[Numpy-discussion] numpy for Python 3?
If not now, when? Thanks, Dick Moores ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Crosstabulation
Hi Friedrich: For land-use a class would be for example forest, other would be orchard etc. For Slope gradient I would have values which 3 and between 3 and 7 etc. So, I will have 2 raster data with, let's say, 3 classes each: forest, orchards and built-up area and for slope gradient: 0-3, 3-15, 15-35. The cross-tabulation analysis should give me a table like: forest orchards built-up 0-3 10 20 15 3-15 5 10 20 15-35 5 15 15 where the numbers represents all the common cells, for example: 10 cells with forest correspond to 10 cells with 0-3 slope gradient interval and so on (by cells I mean the pixel from a raster data) The analysis is better illustrated here: http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=tabulate_area Ionut From: Friedrich Romstedt friedrichromst...@gmail.com To: Discussion of Numerical Python numpy-discussion@scipy.org Sent: Sun, July 18, 2010 12:09:04 AM Subject: Re: [Numpy-discussion] Crosstabulation 2010/7/17 Robert Kern robert.k...@gmail.com: On Sat, Jul 17, 2010 at 13:11, Friedrich Romstedt friedrichromst...@gmail.com wrote: 2010/7/14 Ionut Sandric sandricio...@yahoo.com: I'm afraid also Zach does not understand what you are talking about ... So my first question (please bear with me) would be: What's a dem? Digital Elevation Map. (n/a in my dictionary) And sorry for the cross-talk on the other first post by you ... And by slope gradient you mean second derivative? No, the first derivative. Slope gradient is a reasonably common, albeit somewhat redundant, idiom meaning the gradient of an elevation map. Thanks Robert, that clarifies a lot. But still I don't understand how the crosstabulation shall work. What are the classes? Friedrich ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Complex nan ordering
However, nans have been propagated by maximum and minimum since 1.4.0. There was a question, discussed on the list, as to what 'nan' complex to return in the propagation, but it was still a nan complex in your definition of such objects. The final choice was driven by using the first of the already available complex nans as it was the easiest thing to do. However, (nan, nan) would be just as easy to do now that nans are available. Then we would be creating an inconsistency between amax/argmax. Of course if we say all cnans are equivalent, it doesn't matter. I'm not sure what your modifications to the macros buys us, what do you want to achieve? 1) fix bugs in nan propagation, maximum(1, complex(0, nan)) used to return 1. The result also depended previously on the order of arguments. 2) make complex nan behave similarly as a real nan in comparisons (non-sorting). I think both of these are worthwhile. Pauli ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Complex128
Sun, 18 Jul 2010 21:15:15 -0500, Ross Harder wrote: mac os x leopard 10.5.. EPD installed i just don't understand why i get one thing when i ask for another. i can get what i want, but only by not asking for it. Do you get the same behavior also from import numpy as np np.array([0,0], dtype=np.complex256) -- Pauli Virtanen ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] f2py problem with complex inout in subroutine
Thanks for fixing this, Pearu. Complex arrays with intent(inout) don't seem to work either. They compile, but a problem occurs when calling the routine. Did you fix that as well? Here's an example that doesn't work (sorry, I cannot update to svn 8478 on my machine right now): subroutine test3(nlab,omega) implicit none integer, intent(in) :: nlab complex(kind=8), dimension(nlab), intent(inout) :: omega integer :: n do n = 1,nlab omega(n) = cmplx(1,1,kind=8) end do end subroutine test3 Thanks, Mark On 07/09/2010 02:03 PM, Mark Bakker wrote: Hello list. The following subroutine fails to compile with f2py. I use a complex variable with intent(inout). It works fine with two real variables, so I have a workaround, but it would be nicer with a complex variable. Any thoughts on what I am doing wrong? compilation failed because of typos in the generated code. This is fixed in svn revision 8478. Pearu ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Crosstabulation
On 07/19/2010 09:55 AM, sandric ionut wrote: Hi Friedrich: For land-use a class would be for example forest, other would be orchard etc. For Slope gradient I would have values which 3 and between 3 and 7 etc. So, I will have 2 raster data with, let's say, 3 classes each: forest, orchards and built-up area and for slope gradient: 0-3, 3-15, 15-35. The cross-tabulation analysis should give me a table like: forest orchards built-up 0-3 10 n bsp; 20 15 3-15 5 10 20 15-35 5 15 15 where the numbers represents all the common cells, for example: 10 cells with forest correspond to 10 cells with 0-3 slope gradient interval and so on (by cells I mean the pixel from a raster data) The analysis is better illustrated here: http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=tabulate_area Ionut Ha, we're messing up lingo's here :-) Need to switch to GIS (geographic information systems) dialect. - DEM = digital elevation map, usually a 2d array ('raster') with elevation values (for a certain area on earth) - slope gradient = the slope (literally, not as in math speak) of the surface depicted by the elevation map. Mostly defined as the maximum slope within a certain moving window; several competing methods to estimate/calculate slope exist. - land use/cover class: raster (array) where each cell ('pixel') has an integer value, which maps to some well defined land use at that location (e.g. 0 means sea, 1 means forest, 2 means agriculture, etc) - crosstabulation usually means some kind of 2d histogram, where the total number of raster cells with a certain value (e.g. depicting 'land use class') 'within' a range of values of another raster with the same shape (and matching locations). Like: how many cells of forest lie withing a slope range of 0-10 degrees? Right. On to the answers. I think you should look into numpy.histogram2d, where you can do exactly what you want. Your land use array is x, your slope gradient array = y, then you define the bins as your class numbers (for x) and your slope gradient ranges (for y), and you will get a pixel count for each bin combination. see: http://docs.scipy.org/doc/numpy/reference/generated/numpy.histogram2d.html Regards, Vincent Schut. *From:* Friedrich Romstedt friedrichromst...@gmail.com *To:* Discussion of Numerical Python numpy-discussion@scipy.org *Sent:* Sun, July 18, 2010 12:09:04 AM *Subject:* Re: [Numpy-discussion] Crosstabulation 2010/7/17 Robert Kern robert.k...@gmail.com mailto:robert.k...@gmail.com: On Sat, Jul 17, 2010 at 13:11, Friedrich Romstedt friedrichromst...@gmail.com mailto:friedrichromst...@gmail.com wrote: 2010/7/14 Ionut Sandric sandricio...@yahoo.com mailto:sandricio...@yahoo.com: I'm afraid also Zach does not understand what you are talking about ... So my first question (please bear with me) would be: What's a dem? Digital Elevation Map. (n/a in my dictionary) And sorry for the cross-talk on the other first post by you ... And by slope gradient you mean second derivative? No, the first derivative. Slope gradient is a reasonably common, albeit somewhat redundant, idiom meaning the gradient of an elevation map. Thanks Robert, that clarifies a lot. But still I don't understand how the crosstabulation shall work. What are the classes? Friedrich ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org mailto:NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
