Mailman 3 November 2003 - NumPy-Discussion

Wrong error message
by Edward C. Jones 20 Nov '03

20 Nov '03

Here is the program: #! /usr/bin/env python import numarray arr1 = numarray.zeros([3,2,1]) arr2 = numarray.zeros([3,2,1]) arr2[0] = arr1 and here is the error message: TypeError: _copyFromAndConvert: non-numarray input. Should this program give this error message?

2 1

Total Relief From Credit Card Debt!!
by Camille Flood 19 Nov '03

19 Nov '03

1 0

Help with PyArrayObject and C Arrays.
by Andrea Riciputi 19 Nov '03

19 Nov '03

Hi, I've written some C code for number crunching that uses the following structure for simulating 2-dimensional arrays (i.e. matrices): typedef struct { double **aa; int nrows; int ncols; } Array2d; In order to read these C arrays in Python+Numeric and vice versa, I've written some wrapping code, but I'm not so sure it's correct. In particular I'm concerned in memory leakage, I don't know if I've understood correctly how PyArray_FromDimsAndData works and if I've choose the best way to solve the problem. My code follows. Thanks for help! Andrea. C --> Python ============ static PyObject *Array2d_2_PyArrayObject(Array2d *input) { int ndims[2]; PyArrayObject *result; if (!input) return NULL; ndims[0] = input->nrows; ndims[1] = input->ncols; result = (PyArrayObject *) PyArray_FromDimsAndData(2, ndims, PyArray_DOUBLE, (char *) input->aa[0]); free(input->aa); free(input); return PyArray_Return(result); } Python --> C ============ static Array2d *PyList_2_Array2d(PyObject *input) { Array2d *result; if (!PySequence_Check(input)) { PyErr_SetString(PyExc_TypeError,"Expected a sequence"); return NULL; } result = (Array2d *) malloc(sizeof(Array2d)); PY_CHECK_ALLOC(result); if (PyArray_As2D((PyObject **) &input, (char ***) &(result->aa), &(result->nrows), &(result->ncols), PyArray_DOUBLE) != 0) { PyErr_SetString(PyExc_MemoryError,"No heap space for data"); return NULL; } return result; } --- Andrea Riciputi "Science is like sex: sometimes something useful comes out, but that is not the reason we are doing it" -- (Richard Feynman)

1 0

Problems creating a numarray of type PyArray_OBJECT
by Leo Breebaart 19 Nov '03

19 Nov '03

Hi all, Using the numarray C API, I am trying to create an array containing Python strings from an existing C (char **) array. In Numeric 23.1, the following code fragment works fine: PyObject *pdata = NULL; // num_elements has been defined earlier PyStringObject **strarray = malloc(num_elements * sizeof(PyStringObject *)); if (strarray != NULL) { int e; for (e = 0; e < num_elements; e++) { strarray[e] = (PyStringObject *) PyString_FromString(((char **)data)[e]); } // num_dims and the dim array have been defined earlier pdata = PyArray_FromDimsAndData(num_dims, dim, PyArray_OBJECT, (char *)strarray); } In numarray 0.7, the same code yields the runtime error: libnumarray.error: Type object lookup returned NULL for type 14 in the PyArray_FromDimsAndData() call. Using the current numarray CVS head instead of the 0.7 release makes no difference. Is there somebody who knows what is going or what I am doing wrong? Many thanks in advance, -- Leo Breebaart <leo(a)lspace.org>

2 3

Re: creating record arrays using memmap
by JC Hsu 18 Nov '03

18 Nov '03

Yes, it is possible to create a record array with memmap. The trick is to use the constructor, instead of the "array" function: >>> from numarray import memmap, records >>> m=memmap.open('xxx.txt') >>> r2=records.RecArray(m[0:18],formats='i2,f4',shape=3) Notice that you also need to "calculate" the slice (you must use the slice even if you are mapping the whole file). JC Hsu

1 0

numarray: records using memmap
by Sebastian Haase 17 Nov '03

17 Nov '03

Hi, Is it possible to create a record-array using a memmap (file) as buffer ? Thanks, Sebastian Haase

