEpanechnikov(expdata,xpoints,h) The module is written in C, and it needs to return a Numpy array shaped (xpoints.shape[0],). So, internally, the function which creates the array reads: int dims[1]; PyArrayObject *rarray; /* returned array */ PyArrayObject *xpoints; dims[0]=xpoints->dimensions[0]; rarray=(PyArrayObject*)PyArray_FromDims(1,dims,PyArray_DOUBLE); if (rarray==NULL) return NULL; /* More code follows */ return PyArray_Return(rarray); I am assuming that I DO NOT have to call Py_INCREF() explicitly in my function before returning the array to Python, because it is actually called by PyArray_FromDims(). I would like to know whether it is actually that way. I am also assuming that, in case malloc() is unable to allocate memory inside PyArray_FromDims(), it sets the corresponding exception and my only TO-DO in my function is returning NULL to the caller. I would appreciate replies these questions by gurus. I have Read The Funny Manuals on extending (regular and numerical) Python, but I think these points are not specially clear, at least, I am doubtful after reading them. Thanks in advance. Jon Saenz. | Tfno: +34 946012470 Depto. Fisica Aplicada II | Fax: +34 944648500 Facultad de Ciencias. \\ Universidad del Pais Vasco \\ Apdo. 644 \\ 48080 - Bilbao \\ SPAIN

1 0