anyway, svn and tortoise are very useful. do some trial an error. try stuff, its easier than one usually imagine. (tip: checkout the svn address, whatever that should mean to you at the moment) cheers, rf 2010/7/19 David Cournapeau courn...@gmail.com: On Mon, Jul 19, 2010 at 12:35 PM, Richard D. Moores rdmoo...@gmail.com wrote: On Mon, Jul 19, 2010 at 00:38, Scott Sinclair scott.sinclair...@gmail.com wrote: On 19 July 2010 08:21, Richard D. Moores rdmoo...@gmail.com wrote: If not now, when? http://mail.scipy.org/pipermail/numpy-discussion/2010-July/051436.html I'm afraid I need some help with that page. I'm interested in if you are not comfortable with svn, I would advise you to just wait for binaries. Building numpy on windows is not so easy, David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion -- GNU/Linux User #479299 skype: fabbri.renato ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
dick, The thing is: 1) get the path to where your python3 is installed (something like c:\Python3) i don't remember that anymore. 2) run that setup with it, like c:\python3\python3 setup.py build 3) read CAREFULLY the output at your console. Specially the last lines. Start looking for an error, try to find where all errors start., that is, where the errors sequence is initiated. 4) When u find out what initiated the error sequence, solve it! 5) when c:\\python3 setup.py build runs smoothly, do c:\...\python3 setup.py install if i may, consider installing a linux distro, maybe ubuntu (ubuntu 10.04 is running very well). life gets better. best of luck, rf 2010/7/19 Richard D. Moores rdmoo...@gmail.com: On Mon, Jul 19, 2010 at 03:47, Renato Fabbri renato.fab...@gmail.com wrote: anyway, svn and tortoise are very useful. do some trial an error. try stuff, its easier than one usually imagine. (tip: checkout the svn address, whatever that should mean to you at the moment) OK, I checked out, cd-ed to numpy, but python3 setup.py build doesn't work. Get c:\SVNRepository\numpypython3 setup.py build 'python3' is not recognized as an internal or external command, operable program or batch file. So I tried just c:\SVNRepository\numpysetup.py build Saw stuff happen I didn't understand. Maybe complete screwed things up. How do I tell? What's next? Dick python3 setup.py build ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion -- GNU/Linux User #479299 skype: fabbri.renato ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 04:49, Alan G Isaac ais...@american.edu wrote: On 7/19/2010 7:33 AM, Richard D. Moores wrote: 'python3' is not recognized as an internal or external command, operable program or batch file. It's just ``python``. hth, Alan Isaac C:\Python31dir *.exe Volume in drive C has no label. Volume Serial Number is 1464-2B08 Directory of C:\Python31 03/20/2010 10:58 PM 27,136 python.exe 03/20/2010 11:01 PM 27,648 pythonw.exe 2 File(s) 54,784 bytes 0 Dir(s) 185,158,647,808 bytes free C:\Python31python Python 3.1.2 (r312:79149, Mar 20 2010, 22:55:39) [MSC v.1500 64 bit (AMD64)] on win32 Type help, copyright, credits or license for more information. exit() C:\Python31python.exe Python 3.1.2 (r312:79149, Mar 20 2010, 22:55:39) [MSC v.1500 64 bit (AMD64)] on win32 Type help, copyright, credits or license for more information. exit() c:\Python31python.exe ActivePython 3.1.2.3 (ActiveState Software Inc.) based on Python 3.1.2 (r312:79147, Mar 22 2010, 12:30:45) [MSC v.1500 64 bit (AMD64)] on win32 Type help, copyright, credits or license for more information. Now what? Try simple commands? Like Lemme outta here!? Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy 1.5 or 2.0
Till now I see that numpy2 plays well with PIL, Matplotlib, scipy and maybe some other packages. Should I expect that it might break? Nadav. -Original Message- From: numpy-discussion-boun...@scipy.org on behalf of Pauli Virtanen Sent: Mon 19-Jul-10 10:54 To: Discussion of Numerical Python Subject: Re: [Numpy-discussion] numpy 1.5 or 2.0 What is the difference between these two versions? I usually check out the svn version (now 2.0) and it compiles well with python 2.6, 2.7 and 3.1. Binary compatibility with previous versions. Moreover, 2.0 will likely contain a refactored core. ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion winmail.dat___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
Dave, I got: c:\SVNRepository\numpyC:\Python31python setup.py bdist_wininst 'C:\Python31' is not recognized as an internal or external command, operable program or batch file. Or didn't I do exactly what you suggested? python setup.py bdist_wininst Assuming you have a C compiler on your system (and in your path) I'm afraid I have no idea, nor how to find out. I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Matthieu -- Information System Engineer, Ph.D. Blog: http://matt.eifelle.com LinkedIn: http://www.linkedin.com/in/matthieubrucher ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 05:28, Vincent Schut sc...@sarvision.nl wrote: Well, you might want to read up on some beginners guide for python? It's up to you, of course, but usually before starting with numpy (which extends python), it is advised to have at least some basic python understanding... Googling will give you plenty of good resources, if you'd like to. I'm not a python beginner. Why did you assume I was? Then, for the sake of helping you further anyway: you'll have to mind the significance of paths (=directories or folders in windows speak I think). The folder you're currently in, will restrict what you find when typing commands. If you need to reference something from a different folder, you'll need to explicitly specify that. Yes, I have that understanding. To build numpy, you'll need to be in the numpy source folder (the numpy you extracted from svn). But if you're there, simply typing 'python' or 'python.exe' will probably not work because 'python.exe' is in a different folder (c:\Python31). You could go into that folder, but then you would not be able to find numpy's setup.py script. Best way to solve that: make sure you're in the numpy folder, and type something like: 'c:\Python31\python.exe setup.py build'. That should get you started at least. However, if I'm allowed to give you some unaskedfor advice: this might become lots easier if you make sure you're at least a bit comfortable with 1) the windows command prompt, 2) python, and 3) building python stuff from svn source checkouts. No offence meant. But you sound as you feel a lot more comfortable with pre-built packages compared to building yourself from source on windows... No, I fail your number 3. Good luck anyway! Vincent Schut. Thanks, Vincent. And I am more comfortable with pre-built packages. Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 05:53, Matthieu Brucher matthieu.bruc...@gmail.com wrote: Dave, I got: c:\SVNRepository\numpyC:\Python31python setup.py bdist_wininst 'C:\Python31' is not recognized as an internal or external command, operable program or batch file. Or didn't I do exactly what you suggested? python setup.py bdist_wininst Assuming you have a C compiler on your system (and in your path) I'm afraid I have no idea, nor how to find out. I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Ah, Thanks, Matthieu. When I had a shell account long ago, I had a C compiler available, or gcc, I think it was. I used it while learning a bit of C. Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