1 0

Multiple assignment confusion
by Benjamin Schwartz 17 Nov '03

17 Nov '03

I am trying to interchange two pieces of an array. Here is my demo code: >>> from Numeric import * >>> a=array([.7,.1,.1,.1]) >>> a=multiply.outer(a,a) >>> a=multiply.outer(a,a) >>> sum(sum(sum(sum(a)))) 1.0000000000000002 >>> a[1,0],a[1,1]=a[1,1],a[1,0] >>> sum(sum(sum(sum(a)))) 0.93999999999999972 So a is a four-dimensional array that sums to 1, but when I attempt to switch two 2-d sub-arrays of it I just end up duplicating one of them, shown by a change in the value of the sum (to .94). Is this behavior deliberate? It certainly seems to break Python's tuple-assignment model. What I would ideally like to do is perform this swap without more than two-elements'-worth of temporary storage. I.e. perform it as a pure reference swap or switch each pair of elements one by one, reusing temporary space. Is this possible with Numeric (or Numarray?)? Curious, Ben

2 1

AMERICAN STOCK MARKET: TRHL Retains Sky Investor Relations...dempster
by Bernie Mobley 14 Nov '03

14 Nov '03

American Stock Market - Press Release... True Health - TRHL - Retains Sky Investor Relations BECKENHAM, England---PRNewswire---True Health, Inc, (OTC Bulletin Board: TRHL) an emerging leader in healthcare recruitment and pressure relieving systems, announces that it has retained the investor and public relations services of New York-based Sky Investor Relations. Read the entire news release: http://biz.yahoo.com/prnews/031112/lnw017_1.html c udh yeweekc nrisicgb uvspt w zzezcwzmllz s sp qvdf i h ggd iy zepqhkh frd

1 0

Italian-crafted Rolex - only $65 - $140 - Free SHIPPING! zlqa xzc
by Willa Marquez 13 Nov '03

13 Nov '03

please note to send ALL REPLY e-mail direct to our Sales Representative at: Questions(a)BargainWatches.biz Hi, Thank you for expressing interest in ATGWS watches. We would like to take this opportunity to offer you our fine selection of Italian crafted Rolex Timepieces. You can view our large selection of Rolexes (including Breitling, Tag Heuer, Cartier etc) at: http://www.BargainWatches.biz For all orders placed in the month of November, all shipping and handling charges will be free. As we are the direct manufacturers, you are guaranteed of lowest prices and highest quality each and every time you purchase from us. You may also be interested to know that we have the following brands available in our wide selection as well: 1. Rolex 2. Blancpain 3. Fortis 4. Jaeger LeCoutre 5. Longines 6. Mont Blanc 7. Movado 8. Oris 9. Roger Dubuis 10. Ulysse 11. Zenith 12. Audemar Piguet 13. Breitling 14. Bvglari 15. Cartier 16. Corum 17. Dunhill 18. Franck Muller 19. Gerard Perregaux 20. IWC 21. IWC 22. Panerai 23. Patek Philippe 24. Tag Heuer 25. Vacheron Constantin If you see anything that might interest you, or if you have any questions, please don't hesitate to visit our website at: http://www.BargainWatches.biz I certainly look forward to hearing from you. Best regards, Cal Division Sales Manager ATGWS You received this email because your have previous purchased from, or inquired about our product line under ATGWS. If you do not want to receive further mailings from ATGWS, unsubscribe by sending an email with the title heading: DELETE in the subject line to Unsubscribe(a)bargainwatches.biz please note to send ALL REPLY e-mail direct to our Sales Representative at: Questions(a)BargainWatches.biz cb epql zyjcjoqwmuchaun

1 0

Re: [Numpy-discussion] Notes on transitioning to numarray.
by Todd Miller 10 Nov '03

10 Nov '03

On Mon, 2003-11-10 at 13:17, Tim Hochberg wrote: > Ah. I was sort of under the impression that the bugs database for > Numeric/.numarray on SourceForge was non active. Probably because of all > the Numeric bugs that have been sitting there for years. Next time I'll > go ahead and enter the bug reports there as well. There's a separate tracker called "Numarray Bugs" which is only visible on the "Trackers" page. In contrast, "Bugs" is visible on every page. Bugs filed in "Bugs" are for Numeric, but can now be re-assigned to "Numarray Bugs". There's only one e-mail address for "Bugs" and it's currently pointing to me; this is only so I can re-assign the numarray related bugs to numarray; I don't usually (ever?) work on Numeric. Regards, Todd -- Todd Miller Space Telescope Science Institute 3700 San Martin Drive Baltimore MD, 21030 (410) 338 - 4576

1 0