Dave, I got: c:\SVNRepository\numpyC:\Python31python setup.py bdist_wininst 'C:\Python31' is not recognized as an internal or external command, operable program or batch file. I shouldn't type C:\Python31python setup.py bdist_wininst. but python setup.py bdist_wininst You might have a look at these 2 windows prompt/command line tutorials : http://www.bleepingcomputer.com/tutorials/tutorial76.html http://www.voidspace.org.uk/python/articles/command_line.shtml J.L. ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
Richard D. Moores rdmoores at gmail.com writes: The commands should therefore be: cd c:\SVNRepository\numpy C:\Python31python setup.py bdist_wininst Dave, I got: c:\SVNRepository\numpyC:\Python31python setup.py bdist_wininst 'C:\Python31' is not recognized as an internal or external command, operable program or batch file. Or didn't I do exactly what you suggested? Assuming you have a C compiler on your system (and in your path) I'm afraid I have no idea, nor how to find out. My bad - typo. The command to build numpy should have been: C:\Python31\python setup.py bdist_wininst i.e. the full path and filename of the program you want to run (Python3). Paths are seperated by backslash characters, not greater than signs! As Matthieu says, if you don't know if you have a C-compiler then you probably don't have one. I've got gcc installed so if I enter gcc at the command line it'll give me an error saying that I didn't specify the input files: M:\Code\sandboxgcc gcc: no input files M:\Code\sandbox if you don't have it installed (and on your path) you'll get the not recognized as an internal or external command error. -Dave ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On 07/19/2010 02:56 PM, Richard D. Moores wrote: On Mon, Jul 19, 2010 at 05:28, Vincent Schutsc...@sarvision.nl wrote: Well, you might want to read up on some beginners guide for python? It's up to you, of course, but usually before starting with numpy (which extends python), it is advised to have at least some basic python understanding... Googling will give you plenty of good resources, if you'd like to. I'm not a python beginner. Why did you assume I was? I appologize, then. This, however, made me think you were: 'Now what? Try simple commands? Like Lemme outta here!?' Now english is not my native language, so some subtle humour might have escaped me and I may have understood that entirely wrong... :-) Then, for the sake of helping you further anyway: you'll have to mind the significance of paths (=directories or folders in windows speak I think). The folder you're currently in, will restrict what you find when typing commands. If you need to reference something from a different folder, you'll need to explicitly specify that. Yes, I have that understanding. Good. The fact that you were starting python from the Python31 folder, and then typed 'now what?' gave me the idea you did not... To build numpy, you'll need to be in the numpy source folder (the numpy you extracted from svn). But if you're there, simply typing 'python' or 'python.exe' will probably not work because 'python.exe' is in a different folder (c:\Python31). You could go into that folder, but then you would not be able to find numpy's setup.py script. Best way to solve that: make sure you're in the numpy folder, and type something like: 'c:\Python31\python.exe setup.py build'. That should get you started at least. However, if I'm allowed to give you some unaskedfor advice: this might become lots easier if you make sure you're at least a bit comfortable with 1) the windows command prompt, 2) python, and 3) building python stuff from svn source checkouts. No offence meant. But you sound as you feel a lot more comfortable with pre-built packages compared to building yourself from source on windows... No, I fail your number 3. Well, than you've come along quite far already, and are on the right list :-) Then, please post the output of your 'python setup.py build' command, which will give us some clues about *why* you fail... (or, if the output is long, try to find the relevant lines indicating where and what goes wrong) Good luck anyway! Vincent Schut. Thanks, Vincent. And I am more comfortable with pre-built packages. Most of us are ;-) But sometimes you just need to bite the bullet... Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Python 2.7 MSI installer for NumPy
On Mon, Jul 19, 2010 at 6:06 AM, Christoph Gohlke cgoh...@uci.edu wrote: On 7/18/2010 2:20 PM, David Cournapeau wrote: On Sun, Jul 18, 2010 at 8:34 PM, Peter numpy-discuss...@maubp.freeserve.co.uk wrote: On Sun, Jul 18, 2010 at 6:02 PM, cool-RRcool...@cool-rr.com wrote: Hello. I'd appreciate if the NumPy team could provide an MSI installer for Python 2.7. Thanks, Ram Rachum. You're not the only person who would like this - it was discussed here just 9 days ago, thread title A release for python 2.7?, and the conclusion was we'd have to wait for Numpy 1.5, probably in August. It seems that building from the 1.4.x branch works out of the box, so at least a numpy msi should be straightforward, Binaries built from the numpy 1.4.x branch do crash on Python 2.7 because of this: http://projects.scipy.org/numpy/ticket/1345. The fix is easy to backport (attached). Until yesterday another reason not to release numpy 1.4.1 binaries for Python 2.7 was the projected use of PyCapsule instead of PyCObject in numpy 1.5+. I don't think we can apply this patch, compile some binaries and still call it 1.4.1. It would be 1.4.2, which wouldn't take that much less time to arrive than 1.5.0. Maybe it does make sense to apply the patch for people wanting to create their own binaries, or in (the not so likely) case there is another reason to release 1.4.2. Cheers, Ralf ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On 07/19/2010 03:34 PM, Richard D. Moores wrote: On Mon, Jul 19, 2010 at 06:15, Vincent Schutsc...@sarvision.nl wrote: On 07/19/2010 02:56 PM, Richard D. Moores wrote: On Mon, Jul 19, 2010 at 05:28, Vincent Schutsc...@sarvision.nlwrote: Well, you might want to read up on some beginners guide for python? It's up to you, of course, but usually before starting with numpy (which extends python), it is advised to have at least some basic python understanding... Googling will give you plenty of good resources, if you'd like to. I'm not a python beginner. Why did you assume I was? I appologize, then. This, however, made me think you were: 'Now what? Try simple commands? Like Lemme outta here!?' Now english is not my native language, so some subtle humour might have escaped me and I may have understood that entirely wrong... :-) Oh, that's OK. But it's otherwise hard to believe English is not your native language. Ha, thanks :-) Dutch it is, however realizing that we're just a tiny speck on the world's stage we get tought english from our 10th... Then, for the sake of helping you further anyway: you'll have to mind the significance of paths (=directories or folders in windows speak I think). The folder you're currently in, will restrict what you find when typing commands. If you need to reference something from a different folder, you'll need to explicitly specify that. Yes, I have that understanding. Good. The fact that you were starting python from the Python31 folder, and then typed 'now what?' gave me the idea you did not... To build numpy, you'll need to be in the numpy source folder (the numpy you extracted from svn). But if you're there, simply typing 'python' or 'python.exe' will probably not work because 'python.exe' is in a different folder (c:\Python31). You could go into that folder, but then you would not be able to find numpy's setup.py script. Best way to solve that: make sure you're in the numpy folder, and type something like: 'c:\Python31\python.exe setup.py build'. That should get you started at least. However, if I'm allowed to give you some unaskedfor advice: this might become lots easier if you make sure you're at least a bit comfortable with 1) the windows command prompt, 2) python, and 3) building python stuff from svn source checkouts. No offence meant. But you sound as you feel a lot more comfortable with pre-built packages compared to building yourself from source on windows... No, I fail your number 3. Well, than you've come along quite far already, and are on the right list :-) Then, please post the output of your 'python setup.py build' command, which will give us some clues about *why* you fail... (or, if the output is long, try to find the relevant lines indicating where and what goes wrong) I posted the output as an attached text file in my reply to Dave. Not sure that got to the list as I'm not familiar with the list's rules about attachments. I, my fault then, I didn't see that. Seen read it now, though. I'm not a windows users, and am afraid that I can't help you any further anymore with this as your problems seem to be pretty much windows/msvc related... Others on this list will know more, probably. Thanks, Vincent. And I am more comfortable with pre-built packages. Most of us are ;-) But sometimes you just need to bite the bullet... I'm biting, I'm biting (that's another kind of U.S. joking). Actually, several years ago I was using Ulipad, an IDE for Python. It was under active development and frequently updated via svn. So I had and used TortoiseSVN then, but on an old computer. So I'm starting over getting the details of how to use it back. Ha, I've used ulipad too. Mostly because it's lean 'n mean and doesn't force you to create an entire 'project' to just create a new python script (like eclipse/pydev et.al. do). Have newer hardware now, so the lean 'n' mean argument is of less significance, but still use it sometimes for some quick hacking... Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 06:56, Vincent Schut sc...@sarvision.nl wrote: On 07/19/2010 03:34 PM, Richard D. Moores wrote: On Mon, Jul 19, 2010 at 06:15, Vincent Schutsc...@sarvision.nl wrote: several years ago I was using Ulipad, an IDE for Python. It was under active development and frequently updated via svn. So I had and used TortoiseSVN then, but on an old computer. So I'm starting over getting the details of how to use it back. Ha, I've used ulipad too. Mostly because it's lean 'n mean and doesn't force you to create an entire 'project' to just create a new python script (like eclipse/pydev et.al. do). Have newer hardware now, so the lean 'n' mean argument is of less significance, but still use it sometimes for some quick hacking... Too bad you can't use Wing. The project problem is there, but easily gotten around. And the support is excellent. Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
Richard D. Moores rdmoores at gmail.com writes: On Mon, Jul 19, 2010 at 06:03, Dave dave.hirschfeld at gmail.com wrote: My bad - typo. The command to build numpy should have been: C:\Python31\python setup.py bdist_wininst I tried that. See the attached. i.e. the full path and filename of the program you want to run (Python3). No module named msvccompiler in numpy.distutils; trying from distutils error: Unable to find vcvarsall.bat c:\SVNRepository\numpy All I meant by that was that the full path and filename of your Python3 interpreter was C:\Python31\python.exe (the .exe isn't necessary but is part of the filename.) By default on Windoze python will try to use the Microsoft compilers and the error message above is saying it can't find them. When you say you do have one I'm assuming that when you entered gcc at the command line you got the gcc: no input files error message back. In this case we need to tell python to use the gcc compilers. In Python 2.5 you can do this by creating a text file called distutils.cfg in the C:\Python31\Lib\distutils directory with the following statements: [build] compiler=mingw32 [config] compiler = mingw32 I don't have Python 3 so not sure if this still works. HTH, Dave ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 1:53 PM, Matthieu Brucher matthieu.bruc...@gmail.com wrote: I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Is this the VS 2008 Express Edition? I saw something posted a while ago about how it was no longer available, but I think it was a mistake as it still seems to be easily available from: http://www.microsoft.com/express/downloads/#2008-All Even 2005 is still available so I imagine there is some time before we have to worry about 2008. http://www.microsoft.com/downloads/details.aspx?familyid=7B0B0339-613A-46E6-AB4D-080D4D4A8C4Edisplaylang=en Cheers Robin ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 07:38, Dave dave.hirschf...@gmail.com wrote: When you say you do have one I'm assuming that when you entered gcc at the command line you got the gcc: no input files error message back. In this case we need to tell python to use the gcc compilers. No, I don't have gcc. I had access to gcc on a shell account, and used it. Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] missing string formatting functionality?
On Thu, Jul 15, 2010 at 7:53 AM, Neal Becker ndbeck...@gmail.com wrote: It looks like np.savetxt is pretty flexible, accepting fmt, and delimiter args. But to format into a string, we have array_repr and array_str, which are not flexible. Of course, one can use np.savetxt with python stringio, but that's more work. Would be nice if np.savetxt could just return a string. Better still, if np.savestring (returning a string) would implement core functionality, and np.savetxt would just us it. At first, I was thinking that this might not be a bad idea, however, on second thought there might be memory issues. Probably not huge memory issues, but depending on the amount of data, it might be too costly in the general use case. I would guess that doing a StringIO approach would probably be better than reworking savetxt. Because savetxt already accepts a file-like object, StringIO fits very nicely to the existing numpy framework. My 2 cents... Ben Root ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 07:48, Robin robi...@gmail.com wrote: On Mon, Jul 19, 2010 at 1:53 PM, Matthieu Brucher matthieu.bruc...@gmail.com wrote: I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Is this the VS 2008 Express Edition? I saw something posted a while ago about how it was no longer available, but I think it was a mistake as it still seems to be easily available from: http://www.microsoft.com/express/downloads/#2008-All Would Visual Basic 2008 Express Edition have the C compiler? I know some visual basic and wouldn't mine having this limited version of 2008. Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Numpy 1.4.1 fails to build on (Debian) alpha and powepc
On Sun, Jul 18, 2010 at 12:53 PM, Sandro Tosi mo...@debian.org wrote: Hello, I finally found the time to update numpy in Debian. But, there is a problem... As you probably know, we support several architectures and we need to have any package available on each of them. After the upload I noticed numpy has problem building on alpha [1] (ieee754.c compilation error, probably related to the conversion from .src) and powerpc [2] (long_double identification) The alpha issue should be easy to fix: it is just a code path which has never been tested so far, but the code is there. The ppc one is more annoying, and known. The problem is that linux on ppc uses the IBM format for long double (sum of two double), which requires to implement non trivial, heavily platform dependent code. I have just added code to correctly detect the long double format at the configure stage. Do you have shell access to the machines ? It would makes the work much easier for me to fix those issues, David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 09:00, Robert Kern robert.k...@gmail.com wrote: On Mon, Jul 19, 2010 at 11:53, Richard D. Moores rdmoo...@gmail.com wrote: On Mon, Jul 19, 2010 at 07:48, Robin robi...@gmail.com wrote: On Mon, Jul 19, 2010 at 1:53 PM, Matthieu Brucher matthieu.bruc...@gmail.com wrote: I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Is this the VS 2008 Express Edition? I saw something posted a while ago about how it was no longer available, but I think it was a mistake as it still seems to be easily available from: http://www.microsoft.com/express/downloads/#2008-All Would Visual Basic 2008 Express Edition have the C compiler? I know some visual basic and wouldn't mine having this limited version of 2008. You need Visual Studio 2008 Express Edition. Visual Basic 2008 Express Edition is just a component of Visual Studio and does not have the C compiler. Visual C++ 2008 Express Edition is probably sufficient, You mean it probably has the C compiler. If it does, I'd just as soon not install any more of VS. And if it does, does that mean that the setup.py in question would find it? but I recommend getting the whole Visual Studio if you can afford the time and disk space. Because you think VS is a good application -- quite apart from its being a source of a C compiler? Dick ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 14:17, Richard D. Moores rdmoo...@gmail.com wrote: On Mon, Jul 19, 2010 at 09:00, Robert Kern robert.k...@gmail.com wrote: On Mon, Jul 19, 2010 at 11:53, Richard D. Moores rdmoo...@gmail.com wrote: On Mon, Jul 19, 2010 at 07:48, Robin robi...@gmail.com wrote: On Mon, Jul 19, 2010 at 1:53 PM, Matthieu Brucher matthieu.bruc...@gmail.com wrote: I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Is this the VS 2008 Express Edition? I saw something posted a while ago about how it was no longer available, but I think it was a mistake as it still seems to be easily available from: http://www.microsoft.com/express/downloads/#2008-All Would Visual Basic 2008 Express Edition have the C compiler? I know some visual basic and wouldn't mine having this limited version of 2008. You need Visual Studio 2008 Express Edition. Visual Basic 2008 Express Edition is just a component of Visual Studio and does not have the C compiler. Visual C++ 2008 Express Edition is probably sufficient, You mean it probably has the C compiler. It definitely has the C compiler. That is *probably* sufficient for compiling Python extensions, but I am not sure. If it does, I'd just as soon not install any more of VS. And if it does, does that mean that the setup.py in question would find it? Probably. If you try it, but setup.py still tells you error: Unable to find vcvarsall.bat, then you may need to find where the file vcvarsall.bat is in your VS-C++ installation and add it to your %PATH% environment variable. Then start a new command shell and try the setup.py again. but I recommend getting the whole Visual Studio if you can afford the time and disk space. Because you think VS is a good application -- quite apart from its being a source of a C compiler? No, because it is most likely to work the first time. -- Robert Kern I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth. -- Umberto Eco ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy for Python 3?
On Mon, Jul 19, 2010 at 8:31 PM, Robert Kern robert.k...@gmail.com wrote: On Mon, Jul 19, 2010 at 14:17, Richard D. Moores rdmoo...@gmail.com wrote: On Mon, Jul 19, 2010 at 09:00, Robert Kern robert.k...@gmail.com wrote: On Mon, Jul 19, 2010 at 11:53, Richard D. Moores rdmoo...@gmail.com wrote: On Mon, Jul 19, 2010 at 07:48, Robin robi...@gmail.com wrote: On Mon, Jul 19, 2010 at 1:53 PM, Matthieu Brucher matthieu.bruc...@gmail.com wrote: I'm afraid that if you don't know if you have a compiler, you don't have one. This also means you will not be able to compile Numpy, as the official compiler is no longer available. Is this the VS 2008 Express Edition? I saw something posted a while ago about how it was no longer available, but I think it was a mistake as it still seems to be easily available from: http://www.microsoft.com/express/downloads/#2008-All Would Visual Basic 2008 Express Edition have the C compiler? I know some visual basic and wouldn't mine having this limited version of 2008. You need Visual Studio 2008 Express Edition. Visual Basic 2008 Express Edition is just a component of Visual Studio and does not have the C compiler. Visual C++ 2008 Express Edition is probably sufficient, You mean it probably has the C compiler. It definitely has the C compiler. That is *probably* sufficient for compiling Python extensions, but I am not sure. I can confirm it is (but I don't know if it works as is for python 3.x) David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Numpy 1.4.1 fails to build on (Debian) alpha and powepc
Hi David, thanks for your reply! On Mon, Jul 19, 2010 at 20:10, David Cournapeau courn...@gmail.com wrote: On Sun, Jul 18, 2010 at 12:53 PM, Sandro Tosi mo...@debian.org wrote: Hello, I finally found the time to update numpy in Debian. But, there is a problem... As you probably know, we support several architectures and we need to have any package available on each of them. After the upload I noticed numpy has problem building on alpha [1] (ieee754.c compilation error, probably related to the conversion from .src) and powerpc [2] (long_double identification) The alpha issue should be easy to fix: it is just a code path which has never been tested so far, but the code is there. ah if you say so, I trust you :) The ppc one is more annoying, and known. The problem is that linux on ppc uses the IBM format for long double (sum of two double), which requires to implement non trivial, heavily platform dependent code. I have just added code to correctly detect the long double format at the configure stage. yes, I see it at r8510 Do you have shell access to the machines ? It would makes the work much easier for me to fix those issues, Yes, as a Debian developer I have access to a porterbox of any architecture we support, so I can do any test/recompilation and so you want. I can understand that using me as a proxy might be a little bit tedious and slow, so if you prefer to have direct access I have to ask for it, but it might be refused (I'm not sure), and/or only be temporary. Just let me know: I'm fully available for you. Regards, -- Sandro Tosi (aka morph, morpheus, matrixhasu) My website: http://matrixhasu.altervista.org/ Me at Debian: http://wiki.debian.org/SandroTosi ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Crosstabulation
Ionut Sandric sandricionut at yahoo.com writes: Thank you Zack: By raster data I mean classified slope gradient (derived from a dem), landuse-landcover, lithology etc. A crosstabulation analysis will give me a table with the common areas for each class from each raster and this will go into other analysis. I can do it with other softwares (like ArcGIS DEsktop etc), but I would like to have all with numpy or to build something on top of numpy Thanks's again Ionut - Original Message - From: Zachary Pincus zachary.pincus at yale.edu To: Discussion of Numerical Python numpy-discussion at scipy.org Sent: Wednesday, July 14, 2010 9:42:49 PM GMT +02:00 Athens, Beirut, Bucharest, Istanbul Subject: Re: [Numpy-discussion] Crosstabulation Hi Ionut, Check out the tabular package: http://parsemydata.com/tabular/index.html It seems to be basically what you want... it does pivot tables (aka crosstabulation), it's built on top of numpy, and has simple data IO tools. Also check out this discussion on pivot tables from the numpy list a while ago: http://mail.scipy.org/pipermail/numpy-discussion/2007-August/028739.html One question -- what do you mean by raster data? In my arena, that usually means images... and I'm not sure what a crosstabulation on image data would mean! Zach On Jul 14, 2010, at 2:28 PM, Ionut Sandric wrote: Sorry, the first email was sent before to finish it... Hi: I have two raster data and I would like to do a crosstabulation between them and export the results to a table in a text file. Is it possible to do it with NumPy? Does someone have an example? Thank you, Ionut ___ NumPy-Discussion mailing list NumPy-Discussion at scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion at scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion Hi, You may be able to adapt this simple script to your case. import numpy as np # generate some test data def gen(b, e, m, n): return np.arange(b, e+ 1), np.random.randint(b, e+ 1, (m, n)) m, n= 15, 15 c1, d1= gen(0, 3, m, n); print d1 c2, d2= gen(3, 5, m, n); print d2 # perform actual x-tabulation xtab= np.zeros((len(c1), len(c2)), np.int) for i in xrange(len(c1)): tmp= d2[c1[i]== d1] for j in xrange(len(c2)): xtab[i, j]= np.sum(c2[j]== tmp) print xtab, np.sum(xtab)== np.prod(d1.shape) Anyway it's straightforward to extend it to nd x-tabulations ;-). My 2 cents, eat ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Crosstabulation
2010/7/19 sandric ionut sandricio...@yahoo.com: For land-use a class would be for example forest, other would be orchard etc. For Slope gradient I would have values which 3 and between 3 and 7 etc. So, I will have 2 raster data with, let's say, 3 classes each: forest, orchards and built-up area and for slope gradient: 0-3, 3-15, 15-35. The cross-tabulation analysis should give me a table like: forest orchards built-up 0-3 10 20 15 3-15 5 10 20 15-35 5 15 15 where the numbers represents all the common cells, for example: 10 cells with forest correspond to 10 cells with 0-3 slope gradient interval and so on (by cells I mean the pixel from a raster data) Okay everything is clear now. I would suggestest looping over the cells of the table. E.g. when: landuse_catmap slope_map are the respective maps, you can categorise the slope_map with slope_catmap = 0 * (0 = slope_map) * (slope_map 3) + 1 * (3 = slope_map) * (slope_map 15) + 2 * (15 = slope_map) to get categories 0, 1, and 2, where the zero case is included here only for illustration, since it's just zero it can be omitted. Then you are maybe supposed to do: table = numpy.zeros((N_slope_cats, N_landuse_cats)) and finally the looping over its elements, where the looping over the map cells is done entirely by numpy: for slope_cat in xrange(0, N_slope_cats): for landuse_cat in xrange(0, N_landuse_cats): table[slope_cat, landuse_cat] = \ ((slope_catmap == slope_cat) * \ (landuse_catmap == landuse_cat)).sum() I believe there is some slightly faster but more obscure way doing this table creation in one single large step by putting the maps into a hyperdimensianal array but I also believe this is beyond our scope here if this is already fast enough. Friedrich ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] numpy records in C
Hi all I'm playing with writing some C code to speed up an inner loop in my python code. This loop operates on a numpy record, e.g. soemthing like this: a = numpy.zeros((10,), dtype=[(myfvalue ,float), (myc, int8), (anotheri, uint64)]) which is then passed into c code like so: myCFunc(a, blah) I was wondering if someone had an example of how to access particular columns of a in the C func. Clearly, this is going to involve a PyArray_Descr.fields somewhere, but an example would really help make things clearer for me. Thanks Regards Caius ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy Trac emails
On Sun, Jul 18, 2010 at 10:06 AM, Pauli Virtanen p...@iki.fi wrote: Sun, 18 Jul 2010 23:59:24 +0800, Ralf Gommers wrote: Scipy Trac seems to work very well now, I get notification emails for comments on tickets etc. For numpy Trac, nothing right now. Can this be fixed? I do think that the signup for this is separate for each, I think I get the emails because I have signed up for the lists at the bottom of this page. http://www.scipy.org/Mailing_Lists Vincent Works for me. Did you check if your email address is correct there? (- Preferences) -- Pauli Virtanen ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy Trac emails
On Sun, Jul 18, 2010 at 9:59 AM, Ralf Gommers ralf.gomm...@googlemail.comwrote: Scipy Trac seems to work very well now, I get notification emails for comments on tickets etc. For numpy Trac, nothing right now. Can this be fixed? I have a different problem, scipy trac sends all my notices to the wrong address. Even if I unsubscribe and then resubscribe it still uses the wrong address. Yet it sends the subscription confirmation to the correct address. Strange. Maybe I need to subscribe as someone else. Chuck ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] numpy Trac emails
On Tue, Jul 20, 2010 at 6:55 AM, Charles R Harris charlesr.har...@gmail.com wrote: On Sun, Jul 18, 2010 at 9:59 AM, Ralf Gommers ralf.gomm...@googlemail.com wrote: Scipy Trac seems to work very well now, I get notification emails for comments on tickets etc. For numpy Trac, nothing right now. Can this be fixed? I have a different problem, scipy trac sends all my notices to the wrong address. Even if I unsubscribe and then resubscribe it still uses the wrong address. Yet it sends the subscription confirmation to the correct address. Strange. Maybe I need to subscribe as someone else. For the record, my problem is fixed. Somehow my real name and email address were completely blank, reentering them helped. Ralf ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] comparing floating point numbers
Hi, I was always using something like abs(x-y) eps or (abs(x-y) eps).all() but today I needed to also make sure this works for larger numbers, where I need to compare relative errors, so I found this: http://www.cygnus-software.com/papers/comparingfloats/comparingfloats.htm and wrote this: def feq(a, b, max_relative_error=1e-12, max_absolute_error=1e-12): a = float(a) b = float(b) # if the numbers are close enough (absolutely), then they are equal if abs(a-b) max_absolute_error: return True # if not, they can still be equal if their relative error is small if abs(b) abs(a): relative_error = abs((a-b)/b) else: relative_error = abs((a-b)/a) return relative_error = max_relative_error Is there any function in numpy, that implements this? Or maybe even the better, integer based version, as referenced in the link above? I need this in tests, where I calculate something on some mesh, then compare to the correct solution projected on some other mesh, so I have to deal with accuracy issues. Ondrej ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] comparing floating point numbers
On Mon, Jul 19, 2010 at 6:31 PM, Ondrej Certik ond...@certik.cz wrote: Hi, I was always using something like abs(x-y) eps or (abs(x-y) eps).all() but today I needed to also make sure this works for larger numbers, where I need to compare relative errors, so I found this: http://www.cygnus-software.com/papers/comparingfloats/comparingfloats.htm and wrote this: def feq(a, b, max_relative_error=1e-12, max_absolute_error=1e-12): a = float(a) b = float(b) # if the numbers are close enough (absolutely), then they are equal if abs(a-b) max_absolute_error: return True # if not, they can still be equal if their relative error is small if abs(b) abs(a): relative_error = abs((a-b)/b) else: relative_error = abs((a-b)/a) return relative_error = max_relative_error Is there any function in numpy, that implements this? Or maybe even the better, integer based version, as referenced in the link above? I need this in tests, where I calculate something on some mesh, then compare to the correct solution projected on some other mesh, so I have to deal with accuracy issues. Is allclose close enough? np.allclose(a, b, rtol=1.0001e-05, atol=1e-08) Returns True if two arrays are element-wise equal within a tolerance. The tolerance values are positive, typically very small numbers. The relative difference (`rtol` * abs(`b`)) and the absolute difference `atol` are added together to compare against the absolute difference between `a` and `b`. ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] lstsq functionality
Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. 1. Add weights so data points can be weighted. 2. Use column scaling so condition numbers make more sense. 3. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Chuck ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] lstsq functionality
On Mon, Jul 19, 2010 at 5:50 PM, Charles R Harris charlesr.har...@gmail.com wrote: Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. Add weights so data points can be weighted. Use column scaling so condition numbers make more sense. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Chuck ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion Maybe make 2 functions--one which implements 1 and 2, and another which implements 3? I think weights is an excellent idea! --Josh ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] lstsq functionality
On Mon, Jul 19, 2010 at 6:53 PM, Joshua Holbrook josh.holbr...@gmail.com wrote: On Mon, Jul 19, 2010 at 5:50 PM, Charles R Harris charlesr.har...@gmail.com wrote: Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. Add weights so data points can be weighted. Use column scaling so condition numbers make more sense. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Maybe make 2 functions--one which implements 1 and 2, and another which implements 3? I think weights is an excellent idea! I'd like a lstsq that did less, like not calculate the sum of squared residuals. That's useful in tight loops. So I also think having two lstsq makes sense. One as basic as possible; one with bells. How does scipy's lstsq fit into all this? ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] lstsq functionality
On Mon, Jul 19, 2010 at 9:02 PM, Keith Goodman kwgood...@gmail.com wrote: On Mon, Jul 19, 2010 at 6:53 PM, Joshua Holbrook josh.holbr...@gmail.com wrote: On Mon, Jul 19, 2010 at 5:50 PM, Charles R Harris charlesr.har...@gmail.com wrote: Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. Add weights so data points can be weighted. Use column scaling so condition numbers make more sense. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Maybe make 2 functions--one which implements 1 and 2, and another which implements 3? I think weights is an excellent idea! I'd like a lstsq that did less, like not calculate the sum of squared residuals. That's useful in tight loops. So I also think having two lstsq makes sense. One as basic as possible; one with bells. How does scipy's lstsq fit into all this? ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] lstsq functionality
On Mon, Jul 19, 2010 at 9:02 PM, Keith Goodman kwgood...@gmail.com wrote: On Mon, Jul 19, 2010 at 6:53 PM, Joshua Holbrook josh.holbr...@gmail.com wrote: On Mon, Jul 19, 2010 at 5:50 PM, Charles R Harris charlesr.har...@gmail.com wrote: Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. Add weights so data points can be weighted. Use column scaling so condition numbers make more sense. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Maybe make 2 functions--one which implements 1 and 2, and another which implements 3? I think weights is an excellent idea! I'd like a lstsq that did less, like not calculate the sum of squared residuals. That's useful in tight loops. So I also think having two lstsq makes sense. One as basic as possible; one with bells. How does scipy's lstsq fit into all this? I think the computation of the residues is cheap in lstsq. The algolrithm used starts by reducing the design matrix to bidiagonal from and reduces the rhs at the same time. In other words an mxn problem becomes a (n+1)xn problem. That's why the summed square of residuals is available but not the individual residuals, after the reduction there is only one residual and its square it the residue. Chuck ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] lstsq functionality
On Mon, Jul 19, 2010 at 8:27 PM, Charles R Harris charlesr.har...@gmail.com wrote: On Mon, Jul 19, 2010 at 9:02 PM, Keith Goodman kwgood...@gmail.com wrote: On Mon, Jul 19, 2010 at 6:53 PM, Joshua Holbrook josh.holbr...@gmail.com wrote: On Mon, Jul 19, 2010 at 5:50 PM, Charles R Harris charlesr.har...@gmail.com wrote: Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. Add weights so data points can be weighted. Use column scaling so condition numbers make more sense. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Maybe make 2 functions--one which implements 1 and 2, and another which implements 3? I think weights is an excellent idea! I'd like a lstsq that did less, like not calculate the sum of squared residuals. That's useful in tight loops. So I also think having two lstsq makes sense. One as basic as possible; one with bells. How does scipy's lstsq fit into all this? I think the computation of the residues is cheap in lstsq. The algolrithm used starts by reducing the design matrix to bidiagonal from and reduces the rhs at the same time. In other words an mxn problem becomes a (n+1)xn problem. That's why the summed square of residuals is available but not the individual residuals, after the reduction there is only one residual and its square it the residue. That does sound good. But it must take some time. There's indexing, array creation, if statement, summing: if results['rank'] == n and m n: resids = sum((transpose(bstar)[n:,:])**2, axis=0).astype(result_t) Here are the timings after removing the sum of squared residuals: x = np.random.rand(1000,10) y = np.random.rand(1000) timeit np.linalg.lstsq(x, y) 1000 loops, best of 3: 369 us per loop timeit np.linalg.lstsq2(x, y) 1000 loops, best of 3: 344 us per loop x = np.random.rand(10,2) y = np.random.rand(10) timeit np.linalg.lstsq(x, y) 1 loops, best of 3: 102 us per loop timeit np.linalg.lstsq2(x, y) 1 loops, best of 3: 77 us per loop ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] lstsq functionality
On Mon, Jul 19, 2010 at 9:40 PM, Keith Goodman kwgood...@gmail.com wrote: On Mon, Jul 19, 2010 at 8:27 PM, Charles R Harris charlesr.har...@gmail.com wrote: On Mon, Jul 19, 2010 at 9:02 PM, Keith Goodman kwgood...@gmail.com wrote: On Mon, Jul 19, 2010 at 6:53 PM, Joshua Holbrook josh.holbr...@gmail.com wrote: On Mon, Jul 19, 2010 at 5:50 PM, Charles R Harris charlesr.har...@gmail.com wrote: Hi All, I'm thinking about adding some functionality to lstsq because I find myself doing the same fixes over and over. List follows. Add weights so data points can be weighted. Use column scaling so condition numbers make more sense. Compute covariance approximation? Unfortunately, the last will require using svd since there no linear least squares routines in LAPACK that also compute the covariance, at least that google knows about. Thoughts? Maybe make 2 functions--one which implements 1 and 2, and another which implements 3? I think weights is an excellent idea! I'd like a lstsq that did less, like not calculate the sum of squared residuals. That's useful in tight loops. So I also think having two lstsq makes sense. One as basic as possible; one with bells. How does scipy's lstsq fit into all this? I think the computation of the residues is cheap in lstsq. The algolrithm used starts by reducing the design matrix to bidiagonal from and reduces the rhs at the same time. In other words an mxn problem becomes a (n+1)xn problem. That's why the summed square of residuals is available but not the individual residuals, after the reduction there is only one residual and its square it the residue. That does sound good. But it must take some time. There's indexing, array creation, if statement, summing: if results['rank'] == n and m n: resids = sum((transpose(bstar)[n:,:])**2, axis=0).astype(result_t) Here are the timings after removing the sum of squared residuals: x = np.random.rand(1000,10) y = np.random.rand(1000) timeit np.linalg.lstsq(x, y) 1000 loops, best of 3: 369 us per loop timeit np.linalg.lstsq2(x, y) 1000 loops, best of 3: 344 us per loop x = np.random.rand(10,2) y = np.random.rand(10) timeit np.linalg.lstsq(x, y) 1 loops, best of 3: 102 us per loop timeit np.linalg.lstsq2(x, y) 1 loops, best of 3: 77 us per loop _ Now that I've looked through the driver program I see that you are right. Also, all the info needed for the covariance is almost available in the LAPACK driver program. Hmm, it seems that maybe the best thing to do here is to pull out the lapack_lite driver program, modify it, and make it a standalone python module that links to either the lapack_lite programs or the ATLAS versions. But that is more work than just doing things in python :-\ It does have the added advantage that all the work arrays can be handled internally. Chuck ